Sector: Automotive, Hospitality & Services, Transportation
Function: Sales, Operations, Distribution & Logistics, Procurement/Sourcing & Supply Chain
Type: AI, Agentic

How Gen AI can help

Pricing optimization
A pricing agent can continuously learn the price elasticity of each product and track competitor prices, adjusting in real time to capture revenue opportunities, avoid unnecessary markdowns, and react to changing market conditions.

Inventory management
An inventory agent can monitor stock levels across stores and warehouses, factoring in lead times and supply constraints to align replenishment decisions with projected demand and pricing strategies.

Demand forecasting
A demand forecasting agent can analyze signals from internal sales trends, online search patterns, social media, weather forecasts, and local events to anticipate surges or dips in near-term demand.

Promotions and bundling
A promotions agent can design targeted offers and product bundles (e.g., pairing slow-moving items with high-demand products), scheduling them based on real-time sales velocity and inventory.

Collaborative decision-making
All agents in the process share a common situational awareness and negotiate trade-offs. For example, if the demand agent forecasts a surge, the pricing agent might raise prices while the promotions agent delays discounts conversely, if oversupply is detected, price reductions and targeted promotions might be implemented in specific regions or channels.

Managing risk and promoting trust

Fair and impartial
Because frequent price changes can be perceived as unfair or arbitrary, dynamic pricing agents should operate within clearly defined policies and thresholds to ensure consistent treatment of customers across channels and regions.

Robust and reliable
Bad data can lead to bad decisions. Agents should be designed with strong data validation and filtering processes to avoid reacting to false signals (such as misinterpreted social trends or inaccurate sales figures).

Transparent and explainable
Dynamic price and promotion changes can confuse customers and internal teams alike. To help address the problem, agents should provide clear reasoning for adjustments, including the data sources and logic used, so pricing and category managers can interpret and communicate the rationale.

Responsible and accountable
Rapid pricing and inventory actions can have strategic and reputational impacts. AI-driven decisions should align with the organization’s brand strategy, operating capacity, and regulatory requirements, with final oversight provided by qualified human managers.

Function: R&D/Product Development, Marketing, Manufacturing & Quality, Procurement/Sourcing & Supply Chain
Type: AI, Agentic, GenAI

How AI can help

Market sensing and opportunity identification

A market sensing agent can analyze real-time data from trend reports, social media, consumer sentiment, and competitive intelligence to identify unmet needs and emerging product opportunities.

Concept generation and feasibility analysis

A concept agent can create diverse and innovative product ideas informed by market insights while a feasibility agent evaluates each idea against sourcing options, production cost, manufacturing timelines, and regulatory constraints.

Design development and prototyping

A design agent can produce technical specifications and digital prototypes, enabling rapid iteration and deep product visualization without the time and cost needed to create physical samples.

Validation and dynamic iteration

A validation agent can test designs against historical performance, customer feedback, and simulated market conditions, while a coordination agent can orchestrate updates across product lifecycle management, marketing, and supply chain systems to adjust plans in real time.

Managing risk and promoting trust

Robust and reliable

Because AI agents might produce concepts that are technically or commercially unviable, outputs should be validated in controlled simulation environments and reviewed through human-in-the-loop processes before advancing to production.

Transparent and explainable

Since design decisions can have major cost, brand, and regulatory implications, agents should provide clear reasoning and evidence for their recommendations, including source data and assumptions.

Responsible and accountable

Products must comply with safety and regulatory requirements. Also, legal questions remain over intellectual property (IP) rights for AI-generated outputs. IP protection and ownership rights can be complex when AI is involved in the creative process. To address such issues, AI-driven design activities should align with brand standards and legal constraints, with final approvals retained by qualified human decision-makers.

Function: Operations, Customer Experience, Distribution & Logistics, Compliance & Risk, Field Services
Type: AI, Agentic

How AI can help

Continuous store sensing
A store sensing agent can monitor real-time data streams from cameras, IoT sensors, POS systems, and digital twins to track foot traffic, queue lengths, inventory levels, associate availability, and local events.

Dynamic task allocation
A task management agent can reprioritize and assigns tasks such as restocking, returns processing, online order pickup, or promotional setup based on current demand and available labor.

Automated compliance monitoring
A compliance agent can use computer vision and sensor data to monitor planogram adherence, promotion execution, and safety hazards, triggering immediate corrective actions as needed.

Coordinated multi-agent oversight
A store manager agent can oversee all other agents, resolving conflicts, optimizing labor deployment, and coordinating with upstream systems such as workforce management, ERP, and order management platforms.

Managing risk and promoting trust

Robust and reliable
Because retail environments often involve incomplete or questionable sensor data, agents should be designed to function effectively under imperfect conditions, with human escalation protocols in place when AI output is uncertain or fails.

Transparent and explainable
To avoid blind reliance on AI agents, managers and staff should have access to clear explanations of the rationale behind AI-driven decisions.

Responsible and accountable
Store operations must comply with a wide range of standards, including labor laws, safety requirements, and company policies. All AI-driven actions should align with these standards, with ultimate accountability retained by human supervisors.

Function: Procurement/Sourcing & Supply Chain, Operations, Distribution & Logistics, Compliance & Risk
Type: AI, Agentic

How AI can help

Data readiness and transformation
A data readiness agent can perform quality checks and identify exceptions, while a data generator agent can transform raw inputs into structured data for optimization.

Optimization and demand mapping
A suggestion optimization agent can run AI/ML models to autonomously identify the best-performing options, while a demand mapping agent can align demand signals with the correct product configurations.

Validation and explainability
A validation/explanation agent can review outputs, ensure consistency, and provide transparent reasoning to supply chain managers for greater trust in the system’s recommendations.

Managing risk and promoting trust

Robust and reliable
Inaccurate recommendations can disrupt production or logistics. AI agents should be validated regularly against real-world outcomes and monitored to provide ongoing reliability.

Transparent and explainable
Because supply chain managers need to understand how recommendations are generated, agents should be designed to provide clear explanations of their optimization logic and how the underlying data was used.

Responsible and accountable
Given the risk of AI agents taking inappropriate or inconsistent action, humans should have the final responsibility for supply chain adjustments.

Sector: Automotive
Function: Compliance & Risk, Customer Service, Operations
Type: AI, Agentic

How AI can help

Data validation and fraud detection
AI agents can review incoming claims for completeness, identify unusual patterns, and flag potential fraud, waste, or abuse before the claim progresses.

Customer and claims history analysis
Agents can cross-check claims with customer and vehicle history, uncovering relevant information that strengthens the adjudication process.

Documentation and reporting
An agent can automatically generate detailed reports for human adjudicators, reducing manual work and providing decision-makers with the necessary context to make informed decisions.

Decision support and denial drafting
Agents can propose denial reasons with clear justifications and then draft denial letters for human approval, helping adjudicators operate more efficiently and improving consistency across claims.

Managing risk and promoting trust

Robust and reliable
Errors in root cause analysis or improper routing can lead to incorrect claim outcomes. Actions of AI agents should be validated against historical claim datasets and continuously monitored for performance accuracy.

Transparent and explainable
Warranty decisions can have a big impact on customers, dealers, and regulatory compliance. As such, AI agents should provide clear reasoning for why a claim was flagged, routed, or recommended for denial.

Responsible and accountable
Given the significant financial and reputational risks associated with warranty adjudication, final decisions should be left to human adjudicators, with AI agents providing decision support rather than operating as fully autonomous systems.

Sector: Automotive
Function: Sales, Customer Experience, Marketing, Information Technology
Type: AI, Agentic

How AI can help

Personalized vehicle matching
A central vehicle search and advisor agent helps customers identify models that align with their preferences, budgets, and usage needs, whether buying, leasing, or exploring certified pre-owned (CPO) options.

Comprehensive financial analysis
A buy agent analyzes total cost of ownership—including loan payments, depreciation, maintenance, and taxes—while a lease agent evaluates lease terms, monthly payments, and conditions to help customers compare financing options with full transparency.

Inventory and production visibility
An OEM agent analyzes vehicles in production pipelines and offers booking options, helping OEMs capture demand even when current dealer inventory does not meet customer criteria.

Streamlined communication and support
A communication agent delivers supporting documents, sends summaries, and ensures smooth integration with dealership systems, reducing customer effort and follow-up calls.

Managing risk and promoting trust

Transparent and explainable
Because purchase and lease decisions involve major financial commitments, agents need to provide clear explanations of cost breakdowns, assumptions, and trade-offs in their recommendations.

Robust and reliable
Errors in inventory matching or financial analysis can erode customer trust. AI agents should be validated against real-world dealership and OEM data and continuously updated to ensure accuracy.

Responsible and accountable
AI agents can have a significant influence on consumers’ car-buying decisions. As such, their outputs should be positioned as guidance tools, with customers and dealer staff retaining final responsibility for understanding and confirming selections.

Function: Operations, Sales, Marketing, Procurement/Sourcing & Supply Chain, Cross-functional/Enterprise-wide, Finance
Type: GenAI, AI

How AI can help

Real-time consolidation
AI can consolidate real-time inputs from sales, inventory, marketing trends, and supply chain metrics to produce dynamic forecasts.

Sophisticated analysis
AI enables trend recognition, historical pattern analysis, and early alerting on supply-demand gaps, while also facilitating scenario planning and pricing strategy refinement all through a unified dashboard.

Actionable insights
The system can continuously update itself as new data flows in, signaling demand shifts or regional product affinities and providing decision makers with actionable insights.

Managing risk and promoting trust

Robust and reliable
Given the system’s critical business impact, resilience is key. AI models should be capable of updating in real time and integrating multiple data streams reliably and accurately. Extensive pilot testing can help fine-tune model accuracy before scaling.

Transparent and explainable
A user-facing dashboard that clearly shows inputs, trends, and recommendations can help business leaders understand how forecasts are generated, what assumptions are at play, and what real-world data is influencing outputs—reducing blind reliance on the system and promoting human-AI collaboration.

Safe and secure
To help mitigate security breaches and operational disruptions, robust security protocols should be embedded in both the technology infrastructure and data flows, with IT overseeing access controls, data encryption, and integration with existing ERP systems.

Function: Marketing, Customer Experience
Type: GenAI, AI

How AI can help

Detecting and analyzing trends and events
AI can help detect and analyze influencer trends and brand affinity across a wide range of social media platforms 24/7, identifying opportunities to shape consumer expectations in real time. Retrieval-augmented generation (RAG) capabilities provide real-time access to social data, such as trending hashtags, viral video clips, and current events.

Generating multimodal creative content
AI offers the ability to autonomously generate creative content across modalities, while remaining contextually and culturally aware. Key capabilities include: (1) LLMs for generating social media copy, product descriptions, captions, and hashtags; (2) multimodal image models for visual asset generation, including pack shots, brand imagery, and marketing visuals; and (3) short-form video generation.

Model-agnostic orchestration
Content creation tasks can be dynamically routed to the most cost-effective or best-performing AI models, optimizing output while reducing compute costs.

Managing risk and promoting trust

Robust and reliable
Retrieval-augmented generation can reduce hallucinations and improve model performance over time. Fallback models and safety nets can mitigate failures or inappropriate content generation under unpredictable conditions.

Fair and impartial
The content generation pipeline should be evaluated regularly for potential cultural, social, or representational biases. Human oversight can ensure that outputs reflect brand values.

Private
No personal user data should be used in the generation process; models should be trained and tuned on anonymized or public datasets. Data residency and usage should comply with regional regulations, including the EU AI Act.

Function: Marketing, Customer Experience, Sales
Type: GenAI, AI

How AI can help

Next-gen content generation
With AI, the enterprise can create product descriptions, imagery, video, and more much faster and more consistently than with existing tools and processes.

Personalization at scale
AI models can draw from multimodal data (e.g., text, image, geospatial data) to create personalized and contextually relevant content. The model can be used to catalog content and adapt content and user flow based on language, region, and customer behavior trends.Assisting complianceDue to the consistency AI enables across modes, languages, and contextual factors, the enterprise can enhance regulatory compliance for materials across different geographies, cultures, and topics.

Managing risk and promoting trust

Robust and reliable
While tasked with producing superior marketing materials, AI systems may invent inaccuracies, which will lead to poorer customer engagement and outcomes.

Fair and impartial
Biases in the data (e.g., due to incomplete datasets) could lead to unequal quality of content in the face of different geographical or cultural factors.

Function: Sales, Marketing, Operations, Procurement/Sourcing & Supply Chain
Type: GenAI, AI

How AI can help

Supporting employees
AI can be used to prepare negotiation materials by combing through older campaigns or deals, sorting the relevant information, and generating suggestions. This helps equip the human employee with materials like pre-works (e.g., consolidated material from prior years) and pitch decks, supporting their negotiations.

Predicting outcomes
AI can help optimize trade shelf spacing and investment allocation by predicting outcomes and conducting scenario building and storytelling. It can also be used to build scenarios with cultural customizations for negotiation processes with retailers.

Optimization support
With AI, users can rapidly analyze EPOS data and transactional information to provide insights that help optimize the design of promotional programs, setting the right price points, promotion mechanics, and anticipating sales uplift to inform production processes of the expected demand.

Managing risk and promoting trust

Safe and secure
Because price, margin information, and negotiation strategies are consumed by the model, it must be secured to prevent the leakage of sensitive commercial data.

Fair and impartial
The data used to train and fuel the model may be dated, leaving new target groups and small but growing customer segments potentially underrepresented. As a result of this latent bias, the model may be challenged to provide commensurate accuracy for all groups and segments.

Function: Cross-functional/Enterprise-wide, Information Technology
Type: GenAI, AI

How AI can help

Greater access to insights
An AI system can help stakeholders across all business functions better understand the consumer by simplifying data mining and analysis with user-friendly interfaces and natural language queries. This allows users to ask questions relevant to their work and extract actionable insights without compromising functionality.

Bringing down data barriers
The system can aggregate data from various sources and domains (e.g., purchasing patterns, customer service, website and browsing data, marketing campaign response) to provide comprehensive insights into consumer behaviors. Reaching across data silos, the system can automatically identify outliers and summarize issues to guide decision makers to areas requiring attention.

Managing risk and promoting trust

Safe and secure
The AI model is exposed to sensitive and proprietary enterprise data, which creates a risk of potential data leakage. To mitigate this risk, the enterprise may look at restricting data access to the AI provider, as well as carefully determining what consumer data should be exposed to the model.

Robust and reliable
For business users to make confident decisions informed by AI, they need to be able to trust the outputs. To this end, data inputs must be accurate and up to date, and outputs should be validated and monitored.

Transparent and explainable
Business users require sufficient context to interpret consumer data, and while analysis conducted by a data expert inherently contains a level of "human in the loop," when using an AI model, business users need the capacity to understand context and outputs.

Function: Customer Experience, R&D/Product Development
Type: GenAI, AI

How AI can help

Accurate style transferring
By analyzing images or videos of the customer and the desired style, AI can create realistic representations of how the clothing or product would look in the real world.

Virtual mix-and-match
AI allows customers to more easily explore a wider range of style options, clothing combinations, and accessories.

Greater personalization
By considering factors such as body shape, skin tone, and personal style, AI can suggest suitable products that align with the customer’s preferences.

Managing risk and promoting trust

Private
By working with and augmenting consumer photos and videos, the model is exposed to sensitive or personally identifiable information, which is subject to privacy regulations and standards. Leveraging AI for style transferring requires the enterprise to ensure user data is safely stored, transferred, and used.

Transparent and explainable
When consumers input an image of themselves or their surroundings, they need to understand how that media is used by the enterprise, how consumer-machine interactions are tracked and recorded, and whether there are any privacy risks to the consumer when using the style transferring application.

Fair and impartial
If the training set is unbalanced and therefore biased, renderings for virtual try-ons may be more accurate or realistic for one demographic group over another, potentially impacting customer satisfaction and regulatory compliance.

Function: Information Technology
Type: GenAI, AI

How AI can help

Offloading lower-level work
AI can augment the completion of repetitive tasks, such as the deployment and maintenance of code across different platforms (e.g., iOS, Android, webapps).A developer assistantAI can be used in the development of the code itself, serving as an assistant supporting software developers in writing and maintaining code. It can also promote consistency across platforms and applications, such as by converting functional code to different environments.

Managing risk and promoting trust

Safe and secure
Code created with AI may include vulnerabilities that may be difficult to identify during development and even after deployment. Given the importance of cybersecurity, enterprises need to ensure generated code does not introduce security risks.

Robust and reliable
AI is susceptible to errors, and when using it for development tasks, human validation is necessary to mitigate the risk of bugs or vulnerabilities in code as it is created and maintained for multiple applications.

Function: Customer Service, Customer Experience
Type: GenAI, AI

How AI can help

A conversational agent
AI can enable new ways of engaging with customers, using speech-to-text and natural language inputs to generate empathetic and personalized conversations for aftersales support and handling customer complaints.

Better use of human capital
Because generative AI can provide instant, personalized responses to customer queries, offer relevant solutions, and engage in conversations, customers can gain faster response and resolution, and organizations can free up human agents to focus on more complex customer issues.

Managing risk and promoting trust

Robust and reliable
The quality and accuracy of customer interactions impact the customer experience and brand impression. If an AI-enabled customer assistant fails to provide accurate and personalized advice or product instructions, it could degrade (rather than enhance) the quality of the customer interaction.

Transparent and explainable
Customers should have the opportunity to gain a clear understanding of what the model can and cannot do. Also, to promote transparency and positive engagements, enterprises should set customer expectations for the virtual assistant.

Function: Sales, Customer Experience, Marketing
Type: GenAI, AI

How AI can help

Hyper-personalized recommendations
Based on customer input and preferences, AI can generate tailored recommendations, making the buying process more personalized and convenient. In addition, the interactive and iterative approach to product recommendations that AI enables can yield more targeted suggestions than current search engine capabilities.

Image as input/output
Consumers can enter an image of preferred styles (e.g., a celebrity in a designer outfit), and the AI model can identify products and make recommendations based on the image.

Managing risk and promoting trust

Fair and impartial
Latent bias in training and testing data may lead the model to express a preference toward some products or product combinations when making recommendations. Ongoing monitoring, data updates, and human validation can contribute to continuous improvement and bias mitigation.

Private
The model may be exposed to customer data throughout the course of an interaction, and that personal information may be subject to regulatory protections. Important considerations include how the customer data is stored, transferred, and used, as well as how the data is consumed and used by the model itself.

Function: Marketing, R&D/Product Development, Compliance & Risk, Finance
Type: GenAI, AI

How AI can help

Market intelligence
AI can help simulate market scenarios, generate synthetic data to fill data gaps, predict customer preferences based on existing patterns, offer cross-cultural insights, aid in competitor analysis, suggest compliance strategies, optimize market entry costs, simulate brand perception scenarios, and provide demand forecasting to reduce uncertainties.

Information synthesis
AI enables rapid market research by efficiently reading and summarizing extensive volumes of pertinent material, presenting the information in a readily understandable format for market research teams.

Novel market segmentation
AI-generated data may reveal new and previously unidentified market segments within the target market. This can open up additional opportunities for niche marketing and product customization.

Richer personas
Rather than relying on basic surveys and focus groups for understanding consumer likes and dislikes, AI can identify specific customer preferences and create detailed profiles. Using AI, market research teams can even create fictional yet plausible customer personas based on the market’s unique characteristics, helping the company better understand their potential customers’ behavior and preferences.

Managing risk and promoting trust

Fair and impartial
AI models may learn from biased datasets, leading to biased outputs that do not accurately represent the actual market.

Robust and reliable
Given AI’s potential to hallucinate and produce inaccurate outputs, AI-generated insights should be verified with real-world data and traditional research methods to ensure accuracy and reliability.

Responsible and accountable
While AI can complement market research, it should not replace traditional research entirely, as it may miss qualitative nuances and human expertise.

Transparent and explainable
To trust AI outputs, users require the ability to understand which samples and research methods were used to generate recommendations and insights.

Sector: Industry Products & Construction
Function: Manufacturing & Quality, Operations, Compliance & Risk, Finance
Type: AI, Agentic

How AI can help

Continuous sensor monitoring and anomaly detection
AI agents can analyze vibration, temperature, pressure, and other IoT sensor data in real time, leveraging predictive analytics to flag deviations from baseline performance and predicting impending failures.

Root cause analysis and diagnosis
When anomalies are detected, specialized agents can assess historical failure logs, maintenance records, and environmental conditions to pinpoint likely failure modes.

Automated work order generation and scheduling
Other agents can generate detailed work orders and schedule tasks based on production cycles, resource availability, and cost constraints.Simulation and reinforcement learningMulti-agent reinforcement learning systems can simulate inspection intervals and failure scenarios to reduce maintenance expenses and downtime.

Human-centric integration and continuous improvement
AI agents can collaborate with human maintenance teams: presenting findings in clear, natural language, validating outcomes, helping to prioritize alerts and recommend next steps, and refining models over time based on new data and outcomes.

Managing risk and promoting trust

Robust and reliable
Predictive accuracy should be validated across varied asset types and environmental conditions. Also, AI agents should be tested against historical failure cases and simulated breakdown scenarios, with fallback mechanisms for human review in uncertain situations.

Transparent and explainable
Explainable AI outputs improve adoption and help build trust. Agents need to provide transparent reasoning for their recommendations and actions (e.g., “Vibration on bearing exceeds historical threshold during peak load,” or “Leaf spring failure consistent with past incidents”), supported by traceable data sources.

Safe and secure
Industrial systems are vulnerable to cyber threats, and the consequences of a breach can be severe. Agent platforms should include intrusion detection, secure communication with edge devices, and resilience against malicious sensor tampering or spoofing.

Responsible and accountable
Although AI agents can provide valuable decision support, ultimately human technicians and maintenance managers are responsible for critical decisions and actions. As such, clear escalation protocols need to be established for ambiguous or high-risk alerts.

Function: Field Services, Compliance & Risk, Operations
Type: AI, Agentic

How AI can help

Automated flight and mission control
Drones launch autonomously from preprogrammed docks, fly designated routes, capture high-resolution imagery, and then return for charging and data offload—all without human pilots.

AI-driven defect detection
Onboard analytics can automatically process visual, thermal, or LiDAR data to highlight problems such as cracks, corrosion, vegetation encroachment, and structural anomalies.

Agentic inspection orchestration
An orchestration agent can oversee fleets of drones, schedule inspection missions dynamically, monitor inspection results in near-real time, and then automatically trigger follow-up maintenance workflows when anomalies are detected. This greatly reduces the need for manual coordination.

Extended-range and precision navigation
Advanced AI systems now support beyond-visual-line-of-sight (BVLOS) operations over long distances using perception-aware controllers to maintain accurate positioning and avoid obstacles.

Managing risk and promoting trust

Robust and reliable
Inaccurate or inadequate inspections can lead to safety risks and failures. Drone systems need to be validated against varied situations and adverse environments, not just routine test missions and historical defect cases. Also, escalation paths should be established for manual review of ambiguous findings.

Transparent and explainable
AI should provide tangible evidence (such as high-resolution images, heat maps, or annotated photos) accompanied by clear explanations that enable maintenance teams to confidently verify issues.

Private
Some drone inspections may inadvertently capture sensitive data. Agents need to enforce data minimization and restricted-access policies to protect private or proprietary information during data collection and transfer.

Safe and secure
Since drones operate near power lines and potentially dangerous industrial equipment, systems must include breach-resistant communication, secure command chains, and rigorous safeguards against signal spoofing or unauthorized control.

Function: Field Services, Compliance & Risk, Operations
Type: AI, Agentic

How AI can help

Task management and response automation
AI agents can identify operational issues such as inspection gaps, maintenance alerts, and compliance breaches, while also providing real-time support to field teams, bridging the gap between manual and autonomous operations. They can generate task lists, assign them to the right field crews, and follow up on execution, reducing coordination delays and helping to avoid missed maintenance cycles.

Context-aware adaptation
Agents can respond in real time to changes such as weather shifts, unexpected hazards, and shifting regulatory requirements. They can also reschedule tasks, redirect resources, and escalate safety alerts as needed.

Specialist collaboration
Multi-agent AI teams can mimic expert field units. For example, a maintenance agent can diagnose faults, a scheduling agent can reroute technicians, and a compliance agent can confirm actions comply with regulatory requirements. By working together, AI agents can provide specialized yet cohesive action.

Continuous learning and feedback
Agents can learn from experiences such as safety incidents, unexpected site constraints, and real-world repair times. This feedback loop improves their task allocation and decision logic, boosting long-term effectiveness and responsiveness.

Managing risk and promoting trust

Robust and reliable
Field operations take place in unpredictable environments. Agents should be tested against variable and adverse conditions, such as network outages and natural disasters. Fallback paths should include human intervention for ambiguous cases.

Transparent and explainable
Technicians, supervisors, and regulators need to understand AI decisions. Agents should provide clear rationales such as “rescheduled inspection due to stage-1 weather alert,” backed by task logs and field data.

Safe and secure
Safety in field operations is of paramount importance. As such, agentic systems must include cybersecurity safeguards such as encrypted communications, device authentication, anomaly detection, and protection against unauthorized commands.

Responsible and accountable
Governance structures should define escalation paths, sign-off mechanisms, and oversight protocols for AI-generated plans. Although agents can provide useful support, ultimate responsibility for critical decisions and actions should remain with human field supervisors and operations managers.

Sector: Energy & Chemicals, Industry Products & Construction
Function: Sales, Marketing, Operations
Type: AI, Agentic

How AI can help

Pricing and bid preparation agents
A pricing agent analyzes input costs, market trends, competitor rates, and regulatory constraints. A bid preparation agent then drafts customized proposals based on customer histories, relevant contract terms, and compliance guidelines.

Demand forecasting and scenario modeling
A demand agent pulls in real-time signals—such as weather forecasts, energy prices, and consumption patterns—then simulates demand under different conditions to inform pricing and bid decisions.

Customer engagement coordination
An orchestration agent coordinates the bid, pricing, and demand agents to ensure cohesive action. It sequences tasks, manages version control, and forwards AI output to human commercial teams for review and execution.

Contract rollout and monitoring
After the sale, specialized agents monitor contract performance and market deviations, flagging margin erosion that needs to be addressed and triggering renewals when favorable conditions arise.

Managing risk and promoting trust

Robust and reliable
Pricing errors can reduce margins and trigger regulatory issues. Agents must be tested on historical scenarios and stress-tested for market shocks.

Fair and impartial
Agents must avoid bias toward certain customers or regions. Pricing models should be audited to ensure fairness, and human commercial teams should have the ability to override automated suggestions.

Safe and secure
Given the sensitivity of pricing data and customer value, agentic systems must include secure authentication, intrusion detection, and validation of external inputs to guard against manipulation or cyber interference.

Responsible and accountable
Escalation protocols should be in place for high-value or sensitive proposals, with humans retaining final responsibility for commercial offers and contract decisions.

Sector: Energy & Chemicals
Function: R&D/Product Development, Manufacturing & Quality
Type: GenAI, AI

How AI can help

Streamline experimental process
Using AI to determine the most efficient experimental procedures for probing or optimizing materials can streamline the experimental stages of development by removing redundant experiments and undertaking those that are cost- and time-optimized.

High-entropy alloy (HEA) engineering
raditional techniques for developing HEAs with excellent physical, chemical, and mechanical properties are time-consuming and costly, making AI modelling a promising alternative development pathway.

Managing risk and promoting trust

Safe and secure
Intellectual property or a similar competitive advantage could be compromised by using AI in materials design, as models trained on proprietary or sensitive data could potentially reveal valuable insights or design strategies to competitors.

Responsible and accountable
Companies should be mindful to identify and mitigate unintended negative ramifications of materials designed with the support of AI, such as long-term environmental impacts from materials that cannot be manufactured in responsible and sustainable ways.

Sector: Energy & Chemicals
Function: Manufacturing & Quality, Operations
Type: GenAI, AI

How AI can help

Ore characterization and mapping
AI models can be trained on large datasets of mineral samples to generate synthetic samples that mimic the characteristics of real-world ores. Comprehensive databases can be built for mineral identification, classification, and prediction of ore properties, permitting insights into the behavior and composition of different ores without testing on known processing assays.

Process optimization
Models that simulate the physical and chemical processes involved in mineral processing can help optimize factors like grinding parameters, flotation conditions, and separation techniques. This can improve efficiency, reduce energy consumption, and enhance mineral recovery rates.

Managing risk and promoting trust

Robust and reliable
AI models may struggle to generalize mineral samples and processing scenarios that are significantly different from the training data. The model might not capture the full range of variations and unique characteristics of novel ores, which could lead to suboptimal processing recommendations. Also, if AI models cannot interpret complicated physical and chemical qualities like particle size distribution, mineral composition, and processing conditions, the model may generate suboptimal strategies or overlook critical factors.

Sector: Industry Products & Construction
Function: R&D/Product Development, Operations
Type: GenAI, AI

How AI can help

Automated layout generation
Designers can use AI to analyze site constraints, design requirements, and input from engineers to quickly generate layout options for site plans that consider factors such as zoning regulations, operational use, and user preferences.

Design optimization
AI can help optimize site plans by analyzing parameters like solar orientation, traffic flow, and accessibility to suggest optimal infrastructure placements. This can help improve energy efficiency, support better space utilization, and enhance the user experience.

Efficient documentation and annotation
By analyzing design elements and structures in the generated plans, AI can automatically annotate the plans with relevant information, such as dimensions, materials, and specifications. This automation could save designers considerable time and effort, allowing them to focus on higher level design tasks.

Managing risk and promoting trust

Responsible and accountable
AI for design optimization may focus primarily on efficiencies, such as cost reduction or time savings, while potentially neglecting other important considerations, such as environmental sustainability, community impact, or long-term adaptability. The model should be configured to balance multiple objectives and prioritize trade-offs to achieve better overall outcomes. Using AI for site planning also raises legal considerations around intellectual property, ownership of AI-generated designs, liability for design flaws, and privacy restrictions for sensitive or proprietary data.

Sector: Industry Products & Construction
Function: Learning & Development, Compliance & Risk, Operations
Type: GenAI, AI

How AI can help

Virtual reality (VR) training
Combined with VR, AI can be used to develop virtual training environments that replicate operational conditions. With realistic scenarios that simulate OHS incidents, trainees can navigate hazardous situations, identify risks, and improve their OHS awareness and response capabilities in a safe setting.Customized training contentAI can be used to customize training materials based on specific job roles, site conditions, or regulatory requirements. This technology can analyze large volumes of data, such as incident reports, OHS guidelines, or compliance standards and generate tailored content, including videos, interactive modules, or quizzes.

Managing risk and promoting trust

Safe and secure
Real-life emergencies can be highly stressful and traumatic, and replicating these scenarios virtually could imperil the psychological safety of trainees. The final design of simulations should be reviewed by human trainers to remove potentially harmful visualizations.

Responsible and accountable
AI-generated training materials should be continuously monitored to identify any potential issues, inaccuracies, or outdated information. Regular updates to the training content should be made to reflect the latest safety guidelines, regulations, and best practices.

Fair and impartial
AI-generated training materials should be designed to be accessible to all types of learners, including individuals with disabilities. Organizations should consider providing closed captions for videos, adjustable training scenarios to accommodate different skill levels, and alternative formats for content.

Sector: Energy & Chemicals
Function: R&D/Product Development, Operations
Type: GenAI, AI

How AI can help

Seismic data analysis
To overcome incomplete, low volume, or poor-quality seismic data, AI can support enhanced data analysis and interpretation. AI can be used to generate new data samples that resemble the patterns and characteristics of the existing seismic data, address missing or incomplete seismic data, improve data quality through denoising or resolution enhancement, and more effectively interpret complex data patterns.

Reservoir characterization
By analyzing data sources such as well logs, core samples, and production data, AI can create models that simulate the more complete behaviors of hydrocarbon reservoirs. This enables a better understanding of the reservoir dynamics, which helps optimize production strategies and improve recovery rates.

Managing risk and promoting trust

Robust and reliable
False positives or misinterpretations may result in costly and time-consuming drilling operations that do not yield productive reservoirs, making human expertise crucial to validating insights and decision-making. Also, AI models could overlook critical factors or geological nuances that human geoscientists would recognize and could fail to contextualize the data when generating outputs. Without a contextual understanding, AI models and interpretations may lack accuracy or fail to capture the full complexity of reservoirs.

Sector: Industry Products & Construction
Function: Field Services, Compliance & Risk, Operations
Type: GenAI, AI

How AI can help

Smart summaries
Combined with computer vision solutions, AI can create smart assistive summaries in natural language from thousands of hours of drone footage. Assistive smart summaries can be based on a pre-determined template requested by the user, where observations are generated about elevations, topology, lighting, vegetation, and other factors. Summaries can also be queried in natural language so questions can be asked without the assessor manually reviewing all footage.

Querying the footage
When using OGI to detect leaks, there may be instances where a leak is irreparable but still must be managed. With AI, specific sites can be efficiently reviewed and monitored using simple natural language queries.

Managing risk and promoting trust

Robust and reliable
AI models may struggle to interpret environmental indicators, assess ecological impacts, or consider local conditions and regulations. Training data availability and quality in particular can impact the AI model’s ability to generalize and handle diverse environmental scenarios. Inadequate or biased training data may result in limited or skewed analysis and summaries.

Private
Drone footage may contain sensitive information, including personally identifiable information, facial images, or confidential business information, and the footage may also be captured on private properties or areas with restricted access. In using AI to analyze and summarize the footage, unsecure data handling and access can raise privacy concerns as well as legal and regulatory implications.

Sector: Industry Products & Construction
Function: Procurement/Sourcing & Supply Chain, Operations, Compliance & Risk
Type: GenAI, AI

How AI can help

Supply chain intelligence
AI can help identify and simulate potential disruptions or risks in the supply chain. By assessing port congestion, shipment routes, and tier-n supplier mapping, AI can be used to predict risks and their corresponding impact on operations, then recommend actions to mitigate those risks. This allows supply chain managers to proactively implement mitigation strategies, develop contingency plans, and improve overall resilience.

Scenario analysis and optimization
Supply chain managers can use AI to run what-if scenarios in a digital twin environment that reflects the real-world supply chain. By simulating the impact of changes in demand patterns, production capacity, inventory strategies or supplier reliability, supply chain managers can improve risk assessments and proactive decision-making based on real-time conditions.

Supply chain planning
AI enables supply chain professionals to use natural language to interact with advanced planning solutions. Questions concerning all supply chain areas, such as planning, inventory, supply assurance, order management, and global logistics, can be easily asked, helping even less experienced users navigate complex topics and data.

Supplier assessment
AI can assist in supplier evaluation and relationship management by analyzing financial reports, performance metrics, customer feedback, and other data and then generate insights and predictions around supplier performance, risk factors, and opportunities for collaboration. This helps supply chain professionals make informed decisions when selecting, negotiating with, and managing suppliers.

Managing risk and promoting trust

Robust and reliable
Supply chain management involves complex trade-offs, strategic considerations, and tacit knowledge that AI models may not fully capture. AI outputs may also fail to balance ethical considerations or long-term strategic goals. As such, human judgment and validation is central to the interpretation and augmentation of AI outputs.

Fair and impartial
When using AI for supplier evaluation, negotiating, and contracting, bias in the data or model could lead to unfair recommendations and discriminatory practices. By taking into account factors such as fair contract terms, social responsibility, and ethical sourcing practices, organizations can promote decision-making processes that are fair and transparent.

Function: Operations, Customer Experience, Compliance & Risk
Type: GenAI, AI

How AI can help

Promote informed customer behavior
Energy companies can incentivize consumers to adjust their energy consumption based on their specific energy use patterns using conversational chatbots powered by generative AI. AI models can analyze historical data and customer preferences to recommend personalized strategies to reduce energy usage. When there is an immediate need to reduce peak loads to improve grid stability, AI applications can be used to alert customers about what they can specifically do to help. What is more, conversational chatbots can be used as an educational tool for consumers to understand and optimize their energy usage.

Document and map digitization
AI can be used to digitize documentation, infrastructure maps, and records of energy use, as well as for image-to-image translation or image restoration (e.g., by removing noise, adjusting brightness, and enhancing contrast). This improves the quality of the documents and yields searchable documents that can be used to train existing AI classification and forecasting tools.

Grid layout and expansion
AI can assist in designing optimal configuration and expansion plans for the energy grid. AI models can generate optimized grid designs that minimize transmission losses and maximize efficiency by considering factors such as population density, existing infrastructure, and energy demand projections.

Energy trading and market analysis
AI models can simulate the behavior of electricity markets under different scenarios, such as regulation changes or the introduction of new technologies. This can help energy companies optimize their trading strategies and make more informed investment decisions.

Managing risk and promoting trust

Private
Using AI in customer behavior analysis and chatbot interaction involves handling sensitive customer data. Risks include data breaches and unauthorized access to customer information and chat logs, and risk mitigation requires robust security measures, customer data protection, and adherence to privacy regulations.

Safe and secure
AI models are vulnerable to adversarial attacks, where malicious actors manipulate inputs to deceive or exploit the system, for example, to influence energy trading decisions or disrupt grid operations. Robust security measures and regular testing are necessary to mitigate such risks.

Sector: Banking & Capital Markets
Function: Compliance & Risk, Operations, Information Technology, Finance
Type: AI, Agentic

How AI can help

Risk detection and monitoring
Agents that specialize in different risk areas (such as credit card fraud, anti-money laundering (AML), or cybersecurity) can automatically detect and monitor risks.

Regulatory change monitoring
Other specialized agents can parse new regulations using natural language processing, interpret rules and guidelines, map new requirements against company controls to identify compliance gaps, and then draft reports or suggest control changes for humans to implement.

Stress testing and scenario analysis
A stress testing agent can simulate scenarios such as interest rate jumps or credit crunches, coordinating with agents representing different departments (e.g., loans, trading, deposits) to assess institutional safety.

Managing risk and promoting trust

Robust and reliable
Financial firms face massive penalties and reputational damage for compliance failures. As such, multi-agent systems should be validated against historical fraud cases, past audits, and synthetic stress scenarios to ensure their outputs are dependable and comprehensive.

Safe and secure
Pulling data from various databases to feed AI agents requires significant IT work and raises potential security concerns. Given the sensitive nature of risk management data, agentic systems must be secured against unauthorized access or accidental disclosure of private customer information.

Transparent and explainable
Risk management is overseen by humans who are personally liable and may be wary of trusting “black box” agents. Explainability is key. Agents need to provide clear audit trails (for example, explaining why a transaction was flagged, and including references to rules or past cases). This clarity makes AI outputs easier for human officers and regulators to trust and validate.

Fair and impartial
Biases in data sources or training models can result in unfair or inaccurate risk assessments, raising concerns for customer trust, brand reputation, and regulatory compliance.

Sector: Investment management
Function: Customer Experience, Sales, Compliance & Risk
Type: AI, Agentic

How AI can help

Market monitoring
A market scout agent can track real-time developments such as interest rate changes, stock prices, and macroeconomic indicators to identify relevant opportunities and risks.

Personal financial analysis
A client profile analyst agent can aggregate financial data (e.g., bank accounts, credit card spending, loans) and behavioral patterns (e.g., risk-taking versus saving) to build a detailed and dynamic personal profile for each client.

Planning and strategy generation
A planning agent can synthesize market and client inputs to recommend bespoke financial strategies—for example, rebalancing investment portfolios in response to market changes or personal life events such as a home purchase or retirement.

Compliance integration
A compliance agent can help ensure recommendations adhere to fiduciary standards and regulatory guidelines, making the AI outputs safe and trustworthy for both clients and advisors.

Continuous fine-tuning
Unlike static robo-advisors, multi-agent AI systems can provide adaptive guidance, refining recommendations as circumstances evolve, rather than offering static one-time plans. They can identify portfolio risks in real time, flag anomalies for review, and automate activities such as threshold-based rebalancing, factor-based investing, and tax loss harvesting.

Managing risk and promoting trust

Robust and reliable
Because errors could cause significant client losses, agents should be validated against historical financial data and tested under varied market and life-event scenarios to ensure they generate sound, resilient advice.

Transparent and explainable
AI-driven decisions—such as rebalancing actions or investment recommendations—should be accompanied by clear rationales and audit trails, enabling accountability and regulatory review.

Fair and impartial
Advisory models should be monitored to ensure they do not steer clients toward biased product selections, specific asset classes, or vendor-preferred solutions. Advice must be applied equitably across client segments.

Private
Because AI agents handle highly sensitive financial data, strong safeguards must be in place to protect information, including anonymization and strict access controls.

Sector: Banking & Capital Markets
Function: Operations, Information Technology, R&D/Product Development
Type: AI, Agentic

How AI can help

Market simulation
In artificial market simulations, agents representing different trader archetypes (e.g., retail investors, institutional traders, market makers) can enable researchers, investment firms, and regulators to observe emergent phenomena and test different strategies and scenarios before they go live.

Specialized trading agents
A firm can simultaneously deploy live agents that apply different strategies in the market—such as short-term arbitrage, medium-term trend following, or options hedging—sharing signals to avoid conflicts and improving portfolio resilience.

Coordinated action
Agents can operate semi-autonomously under the supervision of a coordinator agent or human risk manager, ensuring portfolio alignment and preventing overexposure. In some cases, different AI agents might even collude or negotiate with each other to create a competitive advantage.

Reinforcement learning and communication
Advanced systems can use multi-agent reinforcement learning, where agents learn from each other through trial and error in simulated environments. Some of these systems might leverage LLMs to enable agents to communicate and explain their reasoning, making the simulation results easier to interpret.

Managing risk and promoting trust

Robust and reliable
Because unpredictable agent interactions can create market instability, systems should be stress-tested against diverse scenarios, including flash-crash conditions, to ensure agents behave reliably under pressure.

Fair and impartial
Agents should be monitored to prevent behaviors that could unfairly favor certain counterparties or drift into market manipulation.

Sector: Banking & Capital Markets
Function: Operations, Compliance & Risk, Customer Experience
Type: AI, Agentic

How AI can help

Multisource data aggregation
A data gathering agent can compile structured and unstructured information (e.g., bank statements, tax filings, e-commerce history, location signals) to build an accurate and comprehensive borrower profile.

Dynamic scoring and simulations
Another agent can apply adaptive scoring models and run repayment scenario simulations, guiding approval, denial, or escalation recommendations based on risk capacity and context.

Regulatory alignment
A compliance-focused agent can confirm all assessments meet suitability requirements, fiduciary standards, and traceability benchmarks. This solution layer supports audit readiness and decision traceability.

Workflow orchestration
Agents working in concert—each handling discrete steps in the credit workflow, from data gathering to scoring and reporting—can provide continuous fine-tuning of underwriting decisions as new data or market conditions emerge.

Managing risk and promoting trust

Robust and reliable
Credit decisions can have life-changing implications. Agents must be validated continuously against historical performance and varied economic conditions.

Transparent and explainable
Regulators and borrowers require clarity on underwriting decisions. Agents should log their reasoning (e.g., which income streams or risk patterns influenced a score), presenting audit-ready explanations that demystify the model’s logic.

Sector: Banking & Capital Markets
Function: Operations, Compliance & Risk, Information Technology, Finance
Type: AI, Agentic

How AI can help

Autonomous, event-level matching
Multiple AI agents continuously match payment confirmations, Nostro statements, ledger entries, and settlement messages (ISO 20022 rich data) to resolve breaks as they arise—requesting clarifications, attaching evidence, and escalating only true exceptions.

Intraday forecasting and liquidity routing
A forecasting agent aggregates real-time statuses (e.g., cross-border tracking, queued RTGS payments) and predicts cash peaks and troughs. A treasury agent then recommends actions—sweeps, intraday credit, collateral moves, or CLS pay-ins—to minimize buffers while meeting payment obligations.

T+1 and instant-payments readiness
Exception agents accelerate break resolution and funding decisions under compressed timelines, using enriched ISO 20022 data and end-to-end payment visibility to cut manual follow-ups and reduce same-day funding surprises.

Managing risk and promoting trust

Robust and reliable

Because reconciliation errors can trigger fails and liquidity penalties, agents should be tested on historical breaks, simulated queues, and stress days (month-end, index rebalancing). Agents should be designed to fall back to deterministic matching rules and human review for ambiguous cases.

Transparent and explainable
Since treasury, operations, and auditors must understand every break closure, agents should be designed to produce line-item audit trails (source message IDs, fields compared, confidence scores) and plain-language rationales for funding recommendations and payment releases.

Fair and impartial
Because intraday funding decisions affect business lines and clients differently, agents should be designed to apply consistent prioritization policies (e.g., client SLAs, regulatory cutoffs) and be monitored to avoid systematically disadvantaging lower-volume corridors or smaller counterparties.

Safe and secure
Given the high value of payment instructions, agents should be designed with strong cybersecurity controls (segregated runtimes, signed message handling, anomaly detection) and resilience testing to prevent spoofed confirmations or malicious instruction replay.

Sector: Banking & Capital Markets, Insurance, Investment management, Real Estate
Function: Compliance & Risk, Information Technology
Type: GenAI, AI

How AI can help

Evaluating alerts
AI models can evaluate alerts by calculating breachability (likelihood the vulnerability can be tapped in a real-world scenario) and exploitability (likelihood it can be actively taken advantage of to inflict harm).

Setting priorities
An AI-driven system can use enterprise-specific context such as workload risk profiles, application architecture, and network exposure to separate high-risk from low-risk vulnerabilities. AI automatically identifies which threats demand attention and which can be deprioritized, providing a data driven, risk-based prioritization process.

Managing risk and promoting trust

Robust and reliable
The vulnerability management system should be continuously retrained and tested against live alert data and historical incidents to validate its prioritization logic. A human-in-the-loop approach provides important oversight on critical decisions, especially during early-stage deployment.

Safe and secure
The models should be hosted in secure environments (e.g., on-prem or restricted cloud instances), with strict access controls and audit trails. Systems must be designed to prevent data leakage and be regularly tested for robustness.

Transparent and explainable
AI outputs should include the rationale for prioritization with traceable decision paths that help analysts understand and trust the recommendations.

Sector: Banking & Capital Markets, Insurance, Investment management, Real Estate
Function: Information Technology, Operations
Type: GenAI, AI

How AI can help

Supercharge your human capital
AI can be used as a component of cloud and data transformations to empower developers working across the enterprise on applications, data engineering, machine learning, and front-end development.

A helping hand in code development
As organizations explore new digital and cloud capabilities, development teams can accelerate and simplify their work by using AI as a force multiplier when writing, debugging, and documenting code, as well as translating ideas to code.

A shorter path to software
There are opportunities to use AI in software development to shorten the lifecycle and more quickly reach a stable and deployable version, such as by helping rapidly write APIs, ETL, data pipelines, or even front-end code.

Managing risk and promoting trust

Robust and reliable
Partial automation of programming-related tasks requires the system to be reliably available and accurate. If availability cannot be guaranteed to an acceptable extent, the benefits of automation must be weighed against the risk of erroneous or buggy code.

Responsible and accountable
The training data for foundation models may create legal risks related to intellectual property or copyright infringement. If the training data contains copyrighted material, the organization deploying the model needs to evaluate whether the presence of intellectual property in the training set could lead to legal challenges against the enterprise. Also, while the use of AI can accelerate the work of developers, without a human in the loop (e.g., to validate and debug code), critical failures may occur. Shoring up accountability may involve documenting and communicating standards and expectations for employees using AI.

Safe and secure
By using an AI system, proprietary code bases may be exposed to third parties, raising questions around the security of the data and controlled access to it. An inadvertent breach of confidential intellectual property could have significant enterprise impacts.

Sector: Banking & Capital Markets, Insurance, Investment management, Real Estate
Function: Cross-functional/Enterprise-wide, Operations, Finance
Type: GenAI, AI

How AI can help

An enabling interface
AI capabilities can be built on top of an existing solution to facilitate the communication of queries from the user to the search layer. This serves as the interface between search layers and databases, allowing users to easily mine all enterprise data, as well as generate structured analytics reports.

Speed to insight
With the ability to query and analyze disparate data sources using AI as an interface, the enterprise can move past traditional business intelligence techniques and dramatically reduce the time required to generate insights while increasing the workforce accessibility to business intelligence.

Managing risk and promoting trust

Responsible and accountable
When it comes to governance and control, granting more data access to a wider segment of the workforce can create a more complex challenge of restricting who in the organization is permitted to access sensitive business data.

Robust and reliable
Given the known challenges with model reliability and the potential for hallucination, banking and financial services organizations face the risk of inaccurate or false AI-derived insights influencing decision making and leading to negative ramifications even at the market level.

Private
When dealing with sensitive and proprietary information, the organization must contend with securing the data, removing or obscuring it in training and testing sets, and evaluating the model to determine whether it could “leak” protected information, either due to faulty function or a targeted attack.

Sector: Banking & Capital Markets, Insurance, Investment management, Real Estate
Function: Information Technology, Compliance & Risk
Type: GenAI, AI

How AI can help

Improve model training
Generative AI can be used to quickly create synthetic data to supplement machine learning model training data, which is then used to aid and accelerate digital and cloud transformations. In this way, generative AI complements the enterprise’s wider AI initiatives, fueling (rather than replacing) other AI deployments.

Amplify anomaly event detection
The rarity of anomaly events can make it difficult to train machine learning systems to detect instances of fraud, waste, and abuse. Creating synthetic data with generative AI gives ML systems a larger suite of examples that lead to a greater capacity to find patterns and anomalies in the data.

Harden the organization’s cyber posture
Just as synthetic fraud data can be used to train models to identify fraudulent activity, synthetic adversarial data can be used to train models to detect and mitigate cybersecurity risks and deceptive behaviors targeting virtual assistants.

Managing risk and promoting trust

Fair and impartial
A significant risk when generating synthetic data is that historic biases can creep into the generated data, perpetuating those biases (e.g., in the case of certain communities or socio-economic groups being underrepresented in the data because those groups have had lower levels of banking activity in the past). This bias is not necessarily intentional but is still harmful.

Robust and reliable
Synthetic data created with generative AI can be limited in its scope and scale, and it should not be presumed to be accurate or perfectly reflective of real-world data. An over-reliance on synthetic data may inject problems with data reliability, which can hamper the validity and usefulness of outputs and model training.

Sector: Banking & Capital Markets
Function: Customer Service, Compliance & Risk, Operations
Type: GenAI, AI

How AI can help

Condensing results for easier consumption
AI can be used to summarize and filter results from existing search engines to inform meta reports, as well as to summarize information for the customer relationship manager.

Research and analysis to inform reports
Generative AI, alongside other machine learning models, can be used to conduct preliminary data searches and meta-analysis, potentially accelerating the KYC process.

Managing risk and promoting trust

Robust and reliable
When using AI to perform research and analysis, there is a risk that it could skip or misconstrue highly relevant information, which could skew the conclusions in the meta-analysis and hamper sound decision making. If a new customer is erroneously assessed to be higher risk and a relationship manager declines the opportunity to engage the customer, the consequences are missed revenue and diminished customer engagement.

Private
When dealing with a customer’s financial or personally identifiable information, the enterprise faces legal and regulatory standards for data privacy. When using AI, the organization should take steps to ensure sensitive information does not inadvertently leak through model outputs, as well as govern who has access to the model, the underlying data, and the customer data it references.

Sector: Banking & Capital Markets
Function: Customer Experience, Customer Service
Type: GenAI, AI

How AI can help

An empathetic, personalized interface
Generative AI can be paired with other models to create a customer interface that delivers a hyper-personalized experience, such as by training the generative AI model to provide answers or insights with empathy. It also overcomes some of the challenges around more traditional chatbots that can lack an empathetic tone.

A more capable digital agent
Using generative AI can improve the usefulness and accessibility of a chatbot interface. With the capabilities of large language models (LLMs), a personal digital assistant can summarize contracts and answer nuanced questions, and the customer may enjoy a range of interface options, including text, audio, and imagery.

Managing risk and promoting trust

Responsible and accountable
While a generative AI-enabled solution may provide valuable answers and recommendations most of the time, there is a risk that too much confidence may be placed in the validity of the outputs, both by the organization and the customer. Generative AI is not an infallible oracle, and an overreliance on the AI solution may have a detrimental impact on customer actions, which can in turn increase financial risk.PrivateWhen confidential or personally identifiable information is inputted via the digital interface by the customer, the financial institution is obligated to follow the laws and rules that dictate how that sensitive information can be transmitted, stored, and accessed. Failing to do so could raise legal peril and potentially subject the enterprise to greater cyber risks.

Transparent and explainable
End users require a clear understanding of how their information will be processed and that they are interacting with a machine. At the same time, the enterprise needs to be able to interpret outputs and understand how and why the generative AI model created a given output.

Sector: Banking & Capital Markets
Function: Marketing, Sales, Compliance & Risk
Type: GenAI, AI

How AI can help

Customized materials for different audiences
AI can be used to create marketing materials that contain the appropriate tone, language, and cultural references, while also supporting consumer understanding of the product to maintain regulatory compliance.

Personalized sales at scale
With AI, financial services organizations have the ability to create marketing materials that are customized to individual customers—and to do so at scale.

Managing risk and promoting trust

Robust and reliable
For AI-generated marketing to be valuable, organizations must be able to rely on the validity of the output. AI can be prone to hallucinations, and when tasked with creating marketing that touts one product over another, there is a risk the model will return false statements. This injects potential regulatory violations that could result in fines and other penalties. To shore up reliability with AI, outputs should be validated by a human with the subject matter understanding to do so.

Fair and impartial
Datasets may contain latent bias of which the organization is unaware. This could be due to how the data was acquired, recorded, and curated, and the challenge is compounded when operating in multiple geographies. Organizations should consider the datasets used to train and fuel generative AI systems and whether unknown bias could lead to marketing materials that fall short because they fail to reflect important geographical and cultural differences.

Sector: Insurance
Function: Operations, Compliance & Risk, Customer Service
Type: GenAI, AI

How AI can help

Virtual damage rendering
AI can be used to help visualize the damage by replicating it virtually. The basis for the replication and visualization can be customer conversations, damage documents, photos, official reports, and other relevant media. In visualizing the data in this way, the claims agent is empowered to make better decisions when assessing the degree and cost of damage.

Automated claims reporting
With AI, claims reports can be generated automatically based on photographic evidence.

Managing risk and promoting trust

Robust and reliable
Damage visualization requires a high degree of accuracy, and erroneous AI outputs could lead to claims being paid incorrectly, potentially leading to overpayment (a detriment to the organization) or underpayment (a detriment to the customer).

Transparent and explain
ableIf claims agents use AI to automate aspects of claims processing but are unable to articulate to customers how the AI model derived its outputs or contributed to the cost and damage assessment, customers may not accept the outcome of the claims process.

Sector: Banking & Capital Markets
Function: Customer Experience, Customer Service, Operations
Type: GenAI, AI

How AI can help

Bring the bank to the customer
With a generative AI-enabled virtual space, customers can use a VR headset to conduct business with the financial institution and interact with a service representative from the comfort of their own home in a way that is convenient for the customer.

Hyper-personalized service
A generative AI agent can provide conversational, tailored responses to questions about customer accounts and financial needs. This helps the enterprise cater to its customers’ desires for a personalized experience while also avoiding the costs that come with adding more human customer service workers.

Speed and quality of service
In a virtual space, customer data can be accessed in real time by the conversational agent to provide faster, higher quality service and offerings.

Managing risk and promoting trust

Transparent and explainable
Given a generative AI-enabled chatbot’s capacity for conversational outputs, there is a risk that customers may not realize they are interacting with a machine. To promote trust, customers should not only be informed that the chatbot is not human, but they should also understand how their inputs and information are stored, accessed, and used.

Fair and impartial
The datasets used to train and inform the chatbot may contain latent biases, such as underrepresented customer groups or semantic deficiencies in some languages but not others. As a result, the model may simply not work as well for some customers, leading to a variety of negative customer impressions and complaints.

Responsible and accountable
With issues around reliability and data quality, if the chatbot outputs erroneous data or recommendations, a human stakeholder needs to be accountable for the outcome. Organizations can promote accountability by keeping a human in the loop and documenting roles and responsibilities.

Function: Compliance & Risk, Operations, Cross-functional/Enterprise-wide
Type: AI, Agentic

How AI can help

Quantitative analysis
AI agents can extract formulas, run calculations, and compare data against regulatory benchmarks to identify areas of potential concern.

Document review
The systems can process large volumes of reports, filings, and other documents, uncovering key insights and flagging areas that warrant attention from human examiners.

Dynamic examiner support
A conversational interface allows regulators to interact with the system, request clarifications, and guide the direction of analysis, keeping a human in the loop.

Cross-domain applicability
The framework currently being used for financial services examinations can be adapted to other types of regulatory inspections, including food safety, tax compliance, and transportation.

Managing risk and promoting trust

Private
Because examinations can involve sensitive institutional and personal data, agentic systems should employ strict data protection protocols and comply with all relevant privacy regulations.

Robust and reliable
Flawed analyses can lead to missed risks or incorrect enforcement actions. To avoid such problems, AI agents should be validated against known regulatory outcomes and tested continuously for accuracy.

Responsible and accountable
Since regulatory judgments can have serious legal and economic consequences, AI agents should be used primarily as decision-support tools, with human regulators retaining final authority over important decisions and actions.

Fair and impartial
To ensure equitable oversight, agents should be designed to apply standards consistently across institutions and sectors, minimizing the risk of bias in how examinations are prioritized or conducted.

Function: Operations, Compliance & Risk, Customer Service
Type: AI, Agentic

How AI can help

Automated document review
AI agents can scan and extract relevant data from submitted applications, checking for completeness and identifying missing or inconsistent information.

Compliance validation
Specialized agents can evaluate applications against permitting requirements, standard operating procedures (SOPs), and applicable regulations, flagging potential compliance issues early in the process.

Applicant feedback and support
Using AI and other advanced technologies, agents can provide real-time feedback to applicants, helping them correct errors or supply missing documentation before delays occur.

Reviewer assistance
AI agents can support human reviewers by summarizing key findings, verifying that evaluations align with regulatory standards, and highlighting areas requiring human attention.

Managing risk and promoting trust

Fair and impartial
To maintain public trust, agents should be designed to apply requirements consistently across all applicants.

Robust and reliable
Errors in interpreting regulations can lead to incorrect approvals or major delays. To improve consistency and reliability, agents should be trained on verified SOPs and continuously updated as requirements evolve.

Private
Since applications often contain sensitive personal and corporate data, agentic AI systems should be designed with strong data protection measures and full compliance with privacy regulations at all three levels of government.

Responsible and accountable
Permitting decisions can have important legal and financial ramifications. To avoid errors in accuracy or judgment, agentic AI systems should be primarily used for decision support, with final approval authority retained by human reviewers.

Function: Customer Service, Operations
Type: AI, Agentic

How AI can help

Conversational intake
Conversational agents can replace static application forms with guided, human-like interactions that clarify questions and adapt dynamically to applicant responses.

Automated document intelligence
Document analysis agents can extract and validate data from submitted records, reducing manual effort and maintaining accuracy in eligibility assessments.

Interview assistance
AI agents can support human caseworkers by preparing interview outlines, validating applicant responses in real time, and highlighting missing or inconsistent information.

Onboarding and on-the-job assistance
AI can accelerate new caseworker onboarding by generating tailored training materials and insights from historical cases, helping agencies address workforce challenges more efficiently. AI agents can also support human caseworkers by preparing interview outlines, validating applicant responses in real time, and highlighting missing or inconsistent information.

End-to-end applicant support
The specialized agents can work together across the intake and review processes to provide applicants with a seamless experience that is less intimidating and confusing, helping to ensure applications are complete and compliant.

Managing risk and promoting trust

Transparent and explainable
In order for applicants and caseworkers to trust AI-driven recommendations, agents should provide clear explanations of eligibility determinations, including references to applicable rules and policies.

Robust and reliable
Conversational agents might misunderstand user inputs, especially in cases of language barriers or low digital literacy. Agents should be trained on multiple languages and continuously monitored for performance.

Sector: Defense, Security & Justice
Function: Operations, Compliance & Risk, R&D/Product Development
Type: AI, Agentic

How AI can help

Automated data discovery
AI agents can continuously scan for and retrieve information from multiple structured and unstructured data sources, both classified and public—including briefings, news media, and other reports—reducing the need for manual searches and data collection.

Contextual linking of insights
If analysis of an old report produces new leads—such as a location or personal association—AI agents can automatically launch related queries to build a more complete picture of personal networks and activities.

Multimodal data integration
Specialized AI agents can process different types of inputs (e.g., text, imagery, sensor data) and combine the results into comprehensive, integrated insights.

Adaptive intelligence reporting
As the various agents collaboratively draft and refine reports, they can highlight key findings and update them dynamically as new data becomes available.

Managing risk and promoting trust

Robust and reliable
Automated agents might misinterpret ambiguous or misleading data, potentially affecting the accuracy of intelligence outputs. This is especially true for open-source information, which can be intentionally misleading or fake. AI agent actions should be validated against verified sources and continuously monitored to reduce errors.

Safe and secure
Given the sensitive nature of intelligence queries, special care must be taken to prevent adversaries from influencing the model (or gathering their own intelligence from what is queried).

Transparent and explainable
Providing adequate transparency in autonomous information exchange can pose significant technical and operational hurdles, especially given the complexity of integrating data from multiple sources. AI agents should document their reasoning and data sources so the outputs are easier to trust and audit.

Responsible and accountable
Given the life-and-death consequences of intelligence information, AI agents should be viewed as decision-support tools, with human analysts and leaders continuing to have final responsibility for decisions and actions.

Function: Operations, Customer Experience, Customer Service
Type: AI, Agentic

How AI can help

Automated transcription and generation
A generator agent listens to the audio of patient visits, converting it into structured SOAP notes and EHR-ready data, which minimizes the need for manual entry.

Validation and compliance
A validator agent reviews the generated notes for clinical accuracy, completeness, and regulatory compliance, confirming that records meet organizational and legal standards.

Workflow orchestration
An orchestrator agent manages the entire documentation process, routing tasks among the specialized agents and providing smooth, efficient workflows.

Seamless integration
The agentic AI system formats the validated documentation into common formats (e.g., DOCX, HTML) for direct integration into EHR systems, reducing friction and manual effort for providers and staff.

Managing risk and promoting trust

Fair and impartial
Documentation agents should be designed and tested to work effectively across varied patient populations and care contexts.

Robust and reliable
Because errors in transcription or orchestration could compromise documentation quality, AI agents should be validated regularly and continuously updated to reflect evolving medical and compliance requirements.

Transparent and explainable
For providers to trust that AI-generated documentation accurately reflects patient encounters, AI agents must be able to explain how notes were created and why specific items were flagged for correction.

Private
Some patients might be uncomfortable having their medical visits recorded and then analyzed by AI. Every patient should be given the choice to opt out. Also, AI agents must strictly comply with HIPAA and other relevant regulations, encrypt audio and text records, and minimize retention of personally identifiable information.

Function: Compliance & Risk, Operations, Cross-functional/Enterprise-wide
Type: GenAI, AI

How AI can help

Data collection and analysis
AI can automatically gather, structure, and analyze vast volumes of policy documents from government websites and public sources around the world. The technology performs multilingual data extraction, applies natural language processing to categorize and summarize policies, and synthesizes insights into structured outputs that can be validated by subject-matter experts.

Managing risk and promoting trust

Fair and impartial
The AI model should be designed to avoid reinforcing systemic biases. Human experts from different backgrounds and regions are embedded in the feedback loop to validate model outputs, helping to ensure representation across geographies and policy contexts.

Robust and reliable
Automated systems should undergo rigorous, iterative testing to help ensure the reliability of outputs. Policy insights are continuously benchmarked against human analysis and real-world policy documents to maintain a high level of accuracy and dependability, especially in politically sensitive or under-reported regions.

Function: Customer Service, Customer Experience, Operations
Type: GenAI, AI

How AI can help

A digital agent for engagement
An AI-enabled virtual assistant can serve as the interface between citizens and government information, helping with questions and transactions via empathetic, natural language.

Reaching across datasets
The virtual assistant can distill and summarize information from myriad sources on a variety of topics to answer questions in a multitude of languages regarding service requirements and appointment options.

Managing risk and promoting trust

Responsible and accountable
While virtual assistants may be valuable for providing information, they may not be suited to providing true insight and advice. Agencies need to guard against over-reliance on AI solutions and the potential for citizens to take some action based on faulty or improper AI output.

Safe and secure
A model tasked with providing accurate information may be a target for cyber criminals seeking to access sensitive information or manipulate the model and its underlying data. Many government agencies contend with cybersecurity regulations and standards, making model security a priority.

Robust and reliable
Model accuracy and timeliness depends in part on the data sources it can access. If information is outdated or incorrect, it creates a risk of erroneous outputs. Human stakeholders responsible for updating information have a direct impact on model reliability and user trust.

Sector: Defense, Security & Justice
Function: Cross-functional/Enterprise-wide, Operations
Type: GenAI, AI

How AI can help

Greater accessibility
AI can provide a natural language interface that allows non-technical users to access and understand data that might otherwise only be accessible to technical users.

Democratizing insights
Rather than placing all of the burden for data analysis, interpretation, and visualization on a technical team, an AI interface reduces the effort by allowing more stakeholders to work with the data and derive their own insights.

Managing risk and promoting trust

Safe and secure
An AI model that taps into a variety of datasets can make it difficult to control which data is accessed by which stakeholders in which organization, raising important considerations for model security and governance.

Private
When dealing with sensitive and proprietary information that is subject to varying laws and regulations across jurisdictions, organizations are called on to ensure the AI model does not leak, inadvertently divulge, or inappropriately access sensitive or restricted data.

Transparent and explainable
To accurately interpret data and AI outputs, the end user needs to understand which data was referenced for the output, which could not be accessed, and the potential biases in the available data.

Function: R&D/Product Development, Operations, Compliance & Risk
Type: GenAI, AI

How AI can help

Generating 3D city models
Using AI, thousands of 3D images can be rapidly created to help guide and refine a city design. Such 3D images form part of the design brief for urban planners and the master city plan.

Simulate natural disasters
AI can simulate natural disasters like earthquakes, floods, or hurricanes to evaluate the vulnerability of city infrastructure and plan for resilient urban infrastructure.

Planning for the future
By simulating population growth and demographic trends, AI can develop scenarios for urban expansion and plan for adequate infrastructure, housing, transportation, and public services that accommodate urban growth.

Managing risk and promoting trust

Robust and reliable
While AI models may create interesting or attractive designs, they require human review and validation to ensure they meet urban planning requirements and can be feasibly built in the real world.

Transparent and explainable
A lack of contextual knowledge of urban planning may lead AI to develop improbable scenarios, and analysts need to be able to understand how and why the model produced an output in order to confirm and validate it.

Function: Learning & Development, Operations, Customer Experience
Type: GenAI, AI

How AI can help

A digital, adaptive teacher
AI can serve as a virtual instructor, drawing from resources and lesson plans to hyper personalize the learning experience. The model can check the student’s work and comprehension and adapt lessons and learning strategies according to the student’s individual weaknesses, strengths, and preferences.

A force multiplier for teachers
When personalized digital teachers can work with students one-on-one to master new skills and knowledge, the human instructor can focus on higher level planning, interacting with students, evaluation, and student support.

Managing risk and promoting trust

Responsible and accountable
While digital teachers can offer valuable advantages in adaptive learning, the model should not be expected to satisfy all of the important lessons teachers impart, such as social lessons around collaboration, conflict resolution, and empathy. The human element in teaching is essential, and educational institutions need to take a responsible approach to integrating AI-enabled teachers.

Robust and reliable
Because AI is susceptible to outputting inaccuracies and hallucinations, there is a risk that a virtual teacher could teach incorrect facts or produce poor learning strategies.

Private
Student data is subject to education regulations, making model security and data privacy a priority when deploying digital teachers.

Function: Operations, Compliance & Risk, Customer Experience
Type: GenAI, AI

How AI can help

AI assistant
AI can identify data related to common themes and topics and then summarize that information in response to user queries, helping to identify policy differences, conflicts, and gaps.

Citizen engagement in policymaking
Using AI, governments can create interactive platforms and chatbots that encourage citizens to participate in policymaking discussions. The AI-driven interface can gather public opinions and feedback on policies, making it easier for citizens to voice their views.

Managing risk and promoting trust

Private
Some of the data relevant to policy issues may be sensitive or restricted, and the AI model may require controls to limit which users can access which datasets.

Fair and impartial
Various stakeholders aim to influence policymaking. AI might be biased in giving higher weight to comments and input coming from some sources over others. This has the potential to produce biased policies that favor certain businesses or segments of society.

Function: Operations, Compliance & Risk, Procurement/Sourcing & Supply Chain.
Type: GenAI, AI

How AI can help

Automated drafting
AI can automate the RFP and SoW writing processes by generating the initial drafts based on templates, historical documents, or specific prompts provided by procurement officials.

Extracting information
AI’s advanced Natural Language Processing (NLP) capabilities can help extract relevant clauses and requirements from existing contracts, SoWs, and legal documents. Such information can be used to either create new contracts or assess the risks posed by existing contracts.

Managing risk and promoting trust

Transparent and explainable
AI may not be able to explain why certain clauses are added to a contract while others are excluded, which is vital information for the human user validating the outputs.

Private
Ingesting existing and historical contract data may pose data privacy and legal hurdles. Model governance is necessary to ensure the AI model, as well as the underlying data, meet privacy rules, regulations, and standards.

Function: Customer Service, Operations, Customer Experience
Type: GenAI, AI

How AI can help

Aiding frontline workers
AI can be used to create real-time audio and text messages in different languages as frontline workers interact with residents around a variety of services, such as social care, health care, and emergency response.

Translating official documents
Government agencies often handle the publication of official documents, laws, regulations, and policies. AI can streamline the translation process and help produce accurate and consistent translations.

Announcement and website translation
Government websites and public information (e.g., health and travel advisories) can be translated quickly to make essential information more accessible to the population.

Managing risk and promoting trust

Fair and impartial
The data used to train an AI model for use in translation may not be consistently accurate or robust across all languages, which could in turn lead to poorer translations and less access to citizen services for some language speakers than for others.

Private
The translating model may be exposed to sensitive information, necessitating steps to ensure the model does not mishandle or inappropriately divulge protected data and thus violate data privacy regulations.

Function: Operations, Compliance & Risk, Learning & Development Type: GenAI, AI

How AI can help

Summarizing official documents
Auto-generating transcripts of hours-long committee hearings and summarizing important bills and hearings can significantly reduce the administrative burden on staffers.

Process and summarize policy proposals and research
Legislative staff review a large volume of policy proposals and recommendations published by experts. AI can quickly summarize the documents for them, so staffers can spend more time on higher level policy analysis and decision making.

Managing risk and promoting trust
Fair and impartialAI may perpetuate latent biases based on its training set and generate skewed summaries that are partisan and favor certain ideologies.

Private
Ingesting internal policy proposals can expose sensitive information, requiring organizations to take measures that protect the confidentiality of internal documents.

Sector: Health care
Function: Operations, R&D/Product Development, Compliance & Risk
Type: AI, Agentic

How AI can help

Specialized data analysis
Specialized AI agents can apply deep expertise in focused areas. For example, imaging agents can flag abnormalities on scans, NLP agents can parse physician notes and pathology reports, and other agents can interpret lab results and genetic markers.

Contextual integration
A planning or coordination agent can link evidence across modalities—for example, connecting a lung nodule on a CT scan with smoking history and lab findings—before proposing next steps such as a biopsy or targeted treatment.

Continuous learning from records
Agents can draw on the electronic health record as common context and shared memory, adapting recommendations over time based on a patient’s history and treatment responses (in addition to population-level outcomes).

Workflow automation
By triaging scans, pre-annotating records, and uncovering key insights, agents can reduce the time physicians spend on routine analysis, allowing them to focus on patient interactions and issues requiring complex human judgment.

Managing risk and promoting trust

Safe and secure
Errors in diagnosis can be life-threatening. As such, agents should be deployed with strong safeguards, including rigorous validation of outputs and controlled simulation testing before integration into live workflows.

Transparent and explainable
Because clinical decisions carry major consequences, agents should provide transparent reasoning, including references to underlying data. Validator agents can be used to audit and cross-check other agents’ conclusions.

Private
Medical data is highly sensitive, which means that agent systems must follow strict access controls, comply with HIPAA, GDPR, and similar regulations, and use privacy-preserving techniques to minimize unnecessary data exposure.

Responsible and accountable
Since liability in clinical settings is complex and risky, agent recommendations should initially be deployed under physician supervision, with final accountability for diagnosis and treatment decisions retained by human clinicians.

Sector: Health care
Function: Customer Experience, Operations, Compliance & Risk
Type: AI, Agentic

How AI can help

Continuous patient monitoring
A monitoring agent can analyze real-time data from wearable devices, glucometers, or other sensors, detecting subtle patterns such as recurring glucose spikes or irregular heart rhythms.

Evidence-based insights
A knowledge agent can stay updated with medical guidelines and population-level data, interpreting broader patient patterns and suggesting personalized, evidence-based adjustments to care.

Personalized coaching
A coaching agent can engage patients directly through apps or chatbots, offering tailored advice, timely reminders, and personal encouragement that reflects their specific health needs and behaviors.

Automated, coordinated care
Agents can schedule follow-ups, suggest medication adjustments (with clinician approval), offer healthy recipes, and even auto-order groceries—all orchestrated through a shared patient profile.

Managing risk and promoting trust

Fair and impartial
Biased training data could lead to less effective guidance for certain demographic groups. To minimize problems and avoid reinforcing existing disparities, AI agents should be trained and validated on diverse patient datasets.

Private
Because agentic AI systems continuously collect and analyze sensitive patient information, they must provide strict privacy protections and fully comply with data regulations such as HIPAA and GDPR. This includes obtaining clear patient consent.

Transparent and explainable
Patients and clinicians could initially be skeptical of following AI-driven guidance. To help build trust, AI agents should provide understandable explanations for their recommendations, including specific references to the data or guidelines that informed them.

Responsible and accountable
Clinical decisions can have major safety implications. As such, AI agent outputs should be reviewed by care teams where appropriate, with final accountability for treatments and adjustments retained by licensed clinicians.

Function: Customer Service, Customer Experience, Operations
Type: AI, Agentic

How AI can help

Specialized AI agents
Individual agents can handle specific tasks such as answering benefit questions, processing claims inquiries, and assisting with new member onboarding, providing responses that are timely, accurate, and consistent.

Seamless collaboration
AI agents can share context across tasks, improving their service quality and allowing for smooth escalation to human representatives when needed, without requiring members to repeat their information.

Dynamic prioritization
Requests can be triaged based on urgency, member status, or complexity, ensuring that high-priority cases receive timely attention.

Proactive engagement
AI agents can initiate reminders or follow-ups, such as prompting prospective members with enrollment information or reminding existing members of required documentation.

Managing risk and promoting trust

Private
Health plan support requires handling sensitive personal and medical information. Agentic AI systems must be designed with strict data protection protocols and full compliance with regulations like HIPAA.

Robust and reliable
Since members rely on accurate information to make critical financial and health decisions, AI-generated responses should be validated against authoritative data sources and continuously monitored for accuracy.

Responsible and accountable
Health-related interactions often require a strong dose of empathy and humanity. AI agents should be implemented as a complement to human staff, with clear escalation paths to ensure that members always have access to human support when needed.

Sector: Life Sciences
Function: R&D/Product Development, Operations, Compliance & Risk
Type: AI, Agentic

How AI can help

Optimized site and protocol design
AI agents can analyze hospital capabilities, historical trial data, and patient access to recommend optimal trial sites, appropriate sample sizes, and robust protocols that balance scientific rigor with operational feasibility.

Data integration and harmonization
By combining data from electronic health records, registries, and wearable devices, agents can create a holistic, real-time view of trial operations, ensuring decisions are made with the most current and complete information.

Adaptive trial management
Agents can continuously monitor trial progress and outcomes, dynamically adjusting resource allocation or protocol details to keep trials on track and compliant.

Patient stratification and recruitment
Specialized agents can identify and stratify eligible patient populations, improving recruitment efficiency while reducing delays caused by insufficient enrollment.

Managing risk and promoting trust

Fair and impartial
AI models should be trained and validated using demographically broad datasets to avoid systemic bias in recruitment strategies. Representation audits can help ensure no population segments are inadvertently favored or excluded.

Private
Since clinical trials depend on sensitive patient data, agents must be designed with strong privacy protections, data governance controls, and full compliance with regulations such as HIPAA and GDPR.

Robust and reliable
Trial outcomes directly impact patient safety and therapeutic approval. AI-generated recommendations need to be validated against clinical expertise and continuously refined through real-world feedback.

Responsible and accountable
Over-reliance on automation can lead humans to become complacent and miss important clinical nuances. AI agents should be treated as decision-support tools with physicians and trial managers maintaining final accountability.

Function: Procurement/Sourcing & Supply Chain, Operations, Manufacturing & Quality
Type: AI, Agentic

How AI can help

Demand forecasting agents
AI agents can analyze historical sales data, market trends, and external signals to help predict demand fluctuations and proactively adjust inventory levels and plans. AI can also integrate local geographical characteristics, disease prevalence, and socioeconomic factors to generate highly accurate, micro-market-specific demand forecasts.

Supplier and procurement agents
Agents can monitor supplier performance, lead times, and procurement activities, ensuring purchasing decisions align with both short-term needs and longer-term forecasts.

Production and scheduling agents
AI can optimize production schedules, identifying bottlenecks and aligning output with available resources and expected demand. By leveraging cross-silo data and advanced analytics, AI can identify patterns and trends that traditional methods might miss, helping to mitigate product shortages.

Distribution and logistics agents
Agents can monitor distribution flows, detect inefficiencies, and propose adjustments in real time to help ensure inventory is positioned where it is most needed.

Coordinated orchestration
An orchestration layer confirms that actions across procurement, production, and logistics remain aligned, enabling holistic, end-to-end inventory optimization.

Managing risk and promoting trust

Robust and reliable
Poor forecasts or system errors can trigger significant supply chain disruptions. AI agents should be validated against human planners and real-world performance, with continuous monitoring to ensure accuracy and stability.

Transparent and explainable
Inventory decisions can have a major impact on costs, compliance, and patient access. As such, agent-driven decisions should be supported by clear rationales that supply chain professionals can understand and verify.

Responsible and accountable
Automated systems can overlook subtle operational factors that a human expert might consider crucial. Conversely, over-reliance on automation can make human experts complacent. Human oversight of AI is essential, with supply chain professionals empowered to override or adjust agent recommendations as needed.

Sector: Life Sciences
Function: R&D/Product Development, Operations, Compliance & Risk
Type: AI, Agentic

How AI can help

Agent-driven design of novel compounds
Specialized AI agents can generate novel chemical structures, simulate compound interactions, and predict pharmacological properties, providing a broad and high-quality candidate pool. AI can also propose modifications to known molecules in early-stage drug development, rapidly evaluating their therapeutic potential and feasibility.

Multi-modal integration
AI systems can incorporate data from different domains—including chemical, biological, and safety data—creating a holistic view of each compound’s potential and reducing the risk of pursuing weak candidates.

Collaborative refinement
Toxicity, manufacturability, and regulatory-focused agents can evaluate compounds in parallel, sharing data with molecule generation and simulation agents to iteratively improve designs. An orchestration agent can oversee and manage the specialized agents, integrate outputs, and prioritize workflows, facilitating seamless progression from initial ideation to candidate selection.

Clinical trials and regulatory submissions
Agents can design preclinical studies and trial protocols based on predicted outcomes. Also, agents can prepare submission-ready documentation, track compliance requirements across regions, and adapt protocols to evolving regulatory standards.

Managing risk and promoting trust

Robust and reliable
Early design errors can lead to costly downstream failures. AI outputs should be validated against human benchmarks and continuously monitored.

Transparent and explainable
Regulatory approval requires meeting strict standards for safety, efficacy, and traceability. Agent reasoning and decision paths should support explainable outputs, including transparency into the data and assumptions used.

Responsible and accountable
Drug discovery is a complex challenge with life-or-death implications. AI agents should be deployed as co-pilots to human scientists, with humans retaining accountability for final go/no-go decisions.

Sector: Life Sciences
Function: Manufacturing & Quality, Compliance & Risk, Operations
Type: GenAI, AI

How AI can help

Automated, real-time inspection
Combining AI vision capabilities with generative models trained on synthetic and real-world data, manufacturers can automate inspection at scale. High-resolution imaging and computer vision detect anomalies with greater precision and consistency than the human eye.

Continuous learning
AI enhances the system by learning from historical defect data, adapting to new defect types, and identifying potential causes through pattern recognition across vast datasets. The AI system not only flags potential contamination in real time but also enables traceability—helping identify root causes by analyzing patterns across manufacturing lines, geographies, or specific production lots. This insight enables proactive correction and long-term process improvements.

Managing risk and promoting trust

Robust and reliable
AI models should undergo rigorous testing across multiple manufacturing lines and environments to help ensure high accuracy and minimal false positives/negatives. Redundancy checks, human-in-the-loop validation, and performance monitoring help ensure reliable operation even under variable lighting or packaging conditions.

Transparent and explainable
AI-based contamination detection can provide clear, trackable results, allowing manufacturers to understand why a product passes or fails inspection. Detailed imaging and reporting help ensure accountability and regulatory compliance while enhancing confidence in quality control.

Sector: Health care
Function: Operations, Compliance & Risk, Customer Service
Type: GenAI, AI

How AI can help

Retrieving policies and guidelines
An AI-based retrieval model can reach across large volumes of medical policies and member plans to identify the necessary information for a claims appeal.

Extracting patient data
Using extractive algorithms, the organization can rapidly consult unstructured medical notes, medications, lab results, and other electronic health records.

Writing the appeal
With the necessary information gathered by AI, an LLM can be used to generate an appeal letter.

Managing risk and promoting trust

Responsible and accountable
When consulting highly detailed guidelines, policies, and records to appeal a claim denied for vague reasons, AI models may misinterpret the denial or the records, leading to an unsuccessful appeal. Ultimately, a human needs to be accountable for validating appeal letters.

Private
By drawing from electronic health records, the model is consuming health information whose protection is subject to laws and regulations. Organizations must ensure that the data ingested and information outputted aligns with data protection and patient privacy expectations.

Sector: Health care
Function: Operations, Compliance & Risk, Customer Experience
Type: GenAI, AI

How AI can help

Supporting the provider
For providers, AI can help prepare a Prior Authorization submission by analyzing submission requirements and guidelines and cross-referencing with a patient’s medical records to ensure necessary requirements are met. AI can then aid in submission to the payer and continually learn which best practices tend to lead to Prior Authorization approvals.

Supporting the payer
For payers, AI can help reduce the time required to make a Prior Authorization decision, impacting the patient experience. It also helps mitigate fraud by determining if there are anomalies in a provider’s coding practices and supports compliance by analyzing submitted Prior Authorization requests and records against the payer’s policies and procedures.

More efficient operations
For both payers and providers, using AI can reduce work burdens and streamline the ability to handle Prior Authorizations, which can reduce costs while improving patient experiences.

Managing risk and promoting trust

Safe and secure
Prior Authorization requires the provider and payer to communicate sensitive patient data, such as protected health information (PHI) and personally identifiable information (PII), etc., which means this data is exposed to the model. Risks include unauthorized third-party access, as well as AI systems inadvertently revealing sensitive information during the generation process, thus compromising patient data confidentiality.

Fair and impartial
The process for submitting and responding to a Prior Authorization involves a standard set of rules and the patient’s medical history, which introduces the potential for bias in AI models. This bias might arise from the historical data used to train the model (e.g., disparities in health care treatment or outcomes), which could lead the AI model to inadvertently perpetuate and even amplify such biases by making its own biased decisions or recommendations. The use of standardized authorization rules and patient-specific medical history, alongside continuous monitoring and careful evaluation, helps mitigate this risk and promotes fairer outcomes.

Robust and reliable
While the process for submitting and responding to a Prior Authorization revolves around a standard set of rules and the patient’s medical history, there is a risk the model will misinterpret nuanced medical conditions of underrepresented populations that were not in the training dataset, and thus falsely deny the need.

Sector: Health care
Function: Operations, Information Technology, Compliance & Risk
Type: GenAI, AI

How AI can help

Transformed claims processing
Using AI to help categorize incoming claims and analyze and assign accurate codes can improve the overall accuracy, efficiency, and speed of claims processing, resulting in faster reimbursements and a streamlined experience for both the claims department and patients.

Reduced labor burden
By leveraging AI, human effort in the claims submission process can be redirected to higher value tasks, potentially yielding administrative cost savings for the payer.

Managing risk and promoting trust

Fair and impartial
A Large Language Model (LLM) used in medical billing may be susceptible to bias from skewed training data, incorrect labels, and under-represented cases, potentially leading to incorrect claim categorization. To mitigate these issues, careful data collection, diverse model testing, and continuous monitoring and adjustment are vital for ensuring fair and accurate performance.

Private
To assess coding accuracy, the LLM compares the billed codes with the patient medical history, which exposes the patient’s data to the underlying model and creates potential privacy risks that need to be mitigated.

Robust and reliable
Medical coding is highly regulated with strict penalties for over/under coding. The accuracy and reliability of LLM outputs in this regard is essential, as mistakes could have serious consequences. Reliability may be challenging in part because patient medical history may contain multiple modalities (e.g., text, images, and video).

Sector: Health care
Function: Customer Service, Operations, Customer Experience
Type: GenAI, AI

How AI can help

Triaging the in-basket
AI can be used to review routine messages (e.g., Rx refills, scheduling) and delegate simpler tasks to automation.

Message assistant
PCPs can leverage AI to summarize complex clinical messages for review and use the model to draft replies for provider input and response. AI models consult prior in-basket replies and EHR data when creating drafts.

Insights at scale
By using AI-enabled in-basket message systems at scale, organizations can identify issues related to patient negativity in their messages. The insights into complaints, expressions of dissatisfaction, frustration, confusion, or concern about care can inform interventions that may improve the patient experience.

Managing risk and promoting trust

Safe and secure
Use of AI for in-basket systems involves collecting, processing, and storing large amounts of sensitive patient data, such as medical history, diagnoses, and treatment plans. This data is subject to strict privacy laws, and any unauthorized third-party access could result in legal and financial consequences for health care providers.

Responsible and accountable
If messages are composed or summarized with inaccurate information, it could lead the PCP to erroneous decision-making or poor patient engagement, which can have significant consequences for patient health, trust in the health care provider, and the reputation of the organization.

Sector: Health care, Life Sciences
Function: R&D/Product Development, Information Technology, Operations
Type: GenAI, AI

How AI can help

Empowering professionals
With its capacity for learning from and adapting to iterative feedback, generative AI can act as an enabler for professionals across various sectors. It offers the opportunity to continually refine existing domain-specific AI models by adding new training data. This iterative enhancement increases the model’s accuracy, utility, and relevance to the user’s specific professional needs. In this way, generative AI can empower professionals by providing them with tailored, precision AI tools that evolve with their work.

Streamlining health care model development
Generative AI can help simplify model development in the complex and highly regulated health care industry. By focusing on intuitive user interface designs and automated processes, generative AI minimizes UI obstacles, making it more accessible for professionals to refine and improve their existing models.

Improving alignment
Generative AI leverages reinforcement learning (RL) techniques, a type of machine learning where an AI system learns to make decisions by trial and error, to validate and improve its own outputs. This process assists in mitigating prevalent AI challenges, including hallucinations or confabulations, ambiguity, and colloquialism misuse. As a result, it bolsters AI’s reliability and furnishes professionals with more precise models and predictions, thus aligning AI capabilities more closely with user requirements.

Managing risk and promoting trust

Robust and reliable
Hallucinations or confabulation could lead to the execution of incorrect procedures or use of suboptimal reagents and equipment, causing inaccurate experiments and inefficient use of resources. Particularly in medical or pharmaceutical labs, inaccurate information could even lead to compliance or regulatory issues.

Transparent and explainable
The generative AI system incorporates tools that offer transparency into data engineering pipelines, including data preparation stages. This inherent transparency facilitates an understanding of the AI model’s functioning within the organization, fostering trust in the accuracy and reliability of the AI system’s outputs. It is a crucial component of the AI use case, demonstrating the system’s accountability and promoting its acceptance across the organization.

Sector: Life Sciences
Function: R&D/Product Development, Operations, Manufacturing & Quality
Type: GenAI, AI

How AI can help

Generation of novel processes
Leveraging historical data and scientific principles, an AI model could suggest novel experimental designs, more efficient processes, or alternate uses of reagents and equipment, stimulating innovation in laboratory procedures.

Data analysis and interpretation
AI uses a large language model (LLM) to analyze data from lab protocols, equipment specifications, previous experimental designs, reagent usage, and techniques, providing a holistic understanding of laboratory procedures and principles.

Managing risk and promoting trust

Robust and reliable
The integration of multimodal text and images of complex structures and processes in experimental design presents complexity. This can heighten the risk of unworkable, unfeasible, or inefficient designs, as interpreting and accurately representing this diverse and intricate data can be challenging. These challenges could potentially lead to errors in the design and execution of experiments, resulting in failed or less reliable outcomes and unnecessary time and resource expenditure.

Responsible and accountable
In the event of erroneous design recommendations, accountability may be an issue. Determining who bears the responsibility for incorrect designs and their potential consequences is important. The roles of human oversight and system validation need to be clearly defined.

Transparent and explainable
With the application of AI in experimental design, there may be challenges related to explainability. If scientific or academic papers are to be published based on the results, authors need to be able to adequately explain the methodology behind the AI-recommended designs, which can be inherently complex due to the black-box nature of some AI models.

Sector: Technology
Function: Sales, Operations, Customer Service
Type: AI, Agentic

How AI can help

Prospecting and research support
An AI-based business development representative (BDR) agent can research prospects, prepare initial pitch materials, and draft outreach communications, helping to provide consistent and timely engagement.

Lead qualification and CRM automation
Specialized AI agents can automatically update CRM systems, validate customer data, and handle routine record-keeping, reducing errors and freeing sales staff for higher-value activities.

Account executive (AE) enablement
An account executive AI agent can support the next stage of the sales process by preparing detailed materials, assisting with live customer inquiries, and coordinating documents to support the close of the deal.

Workflow integration
The AI agents can connect with sector-specific sales tools and platforms, align processes with industry best practices, and collaborate to provide seamless workflows across the sales organization.

Managing risk and promoting trust

Robust and reliable
Incorrect data entry or poor workflow execution could result in errors that undermine credibility and damage customer relationships. To minimize such problems, AI agents should be continuously monitored, validated, and tested against live sales processes.

Private
Because sales processes involve handling sensitive customer information, agents should comply with all relevant data protection regulations (such as GDPR and CCPA), minimize unnecessary data exposure, and use encryption to secure communications and CRM updates.

Responsible and accountable
While AI can support sales teams with research and recommendations, ultimate responsibility for customer relationships, pricing decisions, and contractual commitments should remain with human sales professionals and business leaders.

Safe and secure
Sales platforms and CRM systems are prime targets for cyberattack. AI agents should be integrated with enterprise-grade cybersecurity safeguards to prevent breaches, account takeovers, or data leaks that could damage trust with customers.

Sector: Technology
Function: Customer Service, Customer Experience, Operations
Type: AI, Agentic

How AI can help

Customer success planning
An overarching customer success agent reviews CRM and sales data to draft initial success plans and then continuously updates them with new inputs about usage, deployments, and engagement.

Ongoing monitoring and insights
Specialized agents analyze customer health scores, product usage, and support interactions, identifying risks and opportunities for proactive intervention by CSMs.

Strategic review preparation
Another AI agent prepares first drafts of materials for quarterly or strategic business reviews, providing customer-facing teams with timely, relevant insights.

Renewal and upsell support
A renewal agent drafts a renewal playbook, updates the CRM system, prepares quotes, and recommends product or pricing mixes, then cycles information back to the customer success agent for onboarding or cross-sell support.

Conversational support
Agentic AI-powered virtual assistants orchestrate advanced language models to deliver natural, adaptive conversations—handling a wide range of customer queries in real time and multiple languages. By combining orchestration with generative AI capabilities, these agents provide fast, accurate, and consistent responses, while automatically documenting interactions for future reference. This enables scalable, personalized support at lower cost.

Managing risk and promoting trust

Robust and reliable
Since inaccurate data or flawed insights could undermine customer trust and loyalty, AI agents should be validated against actual customer usage patterns and regularly updated with new business rules.

Transparent and explainable
For customer-facing teams to trust and act on AI-driven recommendations, agents need to provide clear rationales for their outputs and highlight supporting data.

Responsible and accountable
Strong customer relationships are built on personal empathy and trust. As such, humans will always be a critical part of the sales and support process (pre- and post-sale) and have ultimate responsibility for building and maintaining relationships that last.

Sector: Technology
Function: R&D/Product Development, Information Technology, Operations
Type: AI, Agentic

How AI can help

Specialized coding agents
AI agents can generate code from natural language requirements, translating business or product needs directly into executable code.

Automated testing agents
Testing agents can validate functionality, identify defects, and ensure that new code integrates seamlessly with existing systems.

Debugging and troubleshooting agents
When errors are found, agents can automatically propose fixes or apply patches, quickly resolving problems and accelerating development timelines.

Deployment and operations agents
Specialized agents can handle integration, deployment, and monitoring, facilitating smoother releases and more resilient production environments.

Code summarization and documentation
Agentic AI can automatically generate clear, human-readable documentation alongside code, improving maintainability and accelerating future development—without burdening engineers with manual write-ups.

Managing risk and promoting trust

Robust and reliable
In software development, code quality and integrity are paramount. AI agents should be validated against test suites, subjected to rigorous regression testing, and monitored in production to help ensure reliability at scale.

Transparent and explainable
Since developers must understand, maintain, and extend AI-written code, agents should provide clear documentation and traceability of their logic and coding decisions. This includes explaining why specific code structures, libraries, or fixes were chosen.

Private
Agents must avoid exposing sensitive business logic, proprietary algorithms, or customer data during development and deployment. Data privacy protections and secure handling of intellectual property are essential, particularly when using shared or cloud-based training data.

Responsible and accountable
Although AI agents can perform coding tasks well, ultimate responsibility for design decisions, compliance, and final code deployment should rest with human engineers and engineering managers. Clear governance frameworks should ensure that accountability is not abdicated to autonomous systems.

Sector: Technology
Function: Operations, Customer Service, Information Technology
Type: AI, Agentic

How AI can help

Specialized functional agents
AI agents can be tailored for specific functions such as customer service, network monitoring, billing management, and field support, each handling tasks autonomously with high speed and consistency.

Customer support automation
In support environments, AI agents can triage tickets, resolve common issues, and escalate complex cases as needed, whether through voice or text-based channels.

Operational integration
Integrated into existing enterprise platforms, agents can update records, monitor systems, and provide real-time recommendations, seamlessly aligning with current workflows.

Scalable orchestration
Multi-agent systems can coordinate activities across service lifecycle stages, enabling efficient end-to-end management with minimal human oversight.

On-site troubleshooting support
AI-powered tools equip field technicians with access to troubleshooting guides and generates clear, step-by-step resolution plans, improving on-site problem solving and network performance.

Managing risk and promoting trust

Robust and reliable
Maintaining data integrity and ensuring accurate ticket resolution are critical, especially when agents operate autonomously. AI agents should be continuously validated for accuracy and stability.

Safe and secure
Given the risks of cyberattacks targeting critical telecom and technology infrastructure, agent systems should be hardened against intrusion and designed with robust cybersecurity measures to protect networks from sophisticated attacks.

Private
Since agents often handle sensitive customer and operational data, strong security protocols and strict compliance with privacy regulations are essential.

Responsible and accountable
Agentic AI can make mistakes or overlook nuances in complex service cases. Human supervisors need to closely monitor AI agents and retain final responsibility for actions and decisions, intervening whenever necessary—especially in situations that involve sensitive customer information or regulatory compliance.

Function: Marketing, Operations, Finance
Type: AI, Agentic

How AI can help

Financial forecasting
Marketers need tools that can continuously analyze performance, forecast outcomes, and support decision-making on where best to invest their resources. A financial forecasting agent can analyze marketing budget inputs and business assumptions to generate full-year projections for revenue, spend, and profitability.

Marketing effectiveness analysis
A marketing effectiveness agent can analyze historical and current performance data to provide recurring insights on ROI, ROAS, and customer lifetime value.

Scenario planning
A scenario planning agent can run simulations of different budget strategies to model impacts on subscription starts, retention, and overall profitability.

Coordinated decision support
By sharing insights across forecasting, effectiveness, and scenario planning, AI agents can collaborate to provide a comprehensive view of marketing’s financial contribution.

Managing risk and promoting trust

Robust and reliable
Inaccurate forecasts and simulations can lead to poor budget decisions. To improve reliability, AI agents should be continuously tested and refined to reflect real-world market conditions.

Private
Because customer-level data is used to measure marketing effectiveness, strong data governance and compliance with privacy regulations are essential.

Responsible and accountable
AI agents are fallible and cannot perfectly anticipate all external factors or sudden market shifts. Also, AI hallucinations remain a concern. Ultimately, humans need to be responsible and accountable for all decisions and actions.

Sector: Telecommunications, Media & Entertainment
Function: Finance, Marketing, Operations
Type: AI, Agentic

How AI can help

Cost center analysis
AI agents analyze marketing spend across cost centers and recommend reallocations that optimize key performance indicators (KPIs) such as sign-ups, churn, and viewership.

Product, channel, and vendor optimization
Other agents assess product, channel, and vendor mix within each market, using predictive models to maximize subscription starts and profitability.

Regional and country-level scenarios
Scenario planning agents run simulations of budget allocations at regional and country levels, evaluating the impact of incremental or cost-saving changes on subscriptions, retention, and profitability.

Coordinated recommendations
Agents collaborate to provide a unified view of marketing performance and deliver actionable recommendations for smarter, more flexible budget allocation.

Managing risk and promoting trust

Robust and reliable
Bad marketing budget decisions can have serious consequences for the broader business. Agents should be validated regularly against historical budgets and real-world outcomes, using continuous performance tracking to help ensure accuracy and minimize the risk of misallocation.

Transparent and explainable
Agents should be able to provide clear explanations of their reasoning—e.g., how reallocating X from channel A to channel B affects ROI—so marketing leaders can understand and trust their recommendations.

Responsible and accountable
AI agents should be used as decision-support tools, with final responsibility reserved for marketing and finance leaders who understand strategic business contexts.SecureGiven the sensitivity of financial plans, systems must be built with enterprise-grade security to help ensure budget forecasts and proprietary data are protected from breaches or misuse.

Sector: Technology
Function: Sales, Operations, Customer Service
Type: GenAI, AI

How AI can help

Providing easy access to internal knowledge through chatbots
Salespeople can converse with AI-powered chatbots to quickly and easily retrieve sales playbooks, technical specs, competitive positioning, and customer references directly from internal documentation repositories.

Automatically drafting RFPs
AI models can produce high-quality, tailored RFP responses by finding and summarizing relevant content from existing sales documents, aligning answers with internal knowledge bases, and incorporating reusable proposal components.

Providing individualized sales support with little or no coding
Non-technical users, including sales reps and subject matter experts, can generate summaries, extract insights, and draft proposals through a simple user interface—no prompts or coding required.

Enabling customized sales processes and tools
Technical users can integrate AI tools directly into other internal systems, workflows, or dashboards to build more personalized applications

Managing risk and promoting trust

Robust and reliable
AI-assisted workflows can deliver higher quality outputs—in less time— than traditional approaches. Users can flag incorrect responses or incomplete information; these are logged and reviewed in recurring QA cycles. Also, fallback mechanisms should exist to ensure consistent availability if problems arise with the AI models.

Transparent and explainable
Documentation should be provided for both business users and developers to explain how the system processes inputs and generates outputs. The chatbot interface includes citation tracing, where users can see which source documents were utilized to generate responses. Proposal-generation tools can allow users to edit and review outputs before submission, promoting human-in-the-loop oversight and transparency.

Safe and secure
All data and model interactions should occur within a secured internal environment, with no calls to third-party APIs unless vetted and approved. Systems should support audit logging for all user interactions to help ensure traceability and compliance. Role-based access controls can ensure only authorized personnel are able to view or generate sensitive proposal content.

Private
The system should not log personally identifiable information (PII) unless required by specific business rules and protected under internal data governance protocols. Feedback mechanisms should be anonymized where appropriate, helping to ensure user privacy while supporting continuous improvement. RFPs and customer documents processed in the system should be stored temporarily and purged according to data retention policies.

Sector: Technology
Function: R&D/Product Development, Operations, Information Technology
Type: GenAI, AI

How AI can help

Automating test creation
AI tools can be used to create new test cases from product requirement documents, bug histories, and structured datasets. These tools can assist engineers by proposing a wider set of test scenarios including ones not previously considered and by automating portions of test implementation through code generation.

Identifying test gaps
AI systems can also help identify gaps in testing coverage and can prioritize high risk areas based on historical failure data, although integration with structured data and internal governance systems remains an ongoing challenge.

Managing risk and promoting trust

Robust and reliable
Generated test cases can be validated against known test results and manually vetted to help ensure they hold up under real-world complexity. Also, systems can be stress-tested with increasingly complex product requirement documents to assess scalability and robustness across chip generations.

Transparent and explainable
AI-generated test cases can be accompanied by natural language summaries or rationales explaining why certain logic or edge conditions were selected. Engineers can trace outputs back to source inputs (e.g., PRD sections, bug databases), enabling better understanding and debugging of the AI system itself.

Safe and secure
The development and inference processes can occur in sandboxed environments with strict access controls to prevent accidental leakage of proprietary information. Integration with external AI services should be carefully managed to ensure no sensitive IP or design data is exposed to third-party systems.

Function: R&D/Product Development, Operations, Customer Experience
Type: GenAI, AI

How AI can help

Separating music into its component parts
AI, particularly deep learning-based source separation models, can analyze a fully mixed audio file and isolate its constituent elements— vocals, guitar, bass, drums, ambient noise, etc.—into discrete audio tracks with high fidelity. These models have matured significantly in recent years and can now perform at a level sufficient for commercial use in many scenarios. Rather than depending on traditional DSP (digital signal processing) or manual studio methods, the AI learns from large datasets of music to “de-mix” the sound using learned patterns of frequency and structure.

Leveraging Software-as-a-Service
Most deployments today use AI-powered SaaS platforms that allow internal teams to process catalog tracks quickly and securely. Internal quality control—along with artist or management approval—is then layered on to ensure that the extracted stems meet the creative and technical expectations of the project.

Managing risk and promoting trust

Robust and reliable
All outputs from AI models are subject to expert human review. Because source separation can introduce artifacts, tracks should be assessed case-by-case to determine if the fidelity is suitable for commercial or creative use. Teams should be trained to identify when alternative methods or manual interventions may be more appropriate.

Transparent and explainable
Processes for using AI in audio separation should be clearly defined internally and communicated externally as needed. Stakeholders—including sync partners, artists, and producers—should be informed when AI-generated stems are used, and how those stems were derived from the source material.

Responsible and accountable
All source separation use should be logged, and responsibility for approving commercial use should rest with both label- and artist-facing teams. If stems are to be reused, remixed, or publicly released, the appropriate clearance workflows—including licensing and revenue-sharing— must be followed.

Function: R&D/Product Development, Operations, Customer Experience
Type: GenAI, AI

How AI can help

Document extraction and digitization
AI models can process and extract structured information from legacy formats such as PDFs, microfilm scans, and outdated HTML, even when metadata is missing or inconsistent.

Content reconstruction
AI tools can intelligently identify article structure (headlines, subheads, body text, captions, bylines), reconstruct layout context, and reassemble fragmented articles into coherent, readable documents.

Semantic indexing and search
Large Language Models (LLMs) enable content to be semantically tagged and categorized, improving discoverability across themes, time periods, people, and places—even when specific keywords are not used.

Metadata enrichment and linking of multimodal assets
AI can supplement missing or corrupted metadata (e.g., publication date, author, topic) by analyzing linguistic and contextual clues. Also, the technology can cross-reference and re-link associated images, graphics, or videos from various archives where files may have been separated during prior migrations.

Improved access
AI can provide improved interfaces—such as chat-style queries or timeline exploration—to help users engage intuitively with the archive.

Managing risk and promoting trust

Fair and impartial
Systems are designed to ensure access to historical content across different eras and communities. Bias mitigation strategies are incorporated into model training and metadata tagging to avoid skewed representation of topics, regions, or individuals.

Robust and reliable
Extraction and structuring workflows are tested across various content types and legacy formats to help ensure consistent quality. Human oversight is embedded throughout the process to validate the accuracy and fidelity of reconstructed articles.

Transparent and explainable
A clear audit trail should be maintained for all AI-processed content, including logs of when and how specific items were extracted, tagged, and categorized. Explanatory overlays and metadata annotations help end-users understand the origin and limitations of AI-reconstructed documents.

Function: R&D/Product Development, Customer Experience, Operations
Type: GenAI, AI

How AI can help

Ongoing content development
Post release, developers can rapidly generate and deploy new gaming assets as expansions or microtransactions, such as seasonal or downloadable content (e.g., new characters and skins). Developers can use text prompts to generate new content in line with the existing game and requests from the community, then upload those assets to the game.

Managing risk and promoting trust

Responsible and accountable
Generated content resulting from a model trained with proprietary third-party data may lead to copyright claims if it is deemed to be too similar (without substantial variation).

Safe and secure
Players’ personally identifiable information (PII) could be fed into the models as they interact within the game, which raises risks around cybersecurity and regulatory compliance. The collection of PII, even inadvertently, places an obligation on the organization to secure the data as it is accessed, transferred, and stored.

Fair and impartial
Generated assets may over-index on player segments providing feedback or residing in specific regions. This uneven sampling of the input data could lead to bias in what assets are generated, and it may lead to missed opportunity and revenue as some customers are ignored.

Function: R&D/Product Development, Marketing, Operations
Type: GenAI, AI

How AI can help

Creative assistant tool
AI can be used to create imagery and apply edits using descriptive commands. Features like conversational editing, text-to-template, and text-to-image allow users to expedite the editing phase of the content creation process.

Picture editorial
Producers can automate footage management with video-to-text AI to evaluate and create tags for scenes and content. Text-to-video commands (e.g., “add more rain to this scene”) can be used to enhance and accelerate the editing process.

AI “reshoots”
Content creators can use scripts and 3D scans of actors to generate new content, alter footage to create more realistic special effects, and allow studios to make edits without the need for reshoots.

Managing risk and promoting trust

Responsible and accountable
AI tools may be trained with large databases of media and content, some of which may be copyright protected. As a result, the model outputs may include aspects of a creator’s or studio’s work or style that are not attributed to them, which raises legal risks.

Robust and reliable
Noticeable changes in style and brand quality due to AI content creation and editing may erode consumer trust in the brand and content.

Private
If bad actors access the underlying models or applications, it could contribute to the spread of fake content on behalf of the organization, leading to misinformation. Model owners should ensure strong privacy and access controls to mitigate this risk.

Sector: Technology Function: Marketing, Sales, Customer Experience
Type: GenAI, AI

How AI can help

Cohesive content generation
AI systems can be trained with on-brand content to mimic the style of company marketing materials and generate new, high-quality content rapidly and on demand.

Ideation with generation
Marketing departments can leverage AI to quickly create multiple versions of content in various styles to identify the most compelling and persuasive option.

Tailored, personalized messaging
With AI, organizations can easily create multiple versions of the same on-brand marketing tailored to different customers and audiences.

Managing risk and promoting trust

Transparent and explainable
Personalized advertisements may be customized based on data collected or purchased from individuals. This may be off-putting to consumers who realize the organization has such broad access to their data, potentially harming the enterprise’s brand reputation and undermining consumer trust. One way to mitigate this outcome is to ensure data collection and usage policies are transparent and communicated meaningfully to the consumer.

Responsible and accountable
Content produced by AI systems may not be subject to the same protections as human-generated content. Companies need to be wary of infringing on copyrighted material used to train AI systems.

Safe and secure
When brand data is used to train AI, there is a risk of data leaks that could result in sensitive information or IP being divulged to competitors. Companies need to ensure that their proprietary information is safely stored, transferred, and used, as well as monitor model outputs to validate that protected information is not being revealed.

Sector: Technology
Function: Operations, Customer Experience, Marketing
Type: GenAI, AI

How AI can help

Tools for custom localization and quality assurance
AI can be used to help organize and manage complex file type, analyze content before translation to optimize localization, and integrate glossaries, term bases, and language tools into workflows.

Content personalization across industries
AI-powered content personalization can supercharge localization efforts by helping to improve engagement, build brand loyalty, and increase conversions.

Speech recognition during translation
AI can be leveraged to enable voice user interfaces (VUI), transcribe video and audio content into text, and simultaneously translate spoken content into the target language.

Managing risk and promoting trust

Fair and impartial
Bias in the data used for content personalization could lead to unequal and unfair recommendations for certain groups of customers. In addition, AI applications are often trained on datasets from major languages, which means LLMs may have lower accuracy rates for less common languages and alternative dialects.

Transparent and explainable
Messaging and tone may change with language translation, which may negatively impact the text or audio being generated and the overall quality of the content. Localization should be audited to make sure that the messaging remains consistent with the original intent.

Sector: Technology
Function: R&D/Product Development, Manufacturing & Quality, Operations
Type: GenAI, AI

How AI can help

Iterative chip design
AI can generate and iterate chip designs and improve the outputs by having chip designs “compete” across a set of performance dimensions. At each new iteration, chip parameters are tweaked based on learnings from the best-performing designs in past iterations. These models are trained on existing layouts to learn patterns and constraints and generate new layouts that meet specific design requirements.

Managing risk and promoting trust

Safe and secure
With each new generation of novel designs, there is a risk of IP leakage and data breaches for proprietary chip designs and technical specifications generated by the LLM that could severely damage the organization’s competitive advantage. There should be rigorous security protocols in place to protect against this.

Responsible and accountable
When using AI for design, the organization needs to consider how to secure copyrights or patents and protect the IP of chip designs that are moved into production.

Transparent and explainable
For complex simulation processes, the organization needs the capacity to understand how and why the model determined a scenario or design to be optimal. Design validation requires users and stakeholders to be able to understand the reason for the outputs.

Sector: Technology
Function: Sales, Operations, Customer Service
Type: GenAI, AI

How AI can help

Spec summarization and search
AI can be used to create summaries of technical specifications based on targeted text-based queries to help understand which products meet customer requirements. It can suggest features and integrations that align with the customer’s existing technology stack and vendors, as well as provide links to articles or an internal knowledge base for future reference.

Knowledge management update
Sales case histories and other current documents can be used to update knowledge management so similar technical inquiries in the future can be rapidly addressed with previous resolution steps and summarizations.

Automated technical demos
AI can be used to automate the creation of software demonstrations tailored to specific clients and use cases. This is achieved by training on demo scripts and sample interactions to generate demonstrations showcasing a solution’s key features and benefits.

Managing risk and promoting trust

Private
Customer data (e.g., sales case history, customer tech stack/vendors) must be processed by the model, making it necessary to continuously monitor model outputs and safeguard customer data to mitigate privacy risks.

Robust and reliable
AI models are susceptible to hallucinations, or factual inaccuracies, making human validation essential to establishing trust in the outputs and the decisions they inform. What is needed is a verification process to ensure the accuracy and reliability of information derived from the model (e.g., spec summarization, demos). This has a direct impact on answering customer questions, and by extension, customer satisfaction and sales.