Navigating the different AI assistant and agents
With the rise of Generative AI (GenAI) systems, AI assistants and agents have emerged as common AI helper solutions to support many use cases, such as drafting emails, creating documents, analysing data and more.
With so many different AI assistants and agents available in the market, it can be hard to choose the right one for your specific use case. This article is designed to help you understand what AI assistants and agents are, what they can do and how to choose the right one for your use cases.
What are AI assistants and agents?
An AI assistant is a conversational chat interface (powered by GenAI) that can help to search for specific information, generate responses, or review content and more. For example, helping employees to produce more accurate work by identifying errors and suggesting improvements.
AI agents are AI systems that use reasoning autonomously to perform complex tasks on behalf of a user or another system and require minimal direct control by the user. For example, an AI agent can take real-time notes during online meetings.
AI assistants and agents are used by organisations principally for:
- general productivity - Boosting creativity and productivity by optimising everyday employee tasks. For example, they can:
- automate repetitive tasks such as sending reminders, scheduling messages, and onboarding new team members
- act as a personal assistant for a company’s management and sales teams
- employee services - Supporting employees with general workplace processes and requests. For example, they can:
- support employees with IT and HR related issues.
- assist employees with routine tasks (e.g., raising IT issues)
- domain intelligence - Assisting employees with role-specific processes and information, for example:
- an AI assistant or agent can analyse legal documents and contracts.
- managing a quality control system (powered by AI) in manufacturing process.
The diagram below shows an example of how an AI multi agent solution could work:
Source: Deloitte, 2025.
Type of assistants and agents available
There are three core groups of AI assistants and agents available:
- Off-the-shelf: A ready-to-use AI assistant and agent solution which is pre-integrated with existing platforms. It uses GenAI capabilities without requiring or having to build a minimal development in-house unless you wish to customise or extend AI assistant and agents behaviour based on your use case. Examples include Gemini for Workspace, Microsoft 365 Copilot and other major Cloud and SaaS vendors' AI assistants.
- No-code/Low-code: A graphical development tool to build custom AI assistants and agents for specific use cases. It uses a pre-built LLM model with pre-built data connectors linked to an external repository. Or it allows you to build your own model and custom data connectors. Examples include Google Vertex AI agent builder, Microsoft Copilot Studio & similar AI development platforms from other providers.
- Custom coded/pro-code: A full customisable development toolset. This option is where software developers or engineers build fully customised AI assistants or agents on GenAI platforms available from different vendors. It allows more flexibility for the integration with bespoke applications based on a specific business, data or security requirements. Examples include Azure AI services and AI Foundry, Google AI Studio as well as a wide range of open-source libraries and third-party AI platforms.
Which AI assistant or agent is right for my use case?
In this section we step through an example scenario where an organisation is looking to build an AI copilot or agent for its sales team to improve sales performance and address key use cases within the team.
To identify the most suitable type of AI solution for this use case consider the following options, benefits, and factors based on these examples:
- in app sales solution (off-the-shelf): Using GenAI capability within Copilot for Sales (Microsoft) or Google Gemini. User interacts within the CRM applications. Integration with other off-the-shelf products is possible but depend on the vendor (e.g. Copilot for Sales with M365 Copilot, Google Gemini within Google Workspace or various SaaS providers)
|
Benefits |
Considerations |
|---|---|
|
|
- low-code/no-code solution with external data integration: Build a solution in low-code development tool with integrations with external data sources using connectors or custom coded components. Use combination of low-code development tool generative features and LLM models to orchestrate for specific use cases. User can interact with AI agent in custom websites or productivity applications (e.g., Google Vertex AI Agent Builder or Agentspace, Microsoft Copilot Studio).
|
Benefits |
Considerations |
|---|---|
|
|
- custom built solution (pro-code): Custom built platform to address organisation specific use cases and requirements from the sales team. Custom built application front end and workflows. User interacts with organisation platform that is preconfigured to access custom data sources (e.g., Google Agent Development Kit Teams AI SDK, Azure AI services and AI Foundry).
|
Benefits |
Considerations |
|---|---|
|
|
Adoption considerations
- Get the best with multiple AI assistants and agents: Currently, it is unlikely that a single AI assistant or agent - whether off-the-shelf or bespoke - is able to support and achieve a solution to address every use case across an organisation. However, a combination of AI assistants and agents with multi-agent AI systems architecture can help support the broader needs of your organisation.
- Improve response accuracy: An AI assistant or agent can be more effective if a combination of different AI assistant and agent systems are used. For example, An AI-generated response (where the assistant or agent is built with a low-code tool) can be enhanced with a vector search engine using an AI service offered by a cloud provider. In some cases, it may be possible to have a combination/hybrid of the three different types of assistants and agents as specified above to best address a specific use case with more complex requirements.
- Get data ready: AI assistants and agents cannot provide a comprehensive response if their data sources are fragmented or insufficient. Thus, it is essential to have a well-defined knowledge repository, supported by robust data governance and information architecture.
- Safety: When AI assistants and agents are adopted, organisations need to ensure safety and reliability through appropriate testing, monitoring and governance of AI assistants and agent solutions.
- Empowering your workforce: A successful adoption requires supporting behavioural change and encouraging employees to embrace new AI tools and unlock their advantages. Businesses can achieve this by transforming work culture through comprehensive training, active user engagement, and fostering collaboration.
Conclusion
AI assistants and agents can solve complex problems, help employees to be more productive, and provide organisations with numerous opportunities. To maximise the effectiveness of any solution, it is essential to:
- define a set of use cases that are relevant to business needs;
- understand and select appropriate knowledge source repositories;
- identify the development and integration efforts required based on the use cases;
- evaluate the different AI assistants and agents options available;
- continuously evaluate and improve the chosen AI assistant and agent solutions;
- implement robust governance and security measures to mitigate any associated risks;
- ensure thorough employer training and user adoption.
Selecting the most appropriate AI assistants and agents is crucial for effective adoption and unlocking the full potential of these technologies. By carefully considering the available options and weighing their benefits and considerations, businesses can ensure they choose solutions that align with their unique requirements.
