From “Sorry, I didn’t catch that” to “How can I help?” - The evolution of conversational design
Beyond functionality, the true success of AI Customer Agents lies in their ability to engage users naturally and effectively. This article explores the evolving landscape of conversational design, highlighting key trends and the crucial differences between designing for traditional Natural Language Processing (NLP) bots and designing for the new era of GenAI and agentic AI.
Value realisation in AI Customer Agents is paramount, as we explored in our previous article. However, even the most strategically chosen use cases and well-defined KPIs will fall short without compelling user experience. Conversational design plays a crucial role in this transformation - shifting AI from being merely a potential enabler to becoming a functional, engaging tool that drives truly effective customer interactions.
Understanding the frustration: Why conversational design is crucial for seamless experiences
When we consider conversational design, we refer to the practice of designing interactions between humans and AI systems to feel as natural and intuitive as possible. Just like the interactions we know from organic, interpersonal conversations. With conversational design, you craft and enable seamless dialogue by defining the AI's persona and ability to handle various user inputs. Something that will ultimately guide people towards seamlessly reaching their goals.
Why is this important?
Because perhaps you too have experienced the frustration of trying to reach a certain goal or objective in a conversation with an AI bot. The frustration of being unable to get the help you need or simply being unable to get on the same page.
The problem?
Traditional NLP-based chatbots often struggle to understand varied user inputs or maintain context across a conversation, leading to repeated misunderstandings, dead ends and user frustration. While some of it is caused by the technological limitations, some of it can simply be a result of poor conversational design.
Let us take the following conversation as an example:
Overcoming the frustration: A look into the evolving conversational design landscape
Several key trends are reshaping the future of conversational design. A future that will enable smoother and richer AI conversations. With the below shifts we see a range of new possibilities, but also a redefinition of the scope and complexity of conversational design.
Firstly, hyper-personalisation is moving beyond generic responses. Conversational AI is increasingly leveraging user history, preferences, and real-time data to deliver highly personalised interactions. This means that personalisation is not just about tailoring the information, but also about tailoring the tone and phrasing presented to individual users.
Secondly, personalisation is moving beyond just text. Multimodal interfaces are becoming increasingly prevalent. While text-based chat remains common, the future that is already starting to come into effect integrates voice, visual elements like images and videos, and even gestures to create richer, more flexible, and intuitive interactions.
“Imagine a customer agent that can explain even your most complex product features using a range of hyper personalised multimodal interfaces such as text, voice and videos. User experience is starting to take a monumental leap forward.”
With these emerging trends, conversational AI is evolving from simple scripted chatbots to sophisticated systems capable of natural, human-like interactions. Implementing conversational design will help shape how users experience these AI systems, which in turn directly influences engagement, satisfaction, and operational efficiency.
Let us explore this paradigmatic shift; how traditional NLP-based approaches differentiate from the dynamic, context-rich interactions enabled by Agentic AI – and how this new era of AI customer agents can enable richer interactions.
Exploring the paradigmatic shift: From traditional NLP Bots to GenAI/Agentic AI
Traditionally, conversational AI relied on NLP techniques focused on intent recognition and scripted dialogue flows. However, the advent of Agentic AI - powered by large language models (LLMs), autonomous decision-making and action capabilities - is redefining the scope and complexity of conversational design.
Traditional NLP bots typically operate on a rule-based or intent-driven approach. They excel at structured tasks and predefined workflows, often relying on fixed scripts and decision trees, where conversations are meticulously mapped out with expected user inputs and corresponding bot responses. They use intent and entity recognition to identify a user's intent (e.g., "check order status") and extract entities (e.g., "order number") to trigger a specific, pre-programmed flow.
"These bots often struggle with limited context retention, finding it difficult to maintain context across multiple turns or deviate gracefully from their predefined paths. If a user asks an unexpected question, the bot might loop or fall back to a generic "I don't understand."
Consider this example of a traditional NLP bot:
In contrast, GenAI and Agentic AI bots offer a fundamentally different approach. They feature dynamic learning and adaptability, continuously refining their responses and generating novel, contextually relevant content.
Agentic AI takes this further by autonomously making decisions and taking actions across multiple steps to achieve complex goals with minimal human input. These agents excel at advanced contextual understanding, maintaining context throughout extended conversations, remembering user preferences, and previous interactions.
They also demonstrate cross-domain knowledge integration, drawing information from vast knowledge bases, APIs, and user-specific data to provide comprehensive and nuanced responses. Crucially, they are proactive and goal-oriented, designed to understand a user's overarching goal and plan a sequence of actions to achieve it, even if the initial prompt is vague.
Here is how a GenAI/Agentic AI bot might handle the same scenario:
Notice how the GenAI/Agentic AI bot maintains context, proactively offers a solution, and provides options, making the interaction feel more natural and efficient.
Benefits of excellent conversational design
Investing in strong conversational design for your AI Customer Agents yields significant benefits across the board. By using it correctly within the business setting, it leads to improved customer engagement and Customer Satisfaction (CSAT), as natural, intuitive, and helpful conversations create happier customers.
When you design consciously towards people, and you intend to meet them in their preferred channels (chat, voice, app,), offer support in native language, and keep everything within personalised context, benefits will inevitably show up. For instance in some of the popular metrics like CSAT, Net Promoter Score (NPS), First Time Right (FTR) or increased traffic.
One example for all is Vodafone’s re-designing of rule-based virtual assistants to GenAI virtual assistants, which increased the NPS score by 20% and FTR to 90%1.
Excellent conversational design simultaneously drives increased efficiency and cost savings because well-designed bots can resolve a higher percentage of queries autonomously, reducing the burden on human agents and lowering operational costs 2,3.
A seamless and intelligent AI experience also cultivates an enhanced brand perception, reflecting positively on your brand and fostering trust and loyalty. Trust and transparency are the cornerstones of successful AI adoption; customers must feel confident that their data is handled securely and that the AI understands and respects their needs. Without this trust, even the most advanced conversational designs may fail to gain user acceptance or deliver value. Being transparent in AI use, customers need to understand when AI technology is involved, what its purpose is, and how to exit the loop to get in touch with a human.
Furthermore, users benefit from faster resolution times, getting their answers quickly, which reduces frustration and improves the overall customer journey.
Finally, every interaction provides valuable data insights that can be analysed to identify pain points, optimise processes, and further refine the AI.
However, these benefits are not automatic. To fully realise the potential of conversational AI, organisations must adopt a tactical approach - carefully planning the design, deployment, and continuous improvement of AI Customer Agents. This includes setting clear objectives, defining success metrics, and establishing robust feedback loops to monitor outcomes closely. By doing so, businesses can ensure that conversational AI evolves in alignment with customer needs and business goals, maximising value and avoiding common pitfalls.
At Deloitte, we follow the AI Customer Agent development framework to avoid the common pitfalls and to set the clients for successful AI Agent implementation as shown in example below:
Based on our Deloitte’s State of GenAI report 2025, trust emerges as a critical factor influencing the adoption and effectiveness of AI technologies. Organisations that prioritise transparency, ethical AI practices, and robust conversational design not only enhance user trust but also unlock greater business value from their AI investments.4
The future of conversation is fluid and smart
The shift towards GenAI and agentic AI fundamentally changes the role of the conversational designer. While traditional design focused on predicting and scripting every turn, the new paradigm emphasises defining the AI's intent, constraints, and learning mechanisms. The challenge becomes about empowering the AI to navigate complex, open-ended conversations while maintaining brand consistency and achieving business objectives.
The true mastery of value realisation with AI Customer Agents comes not just from their ability to automate tasks, but from their ability to engage with customers in a way that feels intelligent, empathetic, and genuinely helpful.
By embracing the principles of effective conversational design and leveraging the power of advanced AI, organisations can unlock unprecedented levels of customer satisfaction and operational efficiency. Deloitte practice covers the end-to-end journey for successful virtual agent deployment. From initial assessments to AI strategies and virtual agent implementation, our experts are ready to discuss with you.
Looking ahead, Agentic AI is poised to fundamentally transform customer interactions by evolving from reactive assistants into proactive, autonomous collaborators. Over the next months, we expect AI agents to increasingly orchestrate complex, multi-step workflows that anticipate customer needs, integrate real-time data, and deliver hyper-personalised, context-aware experiences across multiple channels.
This shift will enable organisations to offer seamless 24/7 support that not only resolves queries faster but also drives deeper engagement through intelligent recommendations and adaptive interactions.
As Agentic AI systems mature, their ability to learn continuously from interactions and coordinate across specialised agents will empower businesses to innovate customer service models, reduce operational costs, and build stronger, trust-based relationships. For leaders, embracing this evolution means preparing infrastructure, governance, and talent to harness the full potential of AI agents as strategic partners in delivering exceptional customer experiences.5
Are you interested in discussing how your company can utilise conversational design to achieve these benefits? Reach out to David Schopf (davschopf@deloitte.dk) for a demo of how you can do it in practice.
Click here to explore our AI Customer Agents Series: a series on crafting modular AI Customer Agents that deliver tangible value.
Sources:
1. Case study - Vodafone’s Virtual Assistant, TOBi, CDI, 2024
2. Customer Service Excellence 2025 | Deloitte Digital; Deloitte, 2025
3. Innovation Insight: Generative AI Chatbot to Improve CX and Agent Productivity, Gartner, 2025
4. Now decides next: Generating a new future, Deloitte, 2025
5. Prompting for action How AI agents are reshaping the future of work, Deloitte, 2025
6. The Future of Service and Support in 2028, Gartner, 2025
