From CX to AX: how agentic AI is fundamentally transforming customer interaction

Agentic AI is transforming customer interaction: not more autonomy, but the right role, timing, and boundaries determine where trust is built or breaks down.

In brief:

• AI adoption always depends on context, control, and clear boundaries.

• Trust is not built on promises, but on proven value in real customer moments.

• EY Studio+ translates insights into human-centered, actionable AI solutions aligned with customer experience and strategic goals.

Dutch consumers generally prefer human contact in customer interactions, but in practice turn to AI when it is useful, controllable, and clearly bounded. It is precisely in these moments — when customers ask questions, seek advice, or need support — that a new reality emerges: a shift from customer experience to agentic experience (AX).This tension between stated preference and actual behaviour runs throughout the findings of the new Human Signals report and shapes how agentic AI is adopted in customer-facing services.

Targeted AI in customer interaction

This research does not argue for faster or broader adoption of AI, but for a more focused application of AI in customer contact. Rather than using AI in customer interactions purely for automation, the emphasis is on deliberately designing enriched experiences in which AI supports — rather than replaces — human service.

The research shows that few people trust AI with their most important decisions. Yet agentic AI is increasingly being deployed in customer-facing services that influence our finances, plans, and choices. Agentic services are AI systems that go beyond explanation or advice and are authorized to act, decide, or execute on behalf of the user. This has clear implications for where and how agentic AI can be applied most effectively.

Conditional acceptance in customer contactThe research cuts through clichés and AI hype. Acceptance is not a generic yes or no, but always conditional. Customers do not want AI to take over completely, but they do expect it to help where it adds value. They broadly accept AI as an explainer or navigator, are selective when AI acts as an advisor, and draw a clear line once AI is allowed to act or decide autonomously. These boundaries are surprisingly consistent across generations: both younger and older customers want to retain control, especially when important decisions are involved.

The pragmatism paradox of AI

In practice, people often say they prefer human contact, but turn to AI as soon as it is faster, easier, or clearer. This is the pragmatism paradox: the stated preference for human interaction disappears once AI proves its value in real-world use. Trust in AI is not granted upfront, but built through proven value in simple, low-risk situations. Only when AI proves reliable there does room emerge to apply it in more complex contexts.

Building trust in agentic interactions

The biggest mistake organizations make is trying to do too much, too fast. An AI pilot in customer service that doesn’t work properly causes customers to disengage. It is crucial to first solve customers’ small problems well and demonstrate real understanding. Only then does trust emerge to deploy AI in more critical customer moments.Experiments without clear added value drive customers away. We help organizations accelerate innovation by placing research and experimentation outside existing structures. There is no need to first free up internal processes or IT; you can immediately test what works in a focused way. This enables rapid strengthening of both AI strategy and implementation.

Tangible prototypes

The research stands out for its accessible approach, using concrete and tangible prototypes. Instead of abstract questionnaires, the research is conducted in the context of real customer experiences. Working with concrete examples and prototypes makes it immediately clear what AI means in practice. Together with the EY AI Lab, these prototypes are brought to life in real-world settings, allowing organizations to see and test exactly how AI affects their customer journey.

Validation and design

This approach enables targeted validation of assumptions, points where customers get stuck, and areas where AI can truly make a difference. A design sprint starts by clearly defining the target audience and business opportunity, followed by testing critical assumptions using prototypes. This can be done at scale, but also within three weeks through a limited number of interviews and a working prototype. The result is rapid insight that can be directly applied within the business.

Value for the customer

There is often a gap between what leaders think customers need and what customers actually expect. Many organizations focus on efficiency, for example by automating customer contact. But real differentiation lies in adding value for the customer. In some domains, human attention remains essential, while in others AI can provide fast, pragmatic support.

The research shows that people do not always want to speak to a human, but they do want clarity, structure, and a clear handover to human support when needed. Empathy is not always the answer: Dutch customers often value clarity and concrete next steps more than standard expressions of empathy. Only when the situation truly calls for it are human interaction and empathy genuinely appreciated.
These insights from the Human Signals research are broadly applicable. Although the study was conducted in a financial context, the findings provide clear guidance for deploying agentic AI wherever customer interaction, trust, and responsibility intersect — from healthcare and energy to telecom and public services.

Transparency and accountability

The pragmatism paradox is crucial here: people say they do not want to use AI, but they do when it works well and is easy to use. Trust is built through transparency and accountability — by showing that decisions are based on reliable sources and that a human is always accountable. Especially as AI becomes more autonomous, it is essential to be clear about where AI stops and human responsibility begins.

• People say they do not want to use AI, but do so when it works well (the pragmatism paradox).
• Trust is built through transparency and clear accountability.• It is essential to clarify where AI ends and human control takes over, especially as autonomy increases.

Human handover

Human handover and accountability are not prerequisites, but core design principles. AI must be able to recognize when a situation becomes too sensitive and seamlessly switch to a human. Only then does trust remain intact and customers feel genuinely supported.

The Dutch context calls for a rational approach. While earlier comparable research among UK consumers placed greater emphasis on emotional safety, the focus in the Netherlands is more strongly on control, transparency, and accountability. Younger generations are notably critical: they are aware of the risks associated with AI and big tech, and are therefore more cautious than older generations. This underscores the importance of context-specific research and design.