Building Trust in AI Self-Service

The Challenge

A product team was exploring an AI-powered support experience designed to diagnose customer issues and recommend solutions with minimal effort. The opportunity extended beyond a test of prototype designs and user flows into understanding how existing mental models of AI would shape trust, adoption, and interaction.

The Approach

I conducted qualitative concept testing with customers representing a range of AI familiarity, evaluating both an MVP experience and future-state designs. Sessions focused on expectations, mental models, trust, and decision-making rather than task completion alone.

Key Insights

As AI mental models evolve, product framing becomes part of the experience.
➥ As of summer 2026, participants expected AI experiences to be conversational, adaptive interactions, as informed by their existing AI experiences. Small framing decisions, such as describing a feature as "AI-powered" instead of just "AI", can help establish more accurate expectations and stronger trust.

Curiosity creates an opportunity, but expectations are high.
➥ Nearly every participant said they would try the experience simply because it was AI, creating a valuable first impression. That initial interaction, however, would need to deliver on its promise at launch to build trust and encourage future engagement. A negative first experience would be hard to overcome.

Confidence depends on complete experiences, not isolated recommendations.
➥ Users need enough context and transparency to trust findings and recommendations, along with clear next steps and a sense of resolution before considering an issue solved. In the case of a technical product, like a telecom service, sufficient = enough information to determine what is working and what is not, keeping the use of jargon to a minimum.

Support should feel seamless, not segmented.
➥ Participants didn't want to choose between AI and live support at the outset. Instead, they expect a single support experience that adapts to their needs and escalates naturally when additional help is required. They were willing to engage with AI at the outset, provided the safety net of a human was always perceived to be an option.

Impact

Rather than simply answering whether the concept "worked," the research uncovered broader principles for designing AI experiences. The findings shaped recommendations around product framing, trust, transparency, and the relationship between AI and human support, creating a stronger foundation for future design decisions.