Help Me Think, Not Decide for Me

Overview

I led a discovery study to understand how business decision-makers want to use AI when evaluating a provider, and where they still insist on a human. Across nine in-depth interviews, I mapped a clear and stable line between the two, and surfaced an unscoped insight that reframed what a provider’s AI assistant is really competing with.

The question

On its surface this was a product question: where would an AI assistant help someone evaluating a provider, and where would it get in the way? Underneath, the real question was about people. In a process AI can increasingly mediate from beginning to end, where do people still insist on a human, and why there specifically? I chose one-on-one discovery interviews over usability testing because I wanted to understand mental models and trust thresholds, not measure performance against a screen.

My role

I led the study end to end: framing the research questions, setting the recruiting criteria across company size, conducting every interview, and running the analysis and synthesis on my own.

How I approached it

Nine semi-structured discovery interviews, 30 to 65 minutes each, with decision-makers spanning solo businesses to a 140,000-person enterprise. My analysis followed a deliberate path from raw data to insight:

1. 1. Capture. Rawest impressions right after each session, before reading the transcript.
2. 2. Per-participant review. Each transcript read in full for behavior and contradiction, not just stated preference.
3. 3. Coding. Tagged against recurring concepts, with new codes added as they emerged.
4. 4. Affinity mapping. Grouped into themes, each pressure-tested for whether it held across the sample.
5. 5. Segment lensing. Re-examined by company size to separate universal patterns from segment-specific ones.
6. 6. Confidence weighting. Unprompted, recurring patterns treated as well-supported; single voices held as hypotheses.

What I found

A stable trust gradient

People welcomed AI for exploration, comparing, summarizing, getting a ballpark, and pulled back sharply the moment real commitment entered. The line between “good for research” and “good for committing” was remarkably stable across the whole sample.

One enterprise IT manager rated his likelihood to interact with an AI assistant a ten, and his likelihood to complete an entire provider switch through it a zero.

“Why a human” is really four needs

When I probed why people wanted a human at the point of commitment, the reason split into four distinct mechanisms: reassurance, negotiation leverage, interpretation of nuance, and accountability. They weighted differently by segment, smaller businesses leaned on reassurance, enterprises on negotiation and a named relationship.

The insight nobody scoped for

Several participants were already using general-purpose AI tools to research providers before visiting any provider’s site. That reframed the competitive picture: a provider’s own assistant is not competing with rival chatbots so much as with an opinion the customer has already formed elsewhere. I had not gone looking for this; it surfaced because I asked people to walk me through where they actually started, and I followed it.

Impact

The findings reframed how the team understood the assistant’s role and shaped clear design priorities: it has to earn trust rather than assume it, stay verifiable through cited sources, never gate useful answers behind a contact form, and match its proactivity to whether the person is exploring or executing.

What I’d do next

With nine participants, I treated recurring, unprompted patterns as well-supported and held single voices as hypotheses. The clearest gap is the absence of a public-sector buyer. The natural next phase is quantitative: a survey built from the factors that pull people toward a human, to test how much each one weighs at scale.