Example scenario, not a client project. An insurer rolled out an AI assistant for claims adjusters, but fewer than one in five use it. Observation shows that the assistant summarizes the documentation correctly but doesn't show where the information came from — and the adjuster is still responsible for the decision. The conclusion: the problem lies in trust and responsibility, not in the quality of the summary.
Is this for you?
- The team has strong opinions about users, but few people have spoken to them recently.
- A product or tool was rolled out, and people are going back to the old ways.
- Leadership and the team see the problem differently, and no one has the data to settle it.
- You're planning AI in the work of people who may be afraid of it or not trust it.
What I do
- Stakeholder interviews — goals, constraints, the history of earlier attempts.
- Field research — I sit next to people during real work (contextual inquiry, on site or remotely).
- Tests of the current solution — where people make mistakes, get lost, give up.
- Product data analysis — what the numbers say before we ask people.
The AI layer
- I ask how people already use AI today: for what, in which tools, and why not in the company's own.
- I check when people trust AI output and when they check or reject it, and what they need in order to trust it more.
- I name the concerns: loss of control, responsibility for an AI error, a changing role. Without this, a rollout stalls regardless of the model's quality.
- Context: NN/g calls 2026 "the year of AI fatigue" — people are tired of AI features that don't help them work.
What you get
- A research synthesis — the key findings with quotes and evidence, on a few pages, not in a hundred slides.
- A map of needs and concerns by role.
- Role profiles (not marketing personas) — tasks, tools, context, attitude to AI.
- A list of open questions — what we don't know yet and how to find out.
- Recordings and notes for the team's further use (with participants' consent).
How it runs
- Research plan — questions, participants, methods.
- Recruitment — usually with the client's help; 6 to 12 people [to be confirmed].
- Sessions — interviews and observation.
- Synthesis — a joint session with the team, so the findings are theirs, not just mine.
- Presentation and decisions — what it means for the project.
Roughly 2–4 weeks [to be confirmed].
How we'll know it worked
- The team quotes users instead of its own assumptions.
- At least one important assumption has been confirmed or disproved.
- Decisions in the project refer to specific research findings.
What this doesn't cover
I don't run quantitative research on large samples, or market research. I can plan and interpret them together with your team.
Visualizations
Needs and concerns by role (sample artifact)
Sample artifact: the same change, seen by three roles.
From sessions to decisions
Research ends in decisions, not in a report.