Guide, 12 October 2026, by Anton Hantel
What do you get at the end of an AI interview project?
On first calls, the second question after "is it anonymous" is often "what do I hold in my hands at the end?" Someone has to take it to a leadership meeting, or to the people who were interviewed. Here is what comes back from a project, using an example from frontline interviews.
Themes, counted
Every statement is sorted into themes. A theme counts when six of the ten agents that code a statement agree. Next to each theme you see how many people raised it, out of how many, for example "51 of 120".
The reasons behind a number
If you also ask a scored question, you see the situations behind the score. Example only, not a client result: 120 frontline interviews where 38% said their workload is manageable, against 72% of managers.
| What people described | How many | Suggested next step |
|---|---|---|
| Two open positions in the warehouse have been empty since spring, covered with overtime | 51 of 120 | Approve temporary cover until both are filled |
| Next week's shift plan comes out on Thursday, so people can't plan childcare or appointments | 38 of 120 | Publish shift plans two weeks ahead |
| Ideas from last year were collected, and nobody heard what happened to them | 29 of 120 | Report back on every idea, including the ones turned down |
Each suggested step names who could own it, so the result can go straight into a decision.
Splits by group
You can compare groups you agreed on beforehand: team, site, region, seniority, function. Every figure comes from at least five people, so small teams stay in the totals rather than appearing on their own. More on that in How many people do you need?
What it looks like for other questions
The structure stays the same; what's counted depends on the question.
- Operations: process maps with the handoffs and rework people describe, and ideas ranked by value and effort.
- AI adoption: use cases by function, the blockers ranked by how many people hit them, and who inside already does it well.
- Mergers: overlap by function, and open decisions with a suggested owner.
- Strategy: how the strategy is understood, and the ideas and risks people raise.
The formats you take away
- The results in the app, where you can filter by the groups you agreed.
- Excel and API exports, and downloadable reports.
- Follow-up questions to the results from Claude, ChatGPT or Copilot, through an integration, so you can ask "what did the night shift say about safety?" without building a new chart.
How long it takes
From one question to the results in about a week. Interviews take 10 to 20 minutes each, by phone or in the browser, in each person's own language. When you've acted on the results, you can ask the same people again and compare.
What you don't get
You don't get a rating of any person, a ranking of teams by mood, or answers tied to a name. Results only appear for groups of five or more. Why, and what happens to the data: Are AI voice interviews anonymous?