The AI workflow interview script
Usage dashboards can show that people are using AI. Interviews show whether AI is changing the work in a way worth keeping.
AI adoption is now broad enough that simple usage counts are not enough. Anthropic's recent Claude-at-work survey asks users how AI affects their productivity and day-to-day work. OpenAI's adoption research also points to AI embedding across everyday workflows. Microsoft’s Work Trend Index frames the shift as one where human agency, judgment, and role design matter as agents become more common.
The practical move is to interview the people doing the work. Ask where AI saved time, where it created cleanup, and which repeated workflow deserves the next improvement.
The skill
An AI workflow interview is a 20-minute conversation with someone using AI in real work. It turns vague adoption stories into evidence: tasks, inputs, outputs, review steps, failure modes, and the next workflow to improve.
AI workflow interview
Person or role:
{who you are interviewing}
Workflow discussed:
{specific repeated work}
Before AI:
{how the workflow used to happen}
With AI:
{where AI enters the workflow}
What improved:
{time, quality, confidence, throughput, creativity}
What got worse:
{rework, risk, confusion, review burden, trust}
Next improvement:
{one workflow change, prompt, checklist, tool boundary, or training need}The interview script
Use these questions in order. Keep the focus on one real workflow, not general opinions about AI.
- What is one task you used AI for in the last week?
- What input did you give it: notes, email, files, data, transcript, or a blank prompt?
- What output did you get back?
- What did you do with the output next?
- Where did it save time?
- Where did it create cleanup or review work?
- What did you not trust?
- What would make this workflow repeatable for you or the team?
- What should AI never do in this workflow without approval?
- If we improved one AI workflow for you next week, which should it be?
A worked example
Suppose you interview a customer success manager about renewal preparation.
Person or role:
Customer success manager.
Workflow discussed:
Preparing renewal call briefs.
Before AI:
Read CRM notes, support tickets, emails, and previous call notes manually.
With AI:
AI summarizes account status and drafts call questions from CRM and support context.
What improved:
Prep time dropped from 45 minutes to 18 minutes.
The brief catches more support issues than before.
What got worse:
AI sometimes treats old email threads as current status.
Manager must verify account stage in the CRM.
Next improvement:
Add a source-of-truth rule: CRM owns status, support tickets provide risk context, email is background only.The prompt
After a few interviews, use this to synthesize what you heard:
Help me synthesize AI workflow interviews.
Interview notes:
{paste notes from 3-5 interviews}
Find:
1. Repeated workflows people already use AI for
2. Where AI clearly saves time
3. Where AI creates rework or risk
4. Source-of-truth problems
5. Review steps people already use
6. One workflow we should improve first
7. One habit, checklist, or prompt pattern to teach next
Return a concise adoption insight report with:
- Top workflow opportunity
- Evidence from interviews
- Biggest risk
- Recommended next experimentWhat to listen for
- Workarounds: People are often quietly building better prompts or manual checks.
- Review burden: AI may save drafting time but move effort into checking.
- Source confusion: Watch for cases where AI mixes stale and official information.
- Repeatability: A good workflow should happen often enough to improve.
- Approval boundaries: People usually know which actions feel too risky to automate.
The rule
Do not ask, "Do you like AI?" Ask, "Show me the last workflow where AI changed what you did next." That is where useful adoption evidence lives.