Use whenInterviewing each priority department after shared language is established — bring the department's own baseline profile and walk actual work products, not abstractions.
OutputA structured departmental opportunity map — candidate workflows, friction, constraints, and where AI does not belong.
A
Interview context
Department
Interviewer
Participants
Date
Session length90–120 minutes, scheduled in advance.
1
The five Ds
Framework step + guiding question
Interview prompts
Notes
1 · Discuss — What actually happens?
Listen to the people doing the work. Output: conversation + problem brief.
Listen to the people doing the work. Output: conversation + problem brief.
"Walk me through the last time this work happened, start to finish." · "Which repetitive, high-volume tasks eat the most hours here?"
2 · Document — Is the workflow accurate?
Make actors, decisions, exceptions, inputs, and outputs explicit. Output: validated current-state workflow.
Make actors, decisions, exceptions, inputs, and outputs explicit. Output: validated current-state workflow.
"Who touches this, and what gets handed off where?" · "Where do delayed decisions and manual handoffs stack up?"
3 · Diagnose — Where are friction and risk?
Separate repetition from judgment, agency, privacy, and safety. Output: opportunity + risk map.
Separate repetition from judgment, agency, privacy, and safety. Output: opportunity + risk map.
"What knowledge is hard to retrieve, and what documentation burdens you?" · "Where is quality inconsistent — and what work depends on a few experienced employees?"
4 · Design — What should change?
Assign human and AI responsibilities. Output: future-state workflow.
Assign human and AI responsibilities. Output: future-state workflow.
"Which customer or employee pain points would you fix first?" · "What must stay human no matter what?"
5 · Develop — Does it work in practice?
Prototype, test with real cases, observe, and refine. Output: tested prototype + evidence.
Prototype, test with real cases, observe, and refine. Output: tested prototype + evidence.
"What real cases would prove a change works?" · "Which high-stakes activities here may make AI inappropriate?"
Loop rule: humans validate before the work advances; new evidence returns the team to Discuss.
2
Where AI does not belong
Record the department's own words, verbatim — a hard line is a design constraint, not an obstacle.
3
Accountable outcomes
Outcomes the department is accountable for
Candidate workflows for the inventory With friction quantified where possible.
You should now have