A customer executive says, "We want an AI agent for our operations team." That's all you get. How do you proceed?

30-second answerSay your answer out loud first, then reveal.

Don't start building. Turn the vague request into a concrete problem. Ask what outcome the executive wants and why now, then meet the operations team and observe their work to find specific, high-volume, painful tasks. Quantify them, propose 2–3 candidate use cases with value and feasibility estimates, agree on one with a success metric and owner, then run a time-boxed POC. Very often the right first solution is a narrow workflow with AI steps, not an autonomous agent.

Step-by-step

Clarify intent with the executive (30 min):
• "What would be different in 6 months if this works?" (cost reduction, faster SLAs, scaling without hiring?)
• "Why now? What triggered this?" (board pressure, competitor, backlog crisis)
• "Who should I talk to on the team?"
Discovery with operations (2–5 days): shadow 3–5 people; list tasks with frequency × time × error impact.
Candidate use cases (example):

Task	Volume	Time/task	AI fit	Risk
Classify and route incoming requests	2,000/day	3 min	High	Low
Reconcile vendor statements	300/week	40 min	Medium	Medium
Draft responses to escalations	150/day	15 min	High	Medium (review needed)

4. Recommend and agree: pick one with a clear baseline, success metric and champion. Explain why "agent" may mean a workflow with LLM steps.

5. POC plan with exit criteria (Q7).

What to say in the interview: "I'd reframe 'we want an agent' into 'which outcome, for which users, measured how?' The technology choice comes after the problem is clear."

PreviousHow do FDEs feed what they learn in the field back to the product team?

Little by little, you're building something great.

A customer executive says, "We want an AI agent for our operations team." That's all you get. How do you proceed?

Step-by-step

Related