Skip to content

03 · Drill 02 — Surface the Real Problem

Status: Outline. Body fills in Week 2. Voice: principal-level, BFSI-threaded, Apic-calibrated.

What this drill is. Five customer requests, each containing a hidden real problem. Drill: in 90 sec each, ask 3 probing questions that get from the requested AI feature to the underlying business problem.

What this drill is NOT. A solutioning drill. Don't propose architecture during this drill.

Prompts (5 customer asks)

Ask 1 — "We want Claude to write our compliance reports."

Hidden real problem (don't peek): Auditor finds inconsistencies between sections. Underlying need is structured drafting + consistency-checking with review gates, not free-form generation.

Ask 2 — "We want a chatbot for relationship managers to query customer data."

Hidden real problem: RMs spend 30% of their time switching between 5 systems to assemble a customer 360. Underlying need is integration + UX, not a chat interface.

Ask 3 — "We want Claude to do contract review."

Hidden real problem: Legal team has a 3-week backlog and the bottleneck is triage — which contracts need detailed review. Underlying need is classification + routing, not full review.

Ask 4 — "We want a Claude agent to handle vendor onboarding."

Hidden real problem: Onboarding fails because 4 internal teams pass forms that contradict each other. Underlying need is process redesign, not automation; Claude can help draft, but the process problem is upstream.

Ask 5 — "We want Claude to summarize all our internal Slack channels."

Hidden real problem: Leadership doesn't know what's happening across 12 product teams. Underlying need is structured weekly briefs from team leads, not Slack summarization (which would surface noise + leak signal).

Time box

  • Per ask: 90 sec to deliver 3 probing questions.
  • Total drill: ~10 min including reading the asks cold.

Rubric

Strong Hire

For each ask: - 3 questions delivered, each in a different class (context / problem-surface / constraint). - At least one question that refuses the surface framing without being adversarial. ("If the AI piece weren't part of this, how would you solve this problem today?") - Identifies the real problem before offering any architecture. - No competitor comparison anywhere.

Hire

  • 3 questions delivered but all in the same class (e.g. all "what" questions).
  • Real problem surfaces but only because the prompt obviously hints at it.

Lean No

  • Architect immediately offers solution architecture (Lean No).
  • Questions are pitch-shaped ("would Claude be useful for…?").
  • Surface ask accepted as the requirement.

Strong No

  • Architect skips probing entirely and pitches Claude features.
  • "Yes we can definitely do that" said without further investigation.

The 3-question template

The Strong-Hire structure for any customer ask:

  1. Context question — "Walk me through what's happening today before any AI gets involved."
  2. Problem-surfacing question — "What's the business outcome you're chasing? What would success look like in 6 months?"
  3. Constraint or counter-factual question — one of:
  4. "If the AI piece weren't part of this, how would you solve this problem?"
  5. "What did the previous attempt teach you?"
  6. "Who has to sign off before this goes live?"

If all 3 questions land, the customer surfaces the real problem in their own words.

Common Lean No traps

Trap 1 — Solution-shaped questions

"Would Claude help if it could…" — that's a pitch disguised as a question. Lean No.

Trap 2 — Single-class questioning

3 "what" questions in a row. The constraint and counter-factual classes never get used. Real problem doesn't surface.

Trap 3 — Accepting the AI framing

Treating "we want Claude to X" as the requirement. The requirement is the outcome, not the AI shape.

Trap 4 — Over-leading

"It sounds like what you actually need is…" — finishing the customer's thought for them. They'll agree to be polite, then disagree internally.

Trap 5 — Skipping to constraints

Constraint questions before context questions. Sounds like an audit, not discovery.

How to run this drill

  1. Cover the "hidden real problem" lines.
  2. Read each ask cold; deliver 3 probing questions in 90 sec.
  3. Reveal the hidden problem after answering.
  4. Score: did your questions, in their own logic, surface the hidden problem?
  5. For asks where you missed, write the one question you should have asked.
  6. Application trigger: 4 of 5 asks graded Strong Hire across 2 sessions.

Cross-references

Strong-Hire bar for this drill

  • 3 questions per ask, in 3 different classes, in 90 sec.
  • Real problem surfaces in customer's own words.
  • No solutioning during the drill.
  • 4 of 5 Strong Hires across 2 sessions.