Applied AI / Forward Deployed Engineer · Forward-Deployed Craft & Behavioral · Lesson 3 of 4
The customer-scenario round: a framework that wins
FDE loops include an open-ended customer scenario: "A logistics firm wants an AI agent to reroute shipments; they have SAP data, weather APIs, and 400 warehouse managers on different systems — how do you approach it?" The trap is jumping to a solution. Winners run a repeatable 5-step framework out loud.
- Clarify scope & success criteria: What's the real goal? What does 'working' mean numerically? What's out of scope? (Never skip this — it's the #1 differentiator.)
- Map stakeholders: Who's the exec sponsor, the daily users, the blockers (IT, security, procurement)? Adoption dies without the operators.
- Identify data & constraints: What data exists, how dirty, where does it live, what are the compliance/latency/cost limits?
- Propose an approach with explicit tradeoffs: simplest thing that works first; name what you're trading (quality vs cost vs speed vs build-vs-buy).
- Surface failure modes unprompted: where will this break (bad data, injection, edge cases)? How will you measure it and roll back? This is the move that signals seniority.
The adoption trapA classic curveball: "You deployed the platform but adoption is 12% after 90 days and the customer blames the product." The answer is not technical first — it's discovery: talk to the operators, find the workflow friction, check whether you solved the real problem, then fix with data. Adoption is a people+workflow problem as much as a model problem.
Live-demo-fails curveball"Your demo throws errors on screen in front of the client's execs — what do you do?" Stay calm, acknowledge it honestly, narrate what's happening, pivot to what does work or walk the approach verbally, and commit to a fixed follow-up. Composure under failure is the test.
◆ Lock it in
- Run the 5-step scenario framework: clarify → stakeholders → data/constraints → approach w/ tradeoffs → failure modes.
- Never jump to a solution; clarify success criteria first.
- Low adoption = a discovery/workflow problem; demo failures test your composure, not your luck.
Feynman drill — say it out loudRecite the 5-step customer-scenario framework and apply it to 'an insurer wants to automate claims triage.'
Step 1 rate your confidence · Step 2 pick your answer
In a customer-scenario round, after clarifying scope the strongest next move is to:
Step 1 — how sure are you?