Ten cards for the scenarios that matter most. Each card works as interview prep and as a day-one playbook. Click any card to open the situation, what to watch for, your move, what to say, and what to avoid.
Lead with their problem, not your framework. Observe before you prescribe. Don't invalidate past decisions. Move small and fast. Introduce the framework only after you've delivered.
Tool chosen, licences likely bought, no users yet. You have a window. Run discovery, capture baseline, identify champions, pilot before you scale.
Licences paid, adoption flat, leadership frustrated. It's not a training problem. The infrastructure around the tool is missing — that's what you build.
Shadow AI embedded, multiple tools, no data policy, real value buried in chaos. Discovery, not audit. Frame governance as protection, not constraint.
No framework, no governance, no internal AI capability. A vendor pitch gave them a vision. You are the strategy, not just the execution. Evaluate it, build the foundation, own the roadmap.
Top-down pressure, probably budget, no tool selected. Activity is not value. Translate the mandate into outcomes and a use-case pipeline before teams scramble — or inherit Scenario C in six months.
Tool pulled or abandoned, active distrust, guarded budget. The move is forensic. Diagnose what actually failed, then reframe the next attempt as structurally different — not a retry.
Higher trust bar, stricter governance, different resistance pattern. Trust before efficiency. Run crawl fully, add supervised operation between pilot and full deployment.
Synthesised across healthcare IT, digital agency, and enterprise software — not deployed as a branded package elsewhere. That's a strength, not a gap. Authority is earned by building, not repeating.
The brief says "build an LLM." They need to evaluate, integrate, deploy, and govern one. The value isn't in training the model — it's in knowing what to do with it.