Engineering, production, governance, cost, communication · insight-0292
Negotiate an ambiguous AI request down to one decision, one metric and a baseline
- Evidence type
- practitioner heuristic
- Difficulty
- easy
- Roles
- Data scientist, Data analyst, ML engineer, Applied AI
- Industries
- aviation, logistics, cross-industry
Summary: situation, action, reason, boundary
Situation
A director asks the team to 'use AI to optimise the network' within a quarter.
Action
Ask which decision would change, who makes it, how often, and what it costs to be wrong; propose a first slice with one decision, one success metric, the current baseline and a date; list what is explicitly out of scope.
Reason
A narrow, measurable first slice delivers evidence early and makes trade-offs visible for the next scope decision.
Boundary
New initiatives, vague executive requests, vendor-driven proposals.
Symptom or risk
Broad scope produces months of exploration, no shippable result and a loss of trust.
Exceptions
Exploratory research mandates can be broader, but still need checkpoints and stop criteria.
Trade-offs
Narrow scope may look unambitious; show the roadmap that follows a successful slice.
Pitfalls
Agreeing to scope verbally without writing the out-of-scope list.