1) What business KPI will this improve—and by how much?
Expect a baseline, target range, owner, and measurement plan.
2) What are the top 3 use cases worth doing first?
Ask for a value × feasibility prioritization and sequencing rationale.
3) How do you prevent “pilot sprawl” and AI theatre?
Look for governance cadence, stop-rules, and portfolio discipline.
4) What decisions will AI influence—and who owns them?
Demand clarity on decision rights, escalation, and human oversight.
5) What governance will satisfy our board / regulator?
Expect auditability, explainability, logging, and safe-fail controls.
6) What data do you need—and what if data quality is weak?
A credible plan includes data ownership, quality metrics, and fallbacks.
7) How will AI fit into real workflows (day-2 operations)?
Ask how teams will use it, override it, and monitor it.
8) How do you manage risk, bias, and failure modes?
Expect failure-path design, rollback, and incident response approach.
9) What is the build / buy / partner recommendation—and why?
Look for speed-control trade-offs, not vendor hype.
10) What will it cost—and what ROI should we expect?
A serious partner will outline scenarios, not a single number.
11) How will we know it’s working within 30–90 days?
Ask for milestones, leading indicators, and adoption metrics.
12) Who owns the system long-term—us or you?
Expect a transition plan, documentation, and internal capability uplift.
13) What does “success” change in the operating model?
Great partners identify what to stop, simplify, and redesign.
14) What security and privacy posture do you recommend?
Expect controls around access, data retention, and vendor boundaries.