A polite no beats a six-month pilot with no owner and no metric. Finance teams are under pressure to “do something with AI”. That pressure produces pilots that look busy and never reach the month-end pack. Saying no is not anti-innovation. It is capital discipline.

I still see finance leaders approve AI pilots the way they used to approve innovation labs: interesting demo, unclear owner, no metric that shows up in reporting. The model works in a sandbox; the Monday morning handoff does not. When that happens, the honest move is to decline or redesign — not to extend the pilot because sunk cost feels awkward.

Say no when any of these are true.

No named constraint. If the pitch cannot name whether you are trying to cut close days, aged reconciliations, review hours, cash visibility lag or exception backlog, you are shopping for capability. Capability rarely survives a board review.

No baseline. Without a rough sketch of volume, hours by role and error patterns, you cannot know whether the pilot moved anything. You do not need a perfect activity-based cost model. You do need an honest yardstick.

No owner. Dual ownership (business + technical) or it does not start. A pilot owned only by a vendor success manager is a subscription in waiting.

No human gate. Finance-adjacent work is YMYL territory. If nobody has decided which outputs require approval, which data classes are in scope, and what gets logged, the pilot is unsafe to call “in production” even if the demo impresses.

No stop conditions. Write in advance what would cause a pause or cancel. Endless extensions are how zombie tools enter the budget.

When you do say yes, keep the release thin. One workflow, one team, one metric, two to four weeks beside the old process if risk is material. Review weekly. Capture exceptions in a shared log. If the metric moves and controls hold, document the pattern. If not, stop and explain. A failed pilot you can explain is cheaper than a year of unfocused tooling.

Use a free readiness scorecard or a scoped opportunity audit when the shortlist is real but the organisation is still arguing from anecdotes. The goal is not to delay forever. The goal is to fund only what you can defend in finance language the board already uses.

Public frameworks such as NIST’s AI RMF are useful for structuring risk conversation. They do not replace your chart of accounts, your staffing reality or your auditor’s expectations. Translate them into four one-page rules: ownership, data classes, human gates, logging.

Saying no also protects credibility for the next request. Teams that rubber-stamp every demo find that later, better proposals get the same sceptical shrug. Teams that decline weakly designed pilots earn the right to fund a narrow release that actually lands.

Saying no protects the next yes

Teams that rubber-stamp every demo find that later, better proposals get the same sceptical shrug. Teams that decline weakly designed pilots earn the right to fund a narrow release that actually lands. Use a free readiness scorecard or a scoped opportunity audit when the shortlist is real but the organisation is still arguing from anecdotes.

Translate public frameworks such as NIST’s AI RMF into four one-page rules: ownership, data classes, human gates, logging. They do not replace your chart of accounts, staffing reality or auditor expectations — they structure the conversation so a polite no is easier to defend.

Monday-morning checklist

  • Refuse pilots that cannot name the constraint and the metric.
  • Require a baseline sketch and dual ownership before kickoff.
  • Design the human gate and data classes before any go-live date.
  • Write stop conditions into the pilot charter.
  • Prefer a thin release with weekly review over a six-month theatre project.

Sources

Educational content; not financial, investment or legal advice. Confirm vendor terms and internal policies before approving spend.

Keep the Monday checklist visible: owner, metric, exception path, data the tool may see, weekly review. Blank lines mean pause. Educational commentary only — not financial, tax, legal or investment advice.