Decide when to stop an AI project

Stopping is a delivery outcome when evidence says the project cannot reach an acceptable result under its constraints. Continuing to justify previous spend converts a pilot into an uncontrolled commitment.

Build the decision artifact

Agree stop criteria during scoping: lawful and approved data access, critical quality floor, safe operations, credible value and accountable ownership. Review at fixed gates. Estimate only future cost and opportunity, not sunk effort. Offer redesign or pause only when a specific new fact could change the failed gate.

The commercial leadership lab makes the artifact inspectable with authored inputs:

from leadership_cases import stop_project_case

result = stop_project_case()
assert result["failed"] == ["data_permission", "credible_value"]
assert result["decision"] == "stop"
assert result["sunk_cost_considered"] is False
assert result["reusable"] == ["evaluation set", "adapter contract"]

Data permission and credible value fail in the fixture, while quality, operations and ownership pass. The decision is still stop. The labelled evaluation set and adapter contract remain reusable without pretending the original business case survived.

Protect the decision from weak evidence

Teams often move the threshold, narrow the dataset or rename the objective. Preserve the original charter and record any new project separately. A stop memo should state limitations, affected stakeholders, data disposition and which commitments are being closed.

Keep these artifacts for review:

• original stop/go charter
• gate results and unresolved dependencies
• future-cost and opportunity analysis
• closure, data disposition and reusable-asset record

This practice aligns with the discovery, productisation, client enablement, technical leadership and capstone sequence in the FDE for Professionals course. The course link describes the pathway; the local scenario is fictional and does not claim a client engagement, investment result, hiring decision or certificate.

Practice task

Define five gates before starting a synthetic pilot. Fail two and write the stop memo without proposing extra work unless it changes a named fact.

Continue learning

This article is part of the Commercial judgement and delivery leadership sequence. Use the neighbouring tasks when you need the prerequisite or the next application.

• Review the prerequisite or neighbouring task in Measure the support burden of client-specific customization.
• Continue with Lead a blameless review of an integration incident.

Reference: GOV.UK Service Manual: Discovery.