Full Stack Data EngineeringCommercial judgement and delivery leadership

Decide when to stop an AI project

PK
Pankit Kumar
Sr. Data Scientist at Parexel (a Goldman Sachs–backed company) · 20 September 2026 · 2 min read
Technically reviewed by Ishaan Sharma
In this article (4 sections)

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:

python
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.

Reference: GOV.UK Service Manual: Discovery.

PK
Pankit Kumar
Lead Instructor, NeuraPath Academy

Pankit Kumar has 10 years in Data Science & AI, building and shipping production systems in regulated pharma and clinical environments. He is a freelance trainer at Boston Institute of Analytics, AnalytixLabs and Scaler, and has taught this material to thousands of working professionals.

This article is part of our FDE for Professionals programme — 16 weeks (proposed). An accelerated advanced pathway for IT professionals ready to own enterprise AI delivery.

Explore FDE for Professionals
Counselling is free · no obligation

Not sure which programme fits?

Tell us your background and we will map it to the right entry point — including saying so when a cheaper programme is the better fit. A counsellor replies within one working day.