Pilot Purgatory

The pattern is consistent across industries. A team runs a successful AI pilot — the demo impresses the steering committee, the numbers are strong, the use case is validated. Then the project enters a prolonged review cycle. IT raises integration questions. Security raises data residency questions. Legal raises liability questions. The original sponsor moves on. The next quarter's priorities shift. The pilot is extended, then extended again.

Eighteen months later, the technology has moved on, the team that built the pilot has partially turned over, and the organisation has spent significant budget on something that never went live. A new initiative is proposed.

Pilot purgatory has technical causes and organisational causes, and both must be solved simultaneously. On the technical side: pilots are often built without addressing governance, system integration, or production operations — the things that actually block go-live. A chatbot that works in a demo environment is not the same as a system integrated with SAP, RBAC tied to Active Directory, and a monitored production pipeline.

On the organisational side: pilots are approved as experiments, not as production projects. There is no named owner accountable for the go-live milestone. There is no fixed scope that defines what "done" looks like. Procurement and security reviews are initiated late, when they should be part of the initial scope.

The direct cost is the budget spent on a pilot that never produces a return. The indirect cost is harder to measure: the operations team that keeps manually processing invoices while the AI decision drags on, the customer service team still handling queries the AI could resolve, the competitors who shipped a production system while the organisation was still reviewing.

There is also an organisational cost. Every failed or stalled AI initiative makes the next one harder to fund and harder to drive to completion. Boards become sceptical of AI investment. Teams become reluctant to sponsor new projects.

The exit from pilot purgatory is a fixed-scope production engagement — not another pilot. This means defining the go-live milestone and its criteria upfront, naming an owner accountable for delivery, and scoping governance, integration, and production operations as part of the initial project, not as a follow-on phase.

Organisations that have shipped production AI consistently cite two factors: a named technical owner who stays engaged past go-live, and a clear separation between "proving the concept works" (which should already be done) and "making it production-ready" (which is the actual work).