Eleven pilots. Zero in production.
The pilots keep working. That is the maddening part. Each one demos well, earns applause, and then dies in the gap between working and approved: security review, data governance, operational ownership, budget lines nobody assigned.
By the fourth one, the organization has learned a quiet lesson you never meant to teach: pilots are theater. The next real ask will be met with a sigh.
The question is not whether the pilots work. It is why working is never enough.
Pilots born governed do not stall at review.
On Boxcar, the pilot is built inside the governed lifecycle from day one: intent, autonomy limits, decisions, and evidence captured as the work happens. The artifacts review asks for are generated with the pilot, not reconstructed after it, so the gap between working and approved closes while you build.
And the value is measured from the first week, which changes the ending. The readout is not a demo asking for faith; it is a number asking for a decision.
An emergency department pilot reached the front desk.
The ask was triage by non-clinical staff, which is exactly the kind of pilot that dies in review: patient data, clinical risk, legacy systems with no API. Built governed from day one, the review materials were the build materials. It crossed, and it is in production today, PHI-safe by design, with every patient outcome feeding the roadmap.
Exactly. That is why the fix is not a better demo. Ownership, limits, and evidence are the organizational questions, and the governed lifecycle answers them in writing while the pilot runs.
Make the twelfth one land
One workflow, five weeks, explicit autonomy limits, measured value, and a readout built for the people who say yes. You keep every artifact either way.