Short answer
Pilots stall for four recurring reasons: the demo ran on clean sample data rather than real messy inputs; nobody owned the rollout once the excitement faded; the system was never integrated with the tools people actually work in; and no baseline was captured, so nobody could prove it helped. All four are organisational, not technical.
The demo problem
Demos are built on representative data. Production runs on the actual corpus, which includes the supplier who still faxes, the customer who writes in three languages in one email, and the file that is a photograph of a screen. The gap between those two realities is where most of the engineering effort actually lives, and it is invisible at the point the decision to proceed is made.
The ownership problem
A pilot has a sponsor while it is interesting. Production needs an owner when it is not. That person needs authority to change a process, not just to request a change — otherwise the system is adapted around rather than adopted.
The integration problem
If using the system means opening a separate tool, adoption decays. People work where their work already is. A system that lives inside the CRM, the inbox or the ticketing tool gets used; a separate portal gets used for a fortnight.
The measurement problem
Without a baseline, the conversation at review becomes a matter of opinion. Someone reports it feels faster; someone else reports a case it got wrong. Both are anecdotes. Capturing time-in-stage and volume before the build is cheap and it is what makes the renewal decision straightforward.
A pilot that cannot be measured cannot be defended, and will be cut the first time budgets tighten.
What to do differently
- Run the pilot on real data from day one, including the ugly cases. Expect worse numbers and better decisions.
- Name the production owner before the build starts, and check they have authority over the process.
- Deliver into the tools people already open, not alongside them.
- Capture the baseline first. It takes days and it determines whether the project survives.
- Scope small enough to finish. One completed workflow changes the organisation's mind; three unfinished ones confirm its scepticism.
The path most pilots take
- Demo on clean data
- Enthusiastic sponsor
- Real corpus arrives
Faxes, photos, three languages
- Sponsor moves on
No production owner
- Separate portal ignored
- No baseline to argue from
- Demo ran on representative data1Run the pilot on the ugly cases from day one
- Nobody owned it after the excitement1Name the production owner before the build starts
- It lived beside the work, not inside it1Deliver into the tools people already open
- No baseline was captured1Measure time-in-stage first — it takes days
4 items awaiting a person, grouped by cause so one decision clears a class rather than a case.
All four are organisational rather than technical, which is why better models do not fix them.
Last updated 2026-07-10 · Koderead
Related
Apply this to your own business
Eleven questions, scored in your browser, with the arithmetic shown.
Find my AI opportunities