From prototype to production: the checklist nobody budgets for
Your team built something impressive in a week. Here is the unglamorous list that stands between that demo and software your customers can touch — and why to budget for it before you start.
July 2026 · 7 min read
Here is a scene I see constantly. A team builds an AI prototype in a week. It is genuinely good, everyone in the room is excited, and there is a quiet assumption that production is now a formality. It is not. The demo is the easy 80 percent. The 20 percent that is left is where the time, the cost and the risk actually live, and it is almost never in the business case.
The cliff after the demo
A prototype and a production system are built to do different jobs. A prototype is built to prove an idea, so it takes every shortcut that helps it prove it faster. A production system is built to survive real users, real data and real scrutiny, which means undoing most of those shortcuts. Treating the second as a small step after the first is how a brilliant Friday afternoon becomes the tool people vaguely remember piloting.
The checklist
The gap is not a mystery. It is this list, and each item is a real piece of work:
- 1Security and access
Single sign-on, permissions that mirror your real access controls, and clear data boundaries. The demo ran on an export; production has to respect who is allowed to see what.
- 2Evaluation
A measured answer to "is it actually right?" — a test set, a scoring method, and a threshold below which it does not ship. Impressive is not the same as reliable.
- 3Monitoring and observability
Live visibility of cost, latency, failures and drift. If you cannot see run-cost by team and use case, it will creep, and you will find out from the invoice.
- 4Governance and sign-off
A named owner, an audit trail, an approvals path, and defined behaviour boundaries — the things a risk, compliance or security team needs before they will put their name to it.
- 5Real integration
The system has to act inside the tools your people already use, not in a sandbox. Copy-and-paste is a demo; a connector into your live systems is production.
- 6Handover and documentation
Code, pipelines, prompts and docs your own team can read, run and extend. If the only people who can operate it are the ones who built it, you have bought a dependency, not a system.
Why the business case misses it
The prototype is visible, exciting and cheap, so it gets funded. The production layer is invisible until you need it, unglamorous, and where the real days go, so it gets forgotten. Then the launch slips, the risk committee asks a question nobody prepared for, and the programme stalls at exactly the point it was supposed to start paying back. The fix is not heroics later; it is honesty up front, in the number.
How the timeline compresses
None of this means production has to take quarters. Most of the checklist is the same work every time — auth, evals, monitoring, governance patterns, security sign-offs — which is precisely what accelerators exist to carry from one build into the next. With that groundwork reusable, suitable products go live in days, and the harder ones in weeks, because you are not re-solving the production layer from scratch on every project.
Budget for it before you start
If you take one thing from this, take the order of operations. Cost the checklist at the same time you cost the prototype, not after it impresses everyone. Decide who owns each line before the demo, not when the risk committee asks. Do that, and the cliff between a working demo and a live system stops being a surprise and becomes what it should have been all along: a scoped, costed, ordinary piece of engineering.
Questions we hear next
With the right accelerators, suitable products go live in days, and the harder builds take longer but still weeks rather than the quarters a from-scratch project runs to. The variable is not the model; it is how much of the checklist above is already in place.