Résumer cet article avec :
In brief.
In internal AI projects, the POC (proof of concept, the demonstration prototype) is misleading by design: it is built to succeed, on a chosen scope, with clean documents and a cooperative user. A production system is built not to fail: access rights, edge cases, model updates, support, audit. That second life of the project, invisible in a demonstration, concentrates roughly 90% of the effort and the cost.
The anatomy of the 90%
Security and rights: propagating permissions into the search index, checking them at every query, maintaining them as people move internally. Connectors: plugging in document management, email and the CRM, and re-plugging them at every change. Reliability: annotated test sets, coverage and accuracy measurement, benches replayed at every change. Robustness: scanned documents, complex tables, 300-page files — everything the POC avoided. Operations: monitoring, logging, incidents, version upgrades. People: training, support, documentation.
A recurring undertaking, not a one-off task
Rights change every week, models every quarter, regulation every year: building never stops — it turns a project expense into a permanent team. Many organisations discover this in month twelve, when the flagship prototype becomes a maintenance ticket nobody wants to own. A test before approval: ask for the plan and budget for the 90%, line by line. If they fit on a single slide, they do not exist.
The Optivalue.ai approach
That 90% is a software vendor’s business: Optivalue.ai spreads it across all its clients, where every internal project pays for it alone, in full.
Why is the POC misleading?
Because it lives in a scope chosen to succeed; production lives in the cases the POC avoided.
What is the largest item in the 90%?
Measured reliability: annotated sets, evaluation benches, permanent regression testing.
Turn your quizzes into opportunities, right now
30 days free • No credit card required • No commitment
