A prototype makes an idea concrete enough to review. Production requires evidence that it can support the work, the users and the conditions it will encounter.
Real use introduces exceptions.
Test what happens when information is incomplete, a record is missing, an integration is unavailable or a user lacks permission.
For an agent, uncertainty also needs a defined response: ask for clarification, stop or hand the task to a person.
Data and access become concrete.
Decide where real data comes from, who may access it, how long it is retained and which actions are allowed.
Design access around roles and tasks. Reading a document and changing a business record need different permissions, even within the same interface.
Operation is part of the product.
Agree who supports the system, how users report problems and what happens during an interruption. Include a way to recover from a failed change.
Use checks that reflect the intended use and provide evidence for a launch decision.
Make the next scope explicit.
Record what the prototype has established and what remains untested. Integration behaviour, performance and operating requirements may still need work.
Agree on the remaining scope, responsibilities and launch criteria before deployment.
The next build should answer those open questions and provide evidence that the software is ready for real use.
Back to Insights