What a proof of concept still needs before production
A proof of concept can answer a focused question: can this data be retrieved, can these systems exchange an event, or can this workflow produce a useful draft? Moving...
Always building.
Always thinking.
Practical notes on discovery, architecture, applied AI and the work of getting systems into service.
38 articlesPage 1 of 7
Correcting manufacturing data without losing the original record
When a measurement has the wrong unit, correcting the current value is only part of the job. Someone investigating it later needs to see the imported record, the...
An AI tool approval should name the exact action
A reviewer approving a task needs to see its destination, contents and intended effect. If those details can change after approval, the execution may be authorised by...
A useful presales handover starts with the unresolved questions
A delivery team should be able to read a presales handover and identify what the customer expects, what has been tested and which decisions still need an owner. Those...
When a new constraint changes the proposed architecture
A technical proposal needs to explain why its recommended option fits the customer's constraints. Presales Workbench makes that reasoning inspectable: change a...
Evaluate retrieval before judging the model's answer
A fluent answer can hide a retrieval mistake. The assistant may have selected a draft policy, missed a contradictory source or found a passage that shares vocabulary...