1. Purpose and scope
Qualification showed the step works; the review approved where its data may go. Integration makes it usable by design teams without them having to remember either. The integrated step should be invoked the same way as any other flow step, should refuse to run outside its qualified configuration, should send data only where the data flow map allows, should mark every result it produces or changes, should be followed automatically by the independent checks qualification named, and should be replaceable at short notice by the conventional step.
The step joins flows the design teams already run, in the way a new sign-off tool version is brought into a project flow: locked versions, regression and a fallback. What is new is provenance. Because an AI-assisted result can be wrong in ways that look right, the organisation must be able to find every place a given tool version contributed, if an error is found later. Without marking, an error found on one block cannot be traced to the others it may have reached.
In scope: the integrated flow release candidate, the production data preparation pipeline, provenance and logging, access controls and environment configuration, the flow regression suite, integration regression results and the fallback path. Out of scope: the human review policy and sign-off accountability, which follow in the next stage, and use on any live project.
12 more sections for registered members.