PSD.16:5.2 - A non-contested advice question with a moving evidence basis

A committee asks a separate advisory team for a ninety-day service-development comparison. No disagreement about the question is reported. Internal development and a human–AI support arrangement are live candidates; the organization’s allocation and AI evaluation results remain specialist inputs.

During preparation, the system owner changes the proposed AI configuration from A to B. The evaluation already supplied concerns A. Meanwhile, the allocation analysis can still answer an independent question about internal development. The original plan waits for one whole “evidence package” and then issues one complete comparison. Its all-or-nothing join either delays useful work or tempts the adviser to pass A’s result into B’s claim.

The team considers keeping a historical-A comparison, waiting for every B premise before any return, and separating the returns by their actual dependencies. The first does not answer the committee’s now-B question. The second remains possible but withholds the independent allocation result. Within its agreed control of analysis and delivery, the team retains the qualified internal-development contribution and keeps the cross-candidate comparison open wherever B is material. It considers a B qualification only if an obtainable result could change the recommendation within the receiving window and warrants the acquisition burden. Otherwise it returns the useful internal-development result and the unresolved B limitation.

The arrangement names the configuration owner who reports a further change, the specialist result each claim can consume, the person assembling the comparison and the condition for closing it. It does not freeze the operating system, prevent its owner from changing B, or turn the available internal-development result into the recommended direction.

The test substitutes another A-only result at the receiving join. It must remain an A contribution or an unusable input for the B claim, while the independent contribution stays usable. A later adequate B result can close that particular gap. A further change of subject or horizon reopens only the affected use.

This is reconciliation without a disagreement workshop. The decision changes how qualified results can be returned together; it supplies no AI safety, human learning, organizational feasibility or deployment authority.