Two values, one decision
The first drawing review found a water heater described as 150 litres in one place and 300 litres in another. Choosing the larger number would have made the spreadsheet neat. It would not have established which equipment belonged in the project. The useful output was a coordination question referencing both sources.
Give uncertainty a location
There is a difference between saying “please check the drawings” and identifying a specific discrepancy. A principal-bedroom ceiling appeared at 273 cm on the ceiling plan and 283 cm in elevations. A cabinet width differed between a headline dimension and the sum of its component dimensions. Each issue could be named, located and handed to the person able to resolve it.
Keep different kinds of question separate
Some questions require a design decision. Others require a site survey, a supplier reference or a price. The initial register held 19 questions spanning all of these. A question about an exact appliance template cannot be closed by measuring its illustration. A question about site access cannot be answered by a room plan.
Turn the register into a working habit
A team can add an owner, a target date and a decision reference to each item. A closed item should record what changed and which source now governs. This keeps the register useful beyond the initial review and makes it possible to see whether a later revision reopens a previous assumption.
Confidence should be inspectable
The value of a prepared dossier is not the absence of blank cells. It is the ability to tell which statements are grounded, which are provisional and which depend on someone else. An AI colleague earns a place in the workflow by preserving these distinctions consistently.
See the work behind these observations.
Explore the first assignment