An AI finding without a source is just a rumour
Why we drop findings whose evidence reference does not resolve, and why a dropped finding is a cheaper mistake than an unverifiable one.
Working notes on construction project controls, schedule risk, and the AI that is actually useful on a jobsite.
Why we drop findings whose evidence reference does not resolve, and why a dropped finding is a cheaper mistake than an unverifiable one.
Most weather delay claims fail on arithmetic, not weather. Here is the baseline, the threshold, and the float test that decide whether a lost day is compensable.
The last month is too late to reconstruct a project's evidence trail. Build the closeout map while equipment is installed and decisions are still fresh.
A practical way to turn a folder of drawings, specifications, addenda, scope exhibits, and schedules into a package a reviewer can actually trust.
An XER loading without an error shows that the receiving tool can parse the export. It does not prove semantic completeness, schedule integrity, or broader compatibility.
The tables, the record structure, and the five documented silent drops — baselines, risk data, past-period actuals, secure codes, and resource curves.