An AI result is only useful if you can trust it and act on it. Flaggity builds every finding around three elements that exist precisely so the result holds up under scrutiny: the standard, the citation, and the severity. Once you know how to read those three, every inspection result becomes a clear to-do list instead of a wall of text.
Standard
Each finding is mapped to a recognized standard — for example a clause of AWS D1.1 for structural steel welding, ISO 5817 for weld imperfection levels, or OSHA 29 CFR 1926 for construction safety. The standard tells you which rulebook the observation is being judged against, which matters because different work is governed by different codes.
Citation
The citation is the specific reference within that standard. This is the single most important part of a finding. It's the difference between 'this weld looks bad' and 'this weld's undercut exceeds the limit in clause X' — the first is an opinion, the second is documentation an auditor or customer will accept. When you generate a report, the citation travels with the finding.
Severity
Severity tells you how serious the finding is so you can triage. A high-severity finding generally warrants a formal Non-Conformance Report and tracked corrective action. A lower-severity item might be something you monitor or correct on the spot. Severity is how you turn ten findings into a priority order in seconds.
Flaggity assists — people decide
Flaggity is a documentation and consistency layer, not a replacement for a qualified inspector or engineer. The AI's job is to surface and cite findings the same way every time. The final call on whether work is accepted always belongs to a competent human.
What happens next
For each finding you have three moves: capture it in a report or certificate, escalate it to an NCR if it needs to be corrected and tracked, or re-inspect after a fix to confirm closure. A typical result mixes all three — clean items get reported, failures get NCR'd, borderline items get a judgment call.
Common questions
Why did two similar welds get different findings? Genuinely borderline cases can read differently run to run, the same way two human inspectors might differ on an ambiguous joint. Decisive defects are flagged consistently; the gray area is where your judgment matters most.
Can I change or override a finding? You decide the disposition. The AI's finding and citation are the documented input; your decision as inspector of record is what closes it out.