Core insight: AI outputs can sound credible while being unsupported. The Hallucination Triage is a compact, actionable ritual you can run in under ten minutes to decide whether an AI answer is safe to act on. You get three sequential checks (Source Match: confirm original evidence; Consistency Probe: cross-check with an independent data point or rule; Impact Gate: ask ‘if wrong, what breaks and who pays?’), a paste‑ready Provenance Stamp to attach to accepted outputs, and two constrained assistant patterns to automate the checks without adding friction. The briefing includes three short examples (vendor claim compression, competitor summary, model recommendation), enforcement moves to block downstream automation until the stamp exists, and a 7‑day pilot to triage five recent AI outputs. Outcome: keep speed while preventing plausibly sounding errors from becoming commitments. Close with a single fast action: run the triage on the next AI output you plan to use and subscribe.