Reverse the AI in corporate email. For every message Defluff guesses what prompt the sender gave an LLM, classifies urgency (ACTIONABLE / RESPONSE-NEEDED / FYI / NOISE), and extracts 3–5 bullets of real intent. Handles single emails, threads, and batch triage. Detects invoice fraud / BEC, phishing, fake recruiters, fake interviews, conference scams, and crypto pitches with specifically named red flags — lookalike domains, fake forwarded approvals, urgency + payment redirects.