Job postings are negotiated documents. Most of the language is boilerplate. But each one usually contains a sentence carrying specific operational anxiety, and that line tells you more about the hire than everything around it.
Find it. Scan past collaboration-and-impact language. Look for the verb phrase attached to a consequence: balancing speed against risk, deciding when output is ready, retaining human oversight. Someone wrote that because the cost already showed up.
Use their words. "Deciding when an AI output is ready, needs review, or should not be sent" is how the buyer describes a supervision problem. Pull the phrase into your outreach. Your framework vocabulary earns credibility after you've demonstrated you read the posting at the operational level.
Lead with it. Your first outreach sentence names what breaks if this hire fails. That framing survives forwarding. Your credentials land harder inside a conversation already oriented to the right problem.
Caveat: some postings are aspirational wish lists, not scars. Treat the fear sentence as a working hypothesis you act on while verifying.
Fear sentences from this cycle:
Gusto: "deciding when an AI output is ready, needs review, or should not be sent" — automated service completing a consequential action without human judgment
Capital One: "balancing speed against risk" across an org-wide AI rollout — tooling deployed to hundreds of designers without governance
HackerOne: "humans retain oversight, accountability, and strategic decisions" — agent acting on a security judgment before a human intervenes
Amplitude: "quality to erode unless Design creates coherence without turning into an approval bottleneck" — cheap AI-generated interfaces destroying product coherence
If you can't find it: the posting is too generic to act on, or the real mandate is hidden. Open with a question about what they're actually solving.