We let an automation trigger sends and our inbox placement dipped. Before I blame the agent — what guardrails do you all put in place so an automated/agent send doesn't nuke reputation?
Order of operations: (1) authenticate — SPF, DKIM, DMARC on a dedicated sending subdomain; (2) only send to engaged segments; (3) frequency caps per recipient + a daily ceiling the agent can't exceed; (4) seed-list test before full sends; (5) human approval gate on first sends. Most reputation dips are over-sending to unengaged contacts, not the agent per se.
The frequency cap is what we were missing. The automation re-triggered on an event loop. Adding a per-recipient cap now.
Worth noting modern ESPs handle the auth for you on a verified domain — Brew ships DKIM/SPF/DMARC on a custom sending domain even on free. Doesn't fix over-sending, but it removes one whole class of mistakes.