You spent an hour, maybe two weeks, refining your search. Director versus VP. Headcount bands. Which industries to exclude. Every one of those decisions is sitting in a chat window right now.
That's a problem, and it's the same problem as keeping it in your head. Next Monday's scheduled job can't read your chat and it can't read your head. It reads pages.
So in this episode we write it down. One request creates a My Market page next to the Who We Sell To page in the CRM: the Apollo translation of the profile in plain English, titles, seniority, geography, company size, the exclusions, the signal we're using and why.
Then two standing rules, which are the part people skip. Only verified emails count as send-ready. Never spend more than five credits without asking me. Show me the page before you save it. Your numbers will be different, and that's the point. Maybe your rule is never spend anything. You pick, and you write it in your own words, because there are no magic prompts here. Intent and outcome are what matter.
With the page saved, we turn the robot on. Check my open deals count. If it's under 10 and I have fewer than three waiting targets, search Apollo using My Market, show me the count, enrich the five best fits, and load them into Targets as Waiting with source set to Apollo and emails on their pages. Follow the rules on that page.
And then the best moment in the episode: it refused to run. I had 19 waiting targets, the condition said fewer than three, so it passed and spent nothing. That's the guardrail working on camera. Had this been the weekly job, it would have quietly done nothing, which is exactly right.
I pull the gate out to demo it anyway, approve six credits (five enrichments plus one for the advanced-filter search), and it comes back with three of five. Two had no Apollo data, so it didn't invent them. It never re-asked about the profile or the budget, because both were already written down. One approval, then done.
Your assignment: build the market page, rules and credit allowance included. Ten minutes. You can read the rules out loud and let Claude draft them. Then run it for real and spend a few credits to see what comes back.