Sourcing
Score Exa matches, then a human ships lemlist
Type the job req in Clay. Cap the Exa people search. Clay AI scores must-haves. Review the flags. Only then does the lemlist campaign start.
The loop
- 01Type the req, cap the resultsPut the job in Clay as a natural-language user query. Set default results low (he demos 5) so the first pass is a shortlist, not a dump.
- 02Run Exa from Clay HTTPClay HTTP API hits the Exa Search people endpoint. Poll status until the search completes. Cancel is there if the query is wrong.
- 03Write matches to a second tableCompleted Exa rows land in a People Matches table. That table is the working list, not the query table.
- 04Score must-have vs good-to-haveA Clay AI column audits each row against the original requirements. Strong-fit is common. The column exists for the misses.
- 05Review the flags, then start lemlistDo not auto-push. Walkthrough: the scorer dinged Audible for no FAANG, but Audible is Amazon. A human catches that. Qualified rows go to a lemlist campaign for LinkedIn plus the rest of the channels.
Most agency stacks treat a people-search as a send list. Deep’s version is score-and-hold. Exa finds. Clay grades. A person still decides.
The interesting constraint is the hold, not the HTTP call. The walkthrough miss is Audible counting as FAANG because it is Amazon. That is why the campaign does not fire itself.
