GTMcraft Plays · The Hard-Kill Pre-Filter Play

Your agent pays to judge rows a rule could have killed for free; these five steps put deterministic kills in front of the model.

Play visual

The 90-second read

Your outbound agent scores every account you feed it. So you pay a model to reason about companies that were never buyable, and the list still needs your eyes. The expensive part of an outbound agent is not the model. It is sending it rows a rule could have killed for free. Put deterministic kills in front of the judgment step and you cut the bill and the false positives at once.

• Rules first, model second: anything you can state as a fact never reaches a probabilistic step.
• Nine hard kills, written from your own closed-won pattern, remove most bad rows for free.
• You report two numbers weekly: rows killed pre-model, and euro cost per run.
• The reviewed list gets shorter and better at the same time.
• The move runs on an Apollo or Cognism export plus one Clay table.

Read time: about 5 minutes. First step runnable Monday morning.

The #1 move this week

Write nine hard-kill rules and run them before any model call.

The model is not where your outbound cost or your false positives come from. They come from feeding judgment rows a rule should have removed. Write the kills as plain facts against your closed-won pattern, run them first, and send only survivors to the scoring step.

Start here: pull your closed-won list and write the first three kills, wrong headcount band, wrong country, no EU entity. You need your CRM export and 30 minutes.

Ignore this week: tuning the scoring prompt. Fix the input list before you touch the model.