Lead groups and scoring

Discovery without organization becomes a pile. Lead groups hold the people an agent found so you can review, segment, and attach campaigns without mixing every experiment into one list. Separate groups by offer, geography, or motion such as classic ICP versus Steal Customers.

Deleting an agent does not have to mean throwing away the people it already found. Groups and leads can outlive a single agent configuration when you keep them on purpose.

Scoring helps you rank who looks closest to the buyer you described. It is a prioritization aid, not a guarantee that the top row will buy. Use scores to decide manual review order and campaign priority. Override them when you know the market better than the model.

Weak ICP definitions produce confident but wrong scores. Fix targeting and My Product when the ranking feels consistently off.

Early in a motion, skim leads before you open the fire hose. Ten minutes of review often saves a week of wrong outreach. Later, when patterns are stable, you can trust automation more on well-defined segments.

MCP and REST tools can list and fetch leads, including engagement context for Steal Customers. Operators should report empty results honestly and check activity before inventing prospects that were never discovered.

Keep groups boring and specific: one ICP, one Steal Customers source set, one experiment. Mixing every lead into a single pile makes it impossible to tell which motion produced the reply. Name groups after the buyer or the source, not after the week you created them.

Use scores as a review order, not as a send trigger. A high score still needs a human glance for obvious mismatches. If you cannot explain why a lead scored well, the score is decoration.

A score is not permission to skip reading the profile. It is not a forecast of revenue. And it is not a reason to keep a bloated group you never prune. Groups and scores exist so you can send the next honest batch, then delete the rest.