01what it is
a classifier and decision layer for account qualification. it turns messy company evidence into one defensible verdict, then hands the maybes to a human.
two input flows — company lists and linkedin engagement — land in one shared company table and go through the same classifier.
it answers “is this company worth a rep’s time?” it does not manage deals, sequences or relationships.
dev-signal evidence, an ai verdict, then the safety layer that can only push toward review — never a silent no.
02system at a glance
evidence goes in; a verdict and a funnel come out; the maybes go to a human. the strictness is the feature.
03the decisions that matter
the rules are the product. these are the ones that keep the verdict defensible.
the dev-signal score is evidence for the ai, not the answer itself. the rules decide after the model speaks.
weak evidence becomes review, never no. a review costs 30 seconds; a wrong no deletes a real prospect.
no invented domains, no weak auto-merges, no guessing between two companies with the same name.
the ai suggests, the rules decide, and the admin-editable policy list has the final say — backwards too.
each source owns its fields. every write is logged and every upload can be undone safely.
seen-vs-stored counts, unresolved people reported, failures flagged instead of silently dropped.