How Retriever works
The intelligence behind every list.
You describe who you want to reach in plain language. The agent decides how to get it, so you never pick tools or write queries yourself.
How the agent thinks
It reads your prompt and your Company Brief, so it knows exactly who counts as a match before it spends anything.
For your rarest, hardest criterion, it compares the available sources and chooses the one most likely to deliver.
It probes candidate sources on a small scale first, and only builds the full table once it knows the source works.
If a source underperforms, it rewrites the query or switches method, instead of handing you a weak list.
It scores every row against your criteria and backs each verdict with sources it can cite.
Every row is judged, with a reason
Retriever never hands you a raw dump. Each row is scored from 0 to 100 against your criteria, with evidence you can open and check, so you can trust each match and see why it made the cut.
Green qualifies, red is out, grey is unsure. Every score links back to the sources behind it.
Across a source, those verdicts add up to a qualification rate: the share of rows that pass. Retriever works to keep it high by choosing the source that best fits your target, which is what keeps your cost per qualified lead low.
Shown on every source, measured on your sample and updated as more rows run.