About
Most prospecting asks how many names can be gathered. The more useful question is which companies are worth a careful conversation, and why this quarter rather than last.
Answering that properly means reading, for every company: a funding announcement, a job ad, a stack page, what they actually do. Across a market of a thousand, nobody has time, so the reading gets skipped and replaced with a filter. The filter produces a list, the list produces sends, and the sends wear down the reputation the next round depends on.
So the product does the first pass and shows its working. It doesn’t decide for you. It makes deciding quick enough to actually do.
The result is a shorter list, with a record of why each company is on it.
We were describing the product’s output with a word that implies somebody had shown interest. Nobody had. A penetration-testing firm replied to our outreach and said so, in about one sentence, and there was no version of the argument where they were not correct.
So the site changed: the word, the positioning under it, the claims that leaned on it, and the pages that had overstated what the product can observe. This version of the site is the result.
The boundary, written outEvery expensive step sits behind a cheaper one. Rules throw out clear misfits before any AI runs, an AI review checks the rest, and deep research goes only to companies you release.
You don't need another tool to babysit. The work goes into setup, so day to day you get a queue of decisions you can make quickly, with the reasoning already assembled.
Every outbound message reaches a real person, under your name. So research and review come as standard, and the sending protections are on every plan.
Every score, shortlist, research field and draft carries enough reasoning for a careful reader to disagree with it. A system you can't argue with is one you have to take on faith.
These are behaviours the product will not have. For the separate question of what can and cannot be known about a company from outside it, see what we can’t see.
We don't sell your data.
We don't sell your data or make money from it. Your research is your own copy and no other customer sees it. Only two records cross workspaces: a list of companies that asked us to delete their data, and an anonymous count of common email formats that names no one.
We don't claim to see intent.
Nobody can, from outside a company, before contact. Saying so cost us a positioning we liked.
We don't promise replies.
We don't publish a reply rate yet either, because we haven't sent enough email to have one worth trusting.
We don't run a black box.
Every score carries its reasons, including the discards, and an inference is labelled as an inference.
We don't bill by usage.
You pay a flat monthly fee for research up to your plan's limit. There are no top-ups and no overage charges.
We don't optimise for volume.
The allowance is a ceiling on our research spend. It exists so we never have to pad a list to hit a number.
LeadAtomic is my work. I designed the pipeline, wrote most of it, and I am the person who answers when you write in. There is no shared inbox with a rota behind it.
I mention that because it changes what you are buying. Founder-run software is a bet on judgement, and judgement should have a name attached to it. It also means the limits on this site are limits I chose to publish rather than ones a marketing team missed.
If something here doesn’t hold up, tell me. The last person who did that got the whole site rewritten.
Questions about the product, the pricing, or whether this fits what you do. Not a ticket number.