Paste a web address. In about eight seconds you get a list of the specific ways somebody who wants to give that business money can currently fail to reach them — each finding attached to the exact line of the site that proves it. No email required, and no language model anywhere in the path.
It would be easy to point a language model at a homepage and have it produce five confident paragraphs. We do not, for one reason: a model cannot be held to a finding. It will describe a chat widget that is not there and it will never say so. Every line this page returns comes from code that either found a thing or did not. The same scanner backs the claims in every cold email this company sends, where a separate verifier re-asserts each claim against the captured data before the email is allowed to exist — currently 283 assertions across 44 businesses, never permitted to pass with a failure in it.
Most of what a scan finds comes down to one thing: outside office hours, and often inside them, there is nobody to answer. The fix is not a bigger form. It is something that picks up, on the phone and in the chat box, that knows what the business sells, what it charges, where it goes and what it will not do, and that puts a booked job in the calendar rather than a message in a pile. It ships in 21 days from the moment we have the material and the access, not from the day the contract is signed.
It does not run JavaScript, so a widget that only exists after a script boots will not be seen — which is worth knowing on its own, because most AI crawlers do not run it either. It reads public pages only, at six scans an hour per address, and the result is not stored unless the visitor asks for it by email.