What Enakt is for

Software can already read every filing, docket and sanctions list that matters. What it cannot reliably do is tell you which parts of what it just read are actually true — and, harder, which parts it never managed to check at all.

That second gap is the expensive one. A tool that returns a confident paragraph built partly on a source it could not reach has not saved anyone work; it has moved the checking to someone who no longer knows where to look. Enakt exists to close that gap rather than paper over it.

Everything the product outputs is one of exactly three things, and the distinction is a contract rather than a presentation choice:

Verified. Proven against a primary source — the filing, the official list, the docket — with that source attached, so you can reopen it months later and land on the same record.

Sourced. Quoted from a named document, and not verified. It tells you what a filing said; it does not tell you the statement is correct. We keep these visibly separate from verified facts because collapsing the two is the single easiest way to make a report look stronger than it is.

A gap. We could not reach it, it was out of scope, or it does not exist in a source we cover — and the report says which, by name. A gap is a result. It is the part a reviewer is accountable for, and hiding it is the failure this company was started to stop.

0 in 64 claims re-verified against primary SEC sources carried a silent error — a 95% upper bound of 5.7% on the rate. That result covers value, period and computation. It does not establish that we always pick the right concept or tag scope, and we do not let our own copy imply it does.

Our name matcher was measured against 20,746 real sanctions designations, including the case it handles worst — a spelling the published list does not itself carry, where recall is 27%. That number is on the reliability page next to the flattering one, because a screening tool that only publishes its good measurement is telling you something about its judgement rather than its matcher.

We will not estimate a figure and present it as retrieved. We will not answer a coverage question with “available on request” — our sub-processors and the personal data we touch are named on the security page. We will not publish a benchmark we have not run, or a usage number we cannot point at a record for.

These are stated as commitments because they are the ones that are tempting to break, and because a reader can hold us to them from the outside: every figure this site publishes names the measurement behind it, and any that does not is a bug worth writing to us about.

Enakt is early, and says so. The product is live and you can run a screening or a diligence brief today without talking to anyone — no card and no call, which exists so the artifact can make the argument instead of a deck. What we are looking for at this stage is people who will run it on a subject they already know cold, and tell us where it fell short.

The direction beyond today's product is a straight line: retrieve, verify, act, land. Retrieval and verification are what ships now. Acting on a verified result — and reconciling what landed against what was intended — is the part that makes this infrastructure rather than a report generator.

The contact page reaches us directly. If you are evaluating Enakt for regulated work, say so — the questions that come with it are the ones we most want to be asked early, and the enterprise page covers what changes at that scale.