Know what you're missing before an auditor does.
AYA Coverage takes the rules you've authored, reconciles them into one corpus, and checks that corpus against the list you said you needed. You get back exactly what's missing, what drifted to the wrong version, and what's extra, ranked required-first, with a completeness verdict that only turns green when nothing required is absent. Not a green checkmark you have to trust. A worklist.
Book a coverage audit See a sample gap report
THE PROBLEM
Everybody claims coverage, and almost nobody can defend the claim. The honest state of most compliance corpora is a pile of authored rules, a spreadsheet that says "we've got it", and no mechanical way to prove the two match. So the gaps stay invisible until the worst possible moment, when a customer, a board, or a regulator asks "are you actually complete on this market" and the true answer is "we think so". More than half of organizations don't even have an AI inventory yet, and as of August 2, 2026 the EU AI Act turns "do you know what you have, and is it complete" into a board-level obligation, not optional.
HOW IT WORKS
AYA Coverage runs two steps, in order, and it fails closed. First it aggregates your authored rule trees into one reconciled corpus, deduping and fingerprinting as it goes (so a rule that silently dropped out of the pile gets caught, not swallowed). Then it diffs that corpus against a manifest you declare, the expected-set, the list of what you say this market actually requires. The output is the gap, broken into present, missing, drifted (there but the wrong version), and extra. The completeness flag is true only when zero required items are missing, so a partial answer can never read as a finished one.
WHAT YOU GET (the gap report)
- present: what your corpus actually covers, confirmed against the declared list, - missing: the ranked worklist, required items first, so you know exactly what to author next, - drifted: items you have but at the wrong version, the quiet failures that look like coverage, - extra: what you're carrying that the declared list never asked for, - and one fail-closed completeness verdict, green only when nothing required is missing.
A report your curation team can work down and your board can actually read, not a dashboard you have to take on faith.
WHO IT'S FOR
RegTech and compliance platforms curating regulation corpora (this is where we start, with our design partner). GRC and model-risk teams who have to show a board an honest AI-and-rule inventory. Anyone whose "we cover X" has to survive a diligence review instead of a friendly nod.
PRICING (the ladder)
- Audit: one coverage run on your corpus against your declared expected-set. You see the gap worklist and the completeness verdict before you commit to anything. - Subscription: continuous coverage as your corpus and the rules drift, re-run on a cadence, surfacing new missing and drifted items as they appear. - Platform: the fail-closed coverage gate wired into your curation pipeline, so an incomplete corpus can't ship as complete.
Prices are hypotheses we're validating, not commitments. We charge for a real coverage answer, not for a login.
PROOF (we run it on ourselves)
This isn't a slide. AYA runs the exact same coverage audit on its own pattern library, it aggregates our authored pattern trees into a corpus, then gap-checks that corpus against a declared expected-set of the patterns we say we need. The operation we'd sell you is the one that tells us where our own library is incomplete. The deliverable pattern is witnessed on disk, and the fail-closed gap handler behind it shipped and ran locally (the coverage gap-checker in our recent build wave). We measure our own completeness the way we'd measure yours, against a declared list, failing closed.
HONEST NOTE
Coverage is completeness against the list you declared, not against the entire legal universe, and we're deliberate about that. If your expected-set is wrong or thin, a green verdict just means you're complete on what you asked for, and choosing the right expected-set is on you (we can help, but the manifest is the input, not the magic). The pattern and its fail-closed handler are real and witnessed on AYA's own corpus today; a live run on your data needs your corpus, your declared manifest, and our mesh booted, which is the creds-and-data step we'll do together in the audit. We'd rather tell you that up front than sell you a checkmark.
*This page is a specification. The capability it describes is not built yet, and nothing here is a claim that it runs today.*