Trusted by 200+ compliance teams
What point-in-time checks cannot see
A flag that says “review” and nothing else
Analysts reopen the case and hunt for whatever tripped it.
Every session judged alone
The same face returns under a new name next week and passes again.
Checks stop at the ID
Readable, matches the selfie, not on a watchlist. A quality fake clears all three.
Three questions, every session
Is the document what it claims to be? Does the identity hold together? Has this person been here before, under any name? Every session, in seconds, no added step for the user.
CLEAN or FRAUD. Nothing in between.
A verdict, not a number to interpret. Every signal cites the field, image region, or prior session it came from. Your team makes the final call.
Every signal cites its evidence
Not "document anomaly." The specific field, image region, or earlier session. Reviewers confirm instead of re-investigating.
It only raises what a real analyst would raise
Calibrated on how legitimate users behave. Glare, a retry, a transliterated name, a diaspora naming convention: none of these is a flag. Sherlock hunts for what is hidden, not what is unfamiliar.
CLEAN or FRAUD. Nothing in between.
A verdict, not a number to interpret. Every signal cites the field, image region, or prior session it came from. Your team makes the final call.
Every signal cites its evidence
Not "document anomaly." The specific field, image region, or earlier session. Reviewers confirm instead of re-investigating.
It only raises what a real analyst would raise
Calibrated on how legitimate users behave. Glare, a retry, a transliterated name, a diaspora naming convention: none of these is a flag. Sherlock hunts for what is hidden, not what is unfamiliar.
"I've seen this face before, under a different name."
Each session looks clean alone. Sherlock pulls every earlier session sharing a face, ID number, device, or network into one case file and works them together. In production it has tied one face to sixteen sessions across five names.
"These details are each valid. Not together."
A field-matching engine sees valid strings. An analyst sees a biography. Sherlock checks whether the pieces agree with each other. Each can pass validation alone; the combination is the tell.
Contradiction, not unfamiliarity. Naming conventions and transliteration never count against an identity on their own.
What teamsget back
Based on customer-reported outcomes. Results vary by program configuration and policy.
per session review, no onboarding latency
KYC session investigated after submission
less manual review work per session
evidence cited on every signal
“Sherlock Agent caught something none of us saw. The document looked clean, and it still flagged a mismatch in the issuing country that I never even entered.That's the moment I stopped thinking of it as a tool and started thinking of it as an analyst.”
Frequently asked questions
More of the review, handled
Athena Agent
Reviews the full business profile and recommends a decision.
Website Agent
Reviews a business's website and online footprint, returns a decision.
Document Agent
Reads and cross-checks documents against your policy.
AML Agent
Scores every screening hit for relevance and severity, recommends a disposition.