Proof of Why.
Compliance decisions get questioned long after they are made. Proof of Why is the idea that every one of them should answer for itself: who decided, on what evidence, and why.
Escalated because the funds reached a sanctioned cluster in three hops, scored 87 under rule R-114, and confirmed by J. Doe on 10 September.
The question always comes later.
By the time someone asks why, the analyst may have moved on and the evidence may sit in eight different tools. The answer has to already be in the record.
“Why was this customer accepted, and this transfer allowed?”
“Show me that the control ran, every time, the way the policy says.”
“Who knew, when did they know, and what did they do?”
Three jobs, kept apart.
Machines are good at searching and at consistency. People are accountable. Proof of Why depends on never blurring the two.
- Follows funds across the chain
- Gathers screening, history and exposure
- Drafts the narrative, marked as machine-written
Set a score or make a decision
- Applies rules you approved
- Ties every point to the rule behind it
- Gives the same answer months later
Weigh context it was not given
- Reads the evidence first
- Confirms, overrules or escalates
- Signs with a name, a time and a reason
Close an alert without reading it
What a proof contains.
Every alert and case in Avus carries the same five layers, from the first signal to the final signature.
- 01EvidenceWhat was known
Screening, history and chain exposure, each with the provider that supplied it.
- 02PathHow the funds moved
The route traced across the chain, hop by hop, with amounts.
- 03ScoreWhat the rules said
The score and the exact version of each rule that produced it.
- 04DecisionWho decided, and why
The call, in the words of the person who made it, with where they differed from the agent.
- 05RecordWhat cannot change
An append-only log, each entry linked to the one before, ready to export.
Four rules we build by.
Explain before deciding
Nothing closes until its evidence has been read.
Label the machine
Anything the agent wrote says so, until a person makes it theirs.
Same input, same answer
Scores are deterministic, so a decision can be rerun and checked.
Never tidy the record
Entries are added, never edited, so what you read is what happened.
Ask us why.
Bring a real case and we will show you the proof behind it, from the first signal to the signature.