Brisc matches incoming premium with a ladder of deterministic rules, run in order, each narrower than the last. Every match records its rule and its confidence; nothing is matched by guesswork.
Strongest evidence first: a payment reaches a looser rule only after every stricter one has declined it.
Bank fees, charges and internal transfers are auto-cleared, so the queue only ever holds real premium.
The bank line, the broker’s remittance and the expected premium all agree; the match records all three.
Matched on absolute values, so money going back out finds its policy the same way a receipt does.
Works whichever broker sent the money, including the aggregate case where a remittance header total equals the sum of the obligations it covers.
A payment in one currency matched to premium in another, with both currencies and the rate recorded on the match.
The workhorse: payment straight to policy, and the only rule set that enforces payment dates, so on-time and late mean something.
When the cash is short of gross by the commission, the match says so explicitly. The shortfall is named, never silently tolerated.
OCR typos, truncated policy numbers, near-identical broker names, and a reverse lookup that finds a policy reference buried in raw bank narrative.
Last, and deliberately strict: it fires only when amount and broker point to a single unique candidate.
Every rule set carries a confidence tier that reflects how much evidence it demands. Anything less than certain goes to review.
A match records the rule that made it, and the tier it made it at.
The model argues borderline cases in the review queue; it never decides a match.
We publish what every rule set does; the thresholds underneath are part of the product.
It stops and asks: the payment queues with the Analyst’s reasoning and source documents attached. At steady state that queue is roughly one row in twenty.
Every reviewer decision feeds back as new rules, sharper prompts and confirmed aliases. See bank reconciliation for the three loops.
Brisc never initiates, holds, or moves a payment. It matches, attaches the evidence, and posts to your PAS: your rails, your treasury, your approvals.
Reconciling claims payments going out? See claims payment reconciliation for the four-way match on money out.
Rules. Matching is deterministic, rule-based and explainable: every match names the rule that made it. AI reads the statements and remittances and argues the borderline cases in the review queue; it never decides a match.
The broker-agnostic rule set handles the aggregate case: the remittance header total is balanced against the sum of the obligations it covers, and each line is matched underneath it. One wire, many policies, one evidence trail.
The close-match rule set exists for exactly this: OCR typos, truncated policy numbers, near-identical broker names, and a reverse lookup that finds a policy reference buried in raw bank narrative. Whatever it finds still carries its rule and confidence, and anything short of certain goes to review.
No, deliberately. We publish the behaviour of every rule set; the tuning underneath is part of the product, and some reference data belongs to individual customers. What you always get is the rule and confidence recorded on every match.
About 80% of records match on day one. Over roughly 90 days the rule base absorbs what the model surfaces and the match rate climbs to 92–95%. At steady state the review queue is roughly one row in twenty.
Never. Brisc never initiates, holds, or moves a payment. It matches cash to premium, attaches the evidence, and posts the result to your PAS. Your rails, your treasury, your approvals.
30 minutes, no slide deck. We run your data: a real statement and a real bordereau. Your business is specific; the walkthrough should be too.
Book the walkthrough