Customer risk rating that explains every change in one cited sentence, re-scored the night a fact changes and signed by two people
Every member's rating explained in one sentence, re-scored the day a fact changes.
See one case, screen by screen ↓A customer risk rating is the customer risk profile a bank or credit union keeps for every customer, used to set how closely monitoring watches the relationship and when enhanced due diligence starts. Customer Risk Rating is a solution in which agents re-score a member the night a fact changes, explain each rating change in one sentence whose every clause cites its source and weight, hold type-only and open-activity changes, and route every upgrade and downgrade to a maker and a different checker.
AI use cases in banking: a buyer's guide to agentic AI for bank and credit union operations →
Updated
A customer risk rating is the customer risk profile the FinCEN CDD rule and the FFIEC BSA/AML Examination Manual expect for every customer: it sets how closely monitoring watches the relationship and when enhanced due diligence (EDD) starts. The manual's December 2021 update and the July 2022 interagency joint statement both say that no customer type automatically presents a higher risk, so a rating has to rest on what a customer does, not only on what it is.
Today the rating is usually a scorecard filled in at account opening and refreshed on a calendar. A member whose cash deposits triple in July is still Moderate in December, the reason behind any rating is arithmetic nobody can read aloud, high tiers fill with members whose only elevated attribute is their type, and a downgrade can be signed while an alert is still open. A weight change made in a spreadsheet can move thousands of members with no back-test. Customer Risk Rating puts eight agents in front of that work: they re-score on change, explain, check and draft, and a BSA analyst and a different checker sign every change.
Estimated split of one individual rating change, by hand and with the agents; our arithmetic, kept modest. The 'after' column is the person's reading and deciding time; the agents' overnight run is not counted.
A Watcher finds changed facts, the Rating agent writes the because-sentence, two checkers hold; analyst and checker sign.
Thursday, October 22, 2026, at Norrell Federal Credit Union, a $6.8 billion credit union with 520,000 members. Here is how one night's rating changes reach a decision, screen by screen, in the working solution.
Between 2:00 and 2:41 AM the Fact Watcher found 4,118 members with a changed fact, and the Rating agent proposed 152 rating changes under methodology v4.2: 88 up, 64 down. 133 went into six like-change batches and 19 came to Tamsin, three of them held. A banner says the OFAC screening service answered late at 6:10 AM; 212 members were re-screened at 6:24 AM and nothing was lost.
Kestrel Ridge Supply Co. moves from Moderate to High because monthly cash deposits rose from $22,000 to $71,000 (+14) and wires began to two counterparties in jurisdictions on Norrell's high-risk list (+15).
The cash clause opens the core banking system's cash deposit summary for business checking ••6620: $64,800 in July, $71,300 in August and $76,900 in September, against $25,000 expected. Beside it is the weight in methodology v4.2: rule A-3 cash intensity, in the Activity family, +14 points, validated by Rafael Quint and approved by Helena Marchetti.
Under the sentence: member type gives 6 of 67 points, and without it Kestrel Ridge is still High at 61, so rule G-2 is met. No open alert, case, 314(a) match or pending EDD; the large-cash alert TM-26-0930-4471 was closed on September 30. Confidence is 0.94, above the 0.85 threshold.
It returns to Moderate only if cash deposits stay under $40,000 a month for three months and no wire goes to a listed jurisdiction for 90 days.
Tamsin asks who the two wire counterparties are. The answer is cited: Varnholt Trading Ltd. (list entry HRJ-07), 4 wires for $108,700, and Brisa Coastal Imports S.A. (HRJ-12), 3 wires for $77,700; neither is a Norrell member, and the OFAC screening service cleared both. She approves as maker, and the change waits for a second signature.
Wendell sees the same sentence, sources and checks Tamsin saw: rule A-3 +14, rule G-4 +15, confidence 0.94, both checks passed. He signs second, and the rating is written to the core banking system and the transaction monitoring system. On a batch he prepared himself, the Sign button is locked.
EDD-26-1022-0117 is opened in case management, due November 21, 2026, with investigator Imani Rourke. Four questions are tied to the facts behind them: where the cash comes from, what the wires are for, the relationship to the two counterparties, and what activity to expect from now on.
Tipping-off check passed: no mention of monitoring, alerts, risk ratings, investigations or reports.
Leona Strick is proposed to move from High to Low: her account went dormant after 14 months with no activity, and her international wire service was closed on August 4, 2025. The Hold Checker stops it under rule H-1, because a joint account she shares has open alert TM-26-1009-2281, rapid movement of funds, opened October 9. Tamsin rejects the downgrade with reason code OA, Open activity, and Leona stays High until the alert closes.
Held: your call
Batch B-1022-1 holds 41 consumer members moving from Low to Moderate because P2P payments stayed over $3,000 a month for three months (rule A-7). The agent picks the 5 members closest to the tier boundary; Tamsin reads and agrees with each, then approves the batch once. Any member can be taken out for individual review.
The whole book is tested every night. Rule G-1 caps member type at 10 of 100 points, and rule G-2 says a move into High needs facts from at least one family other than the member profile. Pellham Street Money Services LLC is held today: only its member-profile facts moved, and without them it stays Moderate at 52.
7,140 High ratings tested, 0 resting on type alone.
Wendell drafted v4.3 on October 15: a higher P2P threshold in rule A-7, a three-month cash average in rule A-3, and the September high-risk list in rule G-4. The Methodology Steward back-tested it on all 520,000 members: 6,412 would change tier. Rafael validates it with a memo, and only then can Helena Marchetti approve it.
Five people at Norrell Federal Credit Union who carry one rating change from the overnight run to the exam: what their day looked like, and what it looks like now.
A Watcher finds changed facts, the Rating agent writes the because-sentence, two checkers hold; analyst and checker sign.
Reads the core banking system, transaction monitoring system, digital account opening platform, OFAC screening service and case management each night, and links every changed fact to a member.
Scores the member's five families under the approved methodology, writes the change as one sentence and the reverse sentence, and cites every fact.
Tests each proposed High, and the whole book every night, for ratings that rest on member type alone (rules G-1 and G-2).
Cross-checks each change against open alerts, cases, 314(a) matches, pending EDD and facts newer than the score.
Groups like changes by direction, drivers and evidence, and picks a risk-weighted sample closest to the tier boundary.
On a move into High, drafts the EDD case with questions and documents, and opens it in case management after the second signature.
Back-tests a proposed methodology version on the whole book and writes the validation pack.
Drafts the supervisory committee report and builds examiner and audit packs.
Maker prepares, a different person signs. The agents propose; a named person decides.
Ask about a member, a rating or a rule. The agents answer from the systems of record, with sources, and never change a rating themselves.
Who are the two wire counterparties?
Varnholt Trading Ltd. (list entry HRJ-07): 4 wires, $108,700. Brisa Coastal Imports S.A. (HRJ-12): 3 wires, $77,700. Neither is a Norrell member, and the OFAC screening service cleared both names on every wire. Both jurisdictions are on list v2026-09.
Was the September cash alert closed correctly?
Imani Rourke closed TM-26-0930-4471 on September 30 as not suspicious at this time, citing seasonal contract sales. September cash was $76,900 in 47 deposits, and that trend is what moved the rating. The drafted EDD case asks for the contracts that would explain it.
What is the open alert?
TM-26-1009-2281, opened October 9 on joint share draft ••4470, which Leona Strick owns with Dorian Strick: $48,000 in by ACH and $46,500 out by P2P to 9 recipients within 36 hours. Jonah Pell has it in review.
When can it move down?
Only after TM-26-1009-2281 is closed. The agents re-check the night it closes and propose the change again, with the closure cited.
13 screens from the working solution, on its sample data. Pick one to see it large.
The thresholds, clocks and reason codes every rating change runs on are settings, changed by the administrator and kept on the audit trail.
| Setting | Default | Choose from |
|---|---|---|
| Confidence for a person | 0.85 | Below it, the change is flagged for close reading |
| Confidence for a batch | 0.90 | Only Low ↔ Moderate, no holds |
| Tier cut-offs | Low 0–34 · Moderate 35–59 · High 60+ | Set by the approved methodology version |
| Clocks | 2 business days into High; 5 for other moves; 3 for batches | Second signature within 1 business day; EDD case 30 calendar days |
| Escalation | At 80% of a clock and at breach | To the Senior BSA Analyst, then the BSA Officer |
| Periodic re-score | Quarterly, next January 4, 2027 | The whole book, on top of nightly re-score on change |
| Sampling | 10–12% of each batch, at least 2 | Risk-weighted, closest to the tier boundary first |
| Retention | 5 years from each decision | Account numbers redacted to the last four (31 CFR 1010.430) |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample data · “target” = the design goal, measured in the live solution · “estimated” = our estimate (the split of an analyst's minutes on one rating change). Rules named for context: the FFIEC BSA/AML Examination Manual, Introduction – Customers (December 1, 2021); the interagency joint statement on a risk-based approach to customer relationships (July 6, 2022; NCUA Letter 22-CU-08); the interagency statement on model risk management for BSA/AML systems (April 9, 2021); the FinCEN CDD rule, 31 CFR 1010.230; BSA record retention, 31 CFR 1010.430; SAR confidentiality, 31 CFR 1020.320(e) and 12 CFR 748.1. The credit union, members, businesses, people and identifiers named on this page are fictional: characters and sample data in the working solution.
A customer risk rating is the customer risk profile a bank or credit union keeps for every customer, used to set how closely monitoring watches the relationship and when enhanced due diligence starts. Customer Risk Rating is a solution in which agents re-score a member the night a fact changes, explain each rating change in one sentence whose every clause cites its source and weight, hold type-only and open-activity changes, and route every upgrade and downgrade to a maker and a different checker.
The Rating agent scores five families of facts (member profile, products and services, geography, channel and activity) under the institution's approved methodology version, with tier cut-offs set there: in the working solution, Low 0–34, Moderate 35–59 and High 60 and over. Each rule's points are visible from the clause that used them, and the agent also writes the reverse sentence: what would have to change for the rating to move back.
No. The FFIEC BSA/AML Examination Manual and the 2022 interagency joint statement say no customer type automatically presents a higher risk. A Type-Bias Checker holds any proposed High that rests on member type alone and tests the whole book every night; in the working solution, 7,140 High ratings were tested and none rested on type alone.
On change, every night: the Fact Watcher links each changed fact in the core, monitoring, account opening, screening and case systems to a member, and the member is re-scored. The whole book is also re-scored on a schedule the administrator sets, quarterly in the working solution.
A senior analyst drafts it, the Methodology Steward back-tests it on every member and shows the tier migration, Model Risk validates it with a memo, and the BSA Officer approves it. Nothing scores under a version that has not been approved, and each step names the person who took it.
No. The Hold Checker cross-checks every change against open alerts, cases, 314(a) matches, pending EDD and facts newer than the score, and holds a downgrade during open activity with the evidence. Only a person decides a held change, and only the BSA Officer can override a hold.
Yes. The agents propose, explain, cite, check and draft; they never apply a rating, release a hold, contact a member or show a SAR outside the BSA Unit. A BSA analyst approves, overrides or rejects with a reason code, and a different person signs second before the rating is written to the core.
The Agentic Solution Engine builds and deploys it from your requirements and documents: your risk rating methodology, your reason codes, your high-risk lists and a sample of your members' facts. It goes live once every quality gate has passed. We will walk you through it on your own members first.
The annual risk assessment built in days, every rating cited, refreshed each quarter.
1-min demoFinCEN 314(a) RequestsEvery subject searched and proven, every true match replied to inside the 14 days.
1-min demoBaaS ComplianceEvery fintech partner tested every quarter and tied out every day, with the evidence.
1-min demoWe'll run Customer Risk Rating on a sample of your own members' facts and your current methodology.