Skip links

AML Risk Assessment: Building a Risk-Based Approach That Actually Scales

AML & Compliance
Last updated: November 6, 2026
· 10 min read

The Financial Action Task Force (FATF) wrote the risk-based approach into Recommendation 1 of its standards on 16 February 2012, when it folded the old forty Recommendations and nine Special Recommendations on terrorist financing into a single set and replaced a flatter, same-depth-for-everyone due diligence standard with an explicit requirement to scale controls to assessed risk. Twelve years later, the mandate was still just a mandate at one of North America’s largest banks. In October 2024, TD Bank pleaded guilty to conspiracy charges tied to its anti-money-laundering program and paid the Financial Crimes Enforcement Network (FinCEN) a $1.3 billion civil penalty — the largest in FinCEN’s history — on top of a separate criminal resolution with the U.S. Department of Justice.

The rule existed. The operating mechanism to run it at the size the bank had grown to did not. That gap, between the policy sentence and the workflow underneath it, is where most of an AML risk assessment actually fails, and it’s the subject of this piece: not what a risk-based approach to AML says on paper, but what has to be true operationally for it to hold at volume.

Enhanced due diligence turns on one question underneath all the others: who actually owns the entity, past the first layer of listed directors. Zavia connects directly to government ownership registries in 173 countries to resolve that in one lookup instead of chasing filings jurisdiction by jurisdiction — which matters here specifically because a risk tier built on unverified, self-reported ownership is scoring the disclosure form, not the company.

01Why a fixed checklist doesn’t scale

A checklist scales linearly: more customers, more boxes, more headcount. A risk-based approach scales logarithmically, because it concentrates effort where the actual exposure sits. That is the entire function of Recommendation 1: FATF requires countries, and the entities they regulate, to identify, assess and understand their money-laundering and terrorist-financing risks, then apply resources and controls in proportion to what they find — not apply identical depth to every relationship. The Interpretive Note to Recommendation 1 states this cuts both ways: higher risk requires enhanced measures, and lower risk can justify simplified ones. A program that only knows how to escalate isn’t risk-based. It’s a slower checklist with extra paperwork.

The cost shows up on both sides of the tier, not only the expensive one. Every hour an analyst spends re-verifying a low-risk customer’s source of funds for the third time is an hour not spent on the file where a politically exposed person (PEP) just surfaced as a beneficial owner. Flat controls don’t just waste capacity on the low-risk side — they starve the high-risk side of the attention FATF, FinCEN, and the UK’s Financial Conduct Authority (FCA) now test for directly in supervisory reviews.

02Two risk assessments, not one

Most teams conflate two distinct exercises. FATF’s Recommendation 1 and its Interpretive Note require an institutional, firm-wide risk assessment — a top-down view of the money-laundering and terrorist-financing exposure a bank, payments firm, or fintech carries across its customer base, geography, products, and delivery channels — kept separate from the customer risk assessment applied file by file. The FCA’s Financial Crime Guide, in its April 2025 edition, draws the same line: the business-wide risk assessment should set risk appetite and the design of controls, while the customer risk assessment should directly shape due diligence depth and transaction-monitoring thresholds. Skip the firm-wide layer and the customer scoring model has no anchor. It’s measuring risk against a baseline nobody set.

The FCA tested this distinction directly. Its 2025 multi-firm review of risk assessment processes and controls found firms whose customer-level risk assessments had drifted away from what the firm’s own institutional assessment said the real exposures were — static customer scoring, sitting on top of a risk landscape the firm had already flagged, on paper, as having moved.

03The four risk factor categories

Whatever the institution, the inputs to a customer risk assessment come from the same four categories that FATF’s Recommendation 10 and its Interpretive Note set out: who the customer is, where the customer and their money sit geographically, what product or service is involved, and how the customer is being onboarded and served.

Customer riskLegal structure, ownershipopacity, occupation, PEPstatus, source of fundsGeographic riskHigh-risk / sanctionedjurisdictions, weak AMLregimes, FATF grey listProduct & service riskCash-intensive, correspondentbanking, trade finance,private banking, VASP railsDelivery channel riskNon-face-to-face onboarding,intermediated relationships,introduced business
Source: FATF Recommendation 10 and its Interpretive Note on customer due diligence risk factors.

None of the four is dispositive alone. A high-risk jurisdiction doesn’t make every customer located there high-risk, and a low-risk product doesn’t clear a customer whose ownership structure is opaque — the factors combine, and the combination is the actual score. Geography is the factor that moves fastest. The EU’s Anti-Money Laundering Regulation (AMLR) gives it a binding edge: under Article 29 of Regulation (EU) 2024/1624, the European Commission designates “high-risk third countries” with significant strategic deficiencies in their AML/CFT regime, and Article 34(4) requires obliged entities to apply enhanced due diligence to every business relationship and occasional transaction tied to a person or entity from one of those countries. The list isn’t static. The Commission added Russia to it through Commission Delegated Regulation (EU) 2026/46, adopted 3 December 2025 and published in the Official Journal on 9 January 2026 — which means a relationship that was standard-risk on geography alone can become high-risk without anything about the customer itself changing. A checklist reviewed once a year cannot catch that. A risk-based program has to be built to.

04Risk tiers in practice: SDD, CDD, and EDD

The output of a customer risk assessment is a tier, and the tier is supposed to set the depth of due diligence applied — not just the label on a dashboard. FATF’s Recommendation 10 permits simplified due diligence (SDD) where risk is demonstrably lower, sets standard customer due diligence (CDD) as the default, and requires enhanced due diligence (EDD) where risk is higher, including for every customer identified as a PEP. FinCEN’s 2016 customer due diligence rule, now codified at 31 C.F.R. § 1010.230 as the “fifth pillar” of a Bank Secrecy Act program and binding since its 11 May 2018 compliance date, adds the operational backbone for U.S. institutions: identify the beneficial owner of every legal entity customer, build a customer risk profile, and keep that profile current through ongoing due diligence.

FATF’s own guidance adds one qualification worth sitting with: it warns institutions against applying simplified due diligence automatically. Regulators wrote a rule telling institutions not to turn a risk-based exemption into a new flat rule.

SDD vs. standard CDD vs. EDD — typical trigger, depth and review cadence
Tier Typical trigger Due diligence depth Typical review cadence
SDD Lower-risk customer type, product and geography combined; no adverse indicators Reduced identity verification; no enhanced source-of-funds check; relies on other controls to catch change Event-driven, or long-interval periodic review
Standard CDD Default tier for most customer relationships Full identification and verification, beneficial ownership down to the applicable threshold, understanding the purpose and nature of the relationship Periodic review, commonly every 1–3 years, plus trigger events
EDD PEP status, high-risk third-country nexus, adverse media, opaque ownership, cash-intensive business, correspondent or private banking Senior management sign-off, enhanced source-of-funds and source-of-wealth verification, lower monitoring thresholds At least annual; often continuous or fully event-driven

Tier definitions synthesized from FATF Recommendation 10 and its Interpretive Note, FinCEN 31 C.F.R. § 1010.230 (effective 11 May 2018), and EU AMLR Article 34(4).

Read the table as a depth dial, not three separate programs. The mechanics of identification, screening, and ownership verification don’t change tier to tier. What changes is how much evidence is required before the tier is accepted, and how often the file gets checked again.

A customer scored medium-risk at onboarding can become high-risk the day a beneficial owner changes, and a once-at-intake Know Your Business (KYB) check has no way to notice. Zavia’s ownership-monitoring API flags ownership and structure changes after onboarding closes — closing exactly the gap a point-in-time check leaves open for the rest of the relationship.

05The escalation flow, start to finish

Put the tiers in motion and the program looks less like a static table and more like a loop with exits at every stage. A customer is scored at onboarding using the four factor categories from Recommendation 10; a low score routes to simplified measures, a medium score to standard CDD, and a high score — a PEP hit, a high-risk jurisdiction, adverse media — routes into an EDD queue. FATF guidance expects a second set of eyes on that queue, and senior management sign-off before the relationship is accepted or continued.

Onboarding &risk scoringLOW → SDDreduced depthMEDIUM → CDDstandard depthHIGH → EDDenhanced depthOngoing / perpetualmonitoringtrigger event re-scores the customer
Source: FATF Recommendation 10 (SDD/EDD triggers); FinCEN 31 C.F.R. § 1010.230 (fifth-pillar ongoing CDD).

The part most programs build as an afterthought is the dashed line back into onboarding. A trigger event discovered during monitoring — a new adverse-media hit, an ownership change, a sanctions-list update, a transaction pattern that doesn’t match the stated purpose of the account — has to be able to pull a customer back out of whatever tier they’re sitting in and force a new score. Without that path built as an instrumented workflow, “risk-based” is only true on day one.

06Perpetual monitoring: how re-scoring actually happens

A risk score set once, at onboarding, measures nothing five years later. FinCEN’s fifth pillar makes that explicit: ongoing due diligence isn’t optional documentation, it’s the mechanism that keeps a customer risk profile accurate as facts change, and it requires updating customer information on a risk basis, not a fixed calendar basis. The FCA’s 2025 review of risk assessment practices reaches the same conclusion from the supervisory side. Firms it rated well treat both the firm-wide and customer risk assessments as dynamic and keep them current. Firms it rated poorly run static, point-in-time exercises that quietly go stale.

Institutional (firm-wide)risk assessmentCustomer riskassessment & tieringDifferentiated duediligence (SDD/CDD/EDD)Ongoing monitoring& trigger eventsfeedback loop,not a one-time form
Source: FATF Recommendation 1 Interpretive Note; FCA Financial Crime Guide, April 2025 edition, on BWRA/CRA linkage.

Notice what sits at the center of that loop: not a form, an update. The institutional assessment sets the appetite and the controls. The customer assessment applies it to a specific relationship. Differentiated due diligence acts on the result. Monitoring feeds evidence back in that can move the customer — or, if enough customers move the same way, move the firm-wide assessment itself. A program that can’t close that loop isn’t risk-based, however many factors it lists on an intake form.

07Operationalizing it: what has to change

Making this run at the volume of a real customer base, rather than on a spreadsheet of forty accounts a compliance manager can hold in their head, takes four specific operational changes.

  1. Document the scoring logic. A regulator, or a new analyst, should be able to trace why a given customer landed in a given tier back to the four Recommendation 10 factor categories, not to an individual’s intuition.
  2. Verify the ownership data underneath the score. EDD depends on confirming who actually controls a legal entity customer, which is exactly where independently sourced company data, like a Know Your Business (KYB) check run against government registries rather than a self-reported form, earns its place in the stack.
  3. Wire monitoring into re-scoring, not just into an alert queue. A sanctions update, an adverse-media hit, or an ownership change has to change the tier, not just generate an alert someone clears without touching it.
  4. Check the firm-wide and customer-level assessments against each other on a schedule. This is the single finding the FCA’s 2025 review kept surfacing in the firms it rated poorly.

None of that requires more headcount per customer, on average. It requires less, for the low-risk majority, so the high-risk minority gets more. That reallocation is the mechanism Recommendation 1 was built around in February 2012, and it is also the only version of “risk-based” that survives contact with a growing customer base.

08What regulators are finding when they look

The FCA’s 2025 multi-firm review of risk assessment processes and controls found the same recurring pattern across the firms it examined: risk assessments that were comprehensive on paper but not demonstrably driving CDD depth or monitoring thresholds in practice. The review’s good-practice markers read, in reverse, as the failure modes worth checking for first: a risk assessment that is templated rather than genuinely comprehensive, signed off once a year rather than tested on a real cadence, built from stale sources rather than the best available internal and external information, and static rather than dynamic enough to notice when the facts underneath it have changed.

The EU’s shift to a single, directly applicable AML Regulation — replacing the patchwork of national transpositions of the AML Directives — is itself a bet that inconsistent risk-based implementation across member states was part of the problem. Article 29’s high-risk third-country mechanism, with Russia’s addition still fresh from January 2026, is a live demonstration of how fast the geographic input to a risk assessment can move, and how badly a once-a-year review cycle fits that reality.

None of the tiering in this piece holds up if the ownership and registration data underneath it has to be manually re-pulled every review cycle. Zavia’s plans cover both the single UBO lookup and the ongoing monitoring API across the same 173 countries — because a program that only re-checks risk on a calendar, not on an event, isn’t risk-based. It’s scheduled.

FAQ

What is an AML risk assessment?

An AML risk assessment is the process of identifying and scoring a bank’s or firm’s exposure to money laundering and terrorist financing, at two levels: a firm-wide (institutional) assessment of overall exposure, and a customer-level assessment of each individual relationship, both required under FATF Recommendation 1 and its Interpretive Note.

What’s the difference between a risk-based approach and a checklist-based approach to AML?

A checklist applies the same due diligence depth to every customer. A risk-based approach to AML, as FATF Recommendation 1 requires, applies due diligence depth in proportion to assessed risk — simplified for lower risk, standard for typical risk, and enhanced for higher risk — and keeps re-assessing as facts change.

What’s the difference between institutional and customer risk assessments?

The institutional (firm-wide) risk assessment looks at the whole business’s exposure across its customer base, geography, products and channels, and sets risk appetite and controls. The customer risk assessment scores an individual relationship and directly drives that customer’s due diligence depth and monitoring thresholds, per the FCA’s Financial Crime Guide.

What are the standard AML risk factor categories?

FATF Recommendation 10 and its Interpretive Note point to four categories: customer risk (who the customer is, including PEP status and ownership structure), geographic risk (jurisdictions, including EU AMLR Article 29 high-risk third countries), product/service risk, and delivery channel risk (face-to-face versus remote or intermediated onboarding).

What triggers enhanced due diligence (EDD)?

Common triggers include politically exposed person (PEP) status, a nexus to a high-risk or sanctioned jurisdiction (including the EU’s Article 29 high-risk third-country list), adverse media, opaque or complex beneficial ownership, cash-intensive business models, and correspondent or private banking relationships, per FATF Recommendation 10 and EU AMLR Article 34(4).

How often should a customer risk assessment be reviewed?

There’s no single universal interval. FinCEN’s fifth-pillar rule (31 C.F.R. § 1010.230) and FATF guidance both frame review as risk-based and event-driven rather than purely calendar-based: lower-risk customers get reviewed less frequently, higher-risk customers at least annually or continuously.

What does FATF Recommendation 1 require?

It requires countries and the obliged entities they regulate to identify, assess and understand their money-laundering and terrorist-financing risks, and to apply resources and AML/CFT measures proportionate to those risks — the formal basis for every risk-based approach to AML used worldwide since 16 February 2012.

Does a risk-based approach mean lower-risk customers get no due diligence?

No. Even simplified due diligence under FATF Recommendation 10 still requires baseline identification and verification. Simplified reduces depth and frequency for demonstrably lower-risk relationships; it doesn’t remove due diligence entirely.

This website uses cookies to improve your web experience.