# KYC vs AML vs KYB vs CDD: what each means, and where AI fits in each

KYC verifies who a customer is, AML is the wider programme against money laundering, KYB verifies businesses, and CDD/EDD set how deep checks go. A plain-English guide for founders and engineers, with where automation and AI help.

Published: 2026-11-08

**KYC (Know Your Customer)** checks that a customer is who they say they are. **AML (Anti-Money Laundering)** is the wider legal programme a regulated business runs to stop its services being used for laundering money or financing terrorism; KYC is one part of it. **KYB (Know Your Business)** is KYC for companies: verifying the business and the people who own and control it. **CDD (Customer Due Diligence)** is the level of checking applied to a customer, and **EDD (Enhanced Due Diligence)** is the deeper version for higher-risk customers. Founders building fintech, marketplaces or crypto products meet all five terms in the first compliance conversation, so here's what each means and where software and AI help.

## KYC: who is this person?

The onboarding check most people have done: photograph an ID, take a selfie. Behind it:

* **Document verification:** is the ID genuine and readable? See [how KYC verification works](https://aliahmed.si/blog/how-kyc-identity-verification-works/), [MRZ check digits](https://aliahmed.si/blog/passport-mrz-check-digits/) and [NFC chip reading](https://aliahmed.si/blog/nfc-epassport-chip-verification/).
* **Biometric verification:** does the selfie match the ID ([face match thresholds](https://aliahmed.si/blog/face-match-threshold-far-frr/)), and is it a live person ([liveness detection](https://aliahmed.si/blog/face-liveness-detection-explained/))?
* **Data checks:** address, date of birth, sometimes a database lookup.

This is where computer vision and OCR do most of the work, and where I spent my time at ShuftiPro.

## AML: the whole programme

AML covers everything a regulated business must do, including:

* **Screening** customers against sanctions lists, politically exposed person (PEP) lists and adverse media.
* **Transaction monitoring** for patterns such as structuring (splitting payments to stay under reporting limits) or money moving rapidly through an account.
* **Reporting** suspicious activity to the financial intelligence unit.
* **Record-keeping, training and audits.**

KYC tells you who the customer is; AML asks whether what they do makes sense for who they are.

## KYB: verifying a business

For business customers you verify the company (registration, status, address) and its **ultimate beneficial owners (UBOs)**: the people who ultimately own or control it, typically above a 25% ownership threshold, although rules vary by country. Ownership chains through holding companies in several countries are what make KYB slow. Each UBO then goes through KYC.

## CDD and EDD: how deep to go

* **Simplified due diligence** for demonstrably low-risk cases, where the rules allow it.
* **CDD:** the standard level: identify and verify the customer, understand the purpose of the relationship, and monitor it.
* **EDD:** for higher risk, such as PEPs, high-risk countries or unusual structures: source of funds and wealth, more documents and senior approval.

The level is set by a **risk assessment**, so a good onboarding system adapts the checks to the customer instead of applying the maximum to everyone.

## Where does AI help?

| Area                   | Useful AI                                    | Keep deterministic                       |
| ---------------------- | -------------------------------------------- | ---------------------------------------- |
| Document checks        | Computer vision, multilingual OCR            | MRZ check digits, chip signatures        |
| Biometrics             | Face matching, liveness models               | Thresholds and the review band           |
| Name screening         | Fuzzy and transliteration-aware matching     | The list data and the match decision log |
| Adverse media          | LLMs to summarise and classify articles      | Sources cited for every finding          |
| Transaction monitoring | Anomaly detection to find new patterns       | Rule-based scenarios regulators expect   |
| KYB                    | Extracting ownership from registry documents | The UBO calculation itself               |

Two rules carry across all of it. **Every decision needs a reason a person can read**, because a regulator or a rejected customer will ask. And **LLM output must be verified before it counts**: a summary that invents a fact about a customer is a compliance failure. That's the subject of [how I stop an LLM from inventing numbers](https://aliahmed.si/blog/stop-llm-inventing-numbers/) and [getting reliable structured output from LLMs](https://aliahmed.si/blog/llm-structured-output-validation/), and audit trails are covered in [tamper-evident logs for AI agents](https://aliahmed.si/blog/ai-agent-audit-log-hash-chain/).

## Summary

* **KYC:** verify the person.
* **KYB:** verify the business and its owners.
* **CDD / EDD:** how deeply to check, by risk.
* **AML:** the full programme: screening, monitoring, reporting, and KYC inside it.

I build AI for identity verification and financial intelligence; more on the [home page](https://aliahmed.si/) and the [projects page](https://aliahmed.si/projects/).

---

Written by Ali Ahmed · [Original article](https://aliahmed.si/blog/kyc-vs-aml-vs-kyb/)
