KYC analysts who stay employable learn to judge files an AI agent has already assembled. In retail banking the agent extracts identity data, runs screening and drafts a risk rating; the analyst resolves hard matches, weighs each document, tests the source of funds and writes a rationale an examiner could follow.
What does an agent do in onboarding?
Most onboarding files are clean. An agent can read an identity document, compare it with the applicant's details, run sanctions and politically exposed person screening, check an address and propose a low-risk rating. For periodic reviews it can pull the account's activity and draft "no material change" where that is true.
The analyst's time moves to the files that are not clean: the near-match on a sanctions list, the applicant whose income does not explain their deposits, the company with an ownership chain three layers deep, the review where activity has quietly changed. Under the risk-based approach set out in the FATF Recommendations, these are the cases that deserve the most attention.
Where does the line fall?
- Agent, depth 1: reads documents and screening results.
- Agent, depth 2: drafts a risk rating, a request for more documents or a review summary for you to confirm.
- Agent, depth 3: completes low-risk onboarding inside rules the bank has proven on its own past files.
- Held for you: screening matches, higher-risk ratings, enhanced due diligence, unexplained funds, and any decision to decline or exit a customer.
- Refused: anything outside the agent's role, such as clearing a sanctions match on its own.
Every verdict is filed with the bound that decided it, so you can see whether a file reached you because of a limit, a weak source or a red line.
What does a held periodic review look like?
Take a KYC analyst we will call Daniel at a high-street bank. The agent has drafted "no material change" for a customer's periodic review, but the bank's limits held it because inbound payments have risen sharply. Daniel reads the record. The customer's declared occupation is a part-time retail job. Over six months, regular transfers of around 3,000 have arrived from several unrelated individuals and left within a day.
The agent's draft relied on the customer's declared income and saw no screening hits. Daniel sees a pattern consistent with a money mule. He declines the "no change" outcome, writes "Inbound from multiple unrelated payers, rapid outbound; inconsistent with declared income; refer to financial crime," and raises an internal report. That judgement is the job.
Which skills should you build?
- Match resolution: deciding whether a screening hit is the same person, using date of birth, nationality and other identifiers, and writing down why.
- Document strength: a record issued and signed by an identifiable issuer counts for more than a scan the applicant uploaded. Learn the provenance classes and use them.
- Source of funds judgement: does the story explain the money?
- Pattern reading in periodic reviews, where the agent's draft may miss a slow change.
- Risk rationales that are short, specific and tied to policy.
- Feeding back repeated weaknesses so the agent's sources and holds are adjusted.
Onboarding sits beside monitoring and payments; compare AI skills for a compliance analyst and AI skills for a fraud analyst.
How do you show it?
Describe the decisions: matches resolved, enhanced due diligence files completed, reviews escalated. The AI agent skills CV section for a KYC analyst and rewriting a KYC analyst CV for AI operations jobs give wording. If agents are new to you, how to learn AI agent skills from scratch sets out an order, and AI skills for job seekers explains what employers ask you to prove.
Where the Academy fits
Pryme Intelligence Academy's Operator track trains review of held decisions, keeping the agent's knowledge current and handling incidents, with module tests set in your own industry and role. The practicum asks for nineteen in twenty correct decisions and no missed red line; for a KYC analyst, clearing a sanctions match without evidence is the kind of red line it tests.
Questions people ask
Will agents replace KYC analysts?
Agents take on much of the clean, low-risk work. Banks still need analysts for matches, higher-risk customers and decisions to decline or exit, and they need those decisions documented well.
Is document authenticity checking still a skill?
Yes. Agents can check fields and formats, but recognising a document that does not fit the customer's story, or one that has been altered, still benefits from experience.
Do I need financial crime qualifications as well?
They remain useful and many employers value them. An AI review certificate shows a different thing: that you can review agent decisions to a standard. The two complement each other.