Yes, up to the point of confirmation. An AI agent can run politically exposed person screening at onboarding and on refresh, compare each match against identifiers, and recommend discounting the clear mismatches. Confirming that a customer is a PEP, approving the relationship and deciding the enhanced measures stay with a compliance officer, because the firm answers for each.
What makes PEP screening different from sanctions screening?
A sanctions match is a prohibition: if confirmed, you cannot deal. A PEP match is a risk signal: the customer may be entirely legitimate, but their position exposes them to bribery and corruption, so the firm applies more scrutiny. That difference changes what the agent is for. In sanctions screening the question is "is this the listed person?" In PEP screening there are three questions: is this the person on the list, are they still in a prominent role, and are they a family member or close associate of one?
The second and third questions are where judgement lives. Lists are imperfect on when someone left office and on who counts as a close associate. For the definitions, see what PEP screening is.
What may the agent prepare?
| Task | The agent | Stays with the officer |
|---|---|---|
| Screening | Runs the customer and connected parties against the PEP data at onboarding, refresh and on list updates | Choice of data and match settings |
| Match resolution | Compares date of birth, nationality, address, photo and role; recommends discount where identifiers clearly differ | Any match it cannot clearly discount |
| Classification | Proposes PEP, family member or close associate, with the source for the link | Confirming the classification |
| Role status | Records when the role was held and whether it has ended | Deciding when someone stops being treated as a PEP |
| Source of wealth | Requests evidence and summarises it against the declared story | Whether wealth is explained |
| Approval pack | Drafts the case for senior approval, with risks and adverse media | The approval itself |
Most firms' policies require a senior person to approve taking on or keeping a PEP relationship. The agent's pack makes that approval faster. It does not stand in for it.
What does a match look like when the agent works it?
Take a private bank onboarding a new client, a property investor. Screening returns three matches.
The first is a former minister in another country with the same name. The agent compares dates of birth and finds they differ by twenty years; it recommends discounting and records why. The second is a sitting regional governor with a similar name, no date of birth on file and a different nationality. It recommends discounting but marks the confidence as lower and asks for a person to confirm. The third is the most useful: the client's brother-in-law is a serving deputy head of a state-owned energy company. The agent found the link through a marriage record named in an adverse media article and the company's own published board list.
The officer confirms the first two discounts. On the third, she classifies the client as a family member of a PEP, requests source of wealth evidence and sends the pack for senior approval. The agent then schedules the relationship for more frequent review. The step-by-step method is in how to review PEP screening matches with an AI agent.
Which controls keep the screening sound?
- Discount rules are written and tested. The agent discounts only where your policy says identifiers are sufficient. Everything else is held for a person.
- Low-confidence discounts are flagged. A discount with thin evidence is not treated the same as one with a clear date-of-birth mismatch.
- Sources carry provenance. A link from the list provider is different from a link the agent inferred from a news article, and the record shows which is which. Adverse media screening is a separate input with its own reliability limits.
- Samples are re-checked. A person reviews a sample of discounted matches each cycle.
- No downgrade by the agent. Ending PEP treatment after someone leaves office is a risk decision for the officer.
The broader approach is set out in our guide to AI compliance agents.
Where Pryme fits
In Pryme, screening, match resolution and drafting the approval pack are separate capabilities, each certified on your own past PEP cases before the agent uses it. Discounting a clear mismatch can earn routine depth; confirming a PEP or approving a relationship is outside the agent's role scope. Every recommendation and decision is filed, so an examiner can see who decided what.
Questions people ask
Can the agent tell the customer they have been identified as a PEP?
It can send the standard request for source of wealth evidence that your policy uses for higher-risk customers. Explaining the classification itself is better done by a relationship manager or the compliance team.
How often should existing customers be rescreened?
Whenever the PEP data updates and at each periodic review. The agent can run the delta daily, which catches people who take office after onboarding.
Can the agent treat domestic and foreign PEPs differently?
It can apply whatever distinction your policy makes. The policy itself, including how much extra scrutiny each category gets, is set by people.