How to review PEP screening matches with an AI agent

A PEP match asks two questions: is this the same person, and if so, what risk does their position create? The agent answers the first with a field-by-field comparison and prepares the second with the role, country, time since leaving office and adverse media. An analyst confirms the match, and senior management approves any relationship with a confirmed PEP.

What is a PEP match review, and who owns it?

Politically exposed persons hold, or have held, prominent public functions, and so carry a higher risk of involvement in bribery and corruption. Their family members and close associates are usually in scope too. FATF Recommendations require firms to identify PEPs, apply enhanced due diligence, establish source of wealth and source of funds, and obtain senior management approval for the relationship. What PEP screening is covers the basics.

Screening analysts review matches; a senior manager approves PEP relationships under the firm's policy; the MLRO owns the policy. Unlike a sanctions match, a PEP match is not a prohibition. It is a reason to look closer. That makes the agent's comparison useful, and its framing important. Whether an AI agent can do PEP screening sets out the boundary.

What are the steps, and who acts at each?

1. Agent, depth 1: reads the match. The customer record, the PEP database entry, the match score and the matched fields.
2. Agent, depth 1: compares identity. Name and aliases, date of birth, nationality, place of birth, address, photographs where the database holds them and your policy allows, each marked as matching, differing or missing.
3. Agent, depth 1: reads the role. Position held, country, level of prominence, dates in office, and whether the person is a family member or close associate of a PEP.
4. Agent, depth 1: checks related information. Adverse media on the person, using the approach in whether an agent can do adverse media screening, and the corruption risk of the country concerned from your country risk ratings.
5. Agent, depth 2: drafts the note. Whether the identity evidence supports a match, and, if it does, the role, its risk factors and what enhanced due diligence your policy requires.
6. Person, analyst: decides the match. Same person, different person, or more information needed.
7. Person, analyst: sets the risk treatment for confirmed PEPs under the policy, including source of wealth and funds checks.
8. Person, senior manager: approves establishing or continuing the relationship.
9. Agent, depth 1: schedules ongoing monitoring and the next review date set by the analyst.

Which controls apply, and what evidence does each leave?

Identity and risk kept separate. The note answers "same person?" before "how risky?". Mixing them leads to analysts dismissing true matches because the role seems minor. Evidence: the two-part note structure.

Database content marked external. PEP databases compile public information; they are useful but can be wrong or out of date. Evidence: each fact in the note carries its source.

Senior approval recorded. Evidence: the approver, the date and the conditions.

Declassification by policy. Whether a former PEP still needs enhanced measures depends on your policy and the risk, not on a fixed period alone. Evidence: the analyst's decision with reasons.

Consistency across similar matches. The agent can show how similar roles in the same country were treated. Evidence: the comparison, and the reason where this decision differs.

What should you measure?

• Analyst time per match.
• Confirmed PEPs where the note leaned towards "different person", each reviewed.
• Matches pending more information, and for how long.
• PEP relationships without current senior approval, which should be none.
• Reviews overdue for confirmed PEPs.

Where does it usually go wrong?

Domestic PEPs treated as low risk by default. Many policies treat foreign PEPs as higher risk, but domestic PEPs can carry equal risk. Apply your risk factors to each case.

Family and associates missed. Screening may only flag the PEP. A customer who is the spouse or business partner of a PEP may not match directly. Where the database lists relatives and associates, the agent should check the customer against them.

Name-only comparisons. Common names produce many matches. Without date of birth or other identifiers, the agent should say the comparison is inconclusive rather than lean either way.

Sanctions and PEP confused. Both come from screening, but the treatment differs entirely. The sanctions process is covered in letting an agent read sanctions screening matches.

Adverse media handled as proof. A press report of an allegation is information, not a finding; see what adverse media screening is.

The overall stance on compliance agents is in AI compliance agents.

Where Pryme fits

In Pryme, PEP database entries reach the agent as external facts and verified identity documents as signed, so the comparison shows how strong each side of the match is. Confirming a match and approving the relationship sit outside the agent's role scope; it drafts at depth 2, and its comparison capability is certified on the firm's own past matches.