How to review source of wealth evidence with an AI agent

Source of wealth asks whether the customer's total wealth makes sense given their life. The agent builds a timeline of how the wealth was accumulated, matches each event to evidence, tests whether the pieces add up to the declared total, and drafts the gaps. An analyst decides whether the account is plausible, and senior management approves high-risk relationships.

What is a source of wealth review, and who owns it?

Source of wealth is how a customer came to have their overall net worth: the career, businesses, inheritances, investments and property that built it. It differs from source of funds, which concerns specific money in a transaction; the two are often reviewed together, and reviewing source of funds evidence covers the narrower question. Source of wealth matters most for high-net-worth customers, politically exposed persons and other high-risk relationships, where FATF Recommendations expect it to be established as part of enhanced due diligence.

EDD analysts or private banking compliance teams review it, with senior management approval for the relationships that require it. The hard part is that wealth builds over decades and evidence is partial. The analyst is judging plausibility, not proving every pound or dollar. Whether an AI agent can verify source of funds and source of wealth sets out the boundary.

What does the agent do at each step, and at what depth?

1. Agent, depth 1: records the declaration. The customer's or relationship manager's account of how the wealth was built, and the declared total, both as declared.
2. Agent, depth 1: builds the timeline. Education, career, roles and salaries, businesses founded or sold, inheritances, property, investments, each with approximate dates and values.
3. Agent, depth 1: matches events to evidence. Employment records, company filings, sale agreements, probate records, tax returns where provided, property records and reliable public sources, each marked by provenance.
4. Agent, depth 1: tests plausibility. Whether the evidenced events, with reasonable growth, could produce the declared wealth. It shows its working and assumptions.
5. Agent, depth 1: checks public information for consistency, such as company filings showing a business sale or reports of a role, marking it as external.
6. Agent, depth 2: drafts the source of wealth narrative and gap list. What is evidenced, what is only declared, what does not fit, and what further evidence would help.
7. Person, analyst: decides whether the source of wealth is established to the standard your policy requires.
8. Person, senior manager: approves the relationship where policy requires.

Which controls apply, and what evidence does each leave?

Declared and evidenced kept apart. The narrative shows clearly which parts rest on documents and which only on the customer's word. Evidence: provenance on every event in the timeline.

Plausibility working shown. Any estimate, such as the likely value of a business sold years ago, shows its assumptions. Evidence: a calculation the analyst can challenge.

Relationship manager independence. Relationship managers have an interest in the customer being accepted. Their account is declared, like the customer's. Evidence: the source of each statement.

Senior approval recorded. Evidence: approver, date, any conditions such as further evidence by a deadline.

Narrative refreshed. Wealth changes. Evidence: review dates and triggers, such as a large new inflow or a change of role.

What should you measure?

• Proportion of declared wealth evidenced, by case, as a guide to file quality.
• Gaps closed after requests.
• Cases where the plausibility test failed, and the outcomes.
• Time to complete reviews for new high-net-worth relationships.
• QA findings on narratives.

Where does it usually go wrong?

The narrative written by the relationship manager. A polished story with little evidence. The agent's timeline and gap list put the evidence first.

False precision. Trying to account for every unit of wealth makes the review endless and does not reduce risk. Aim for a plausible, evidenced account of the main sources.

Public information treated as proof. A press profile saying someone sold a company is useful but can be wrong. Company filings are stronger.

Ignoring how wealth was made. Wealth from a sector or country with high corruption risk deserves more scrutiny, even if well evidenced. The analyst should weigh it.

Joint and family wealth blurred. Wealth held jointly with a spouse, or through a family trust or company, needs to be attributed clearly. A narrative that counts family assets as the customer's own overstates what is explained, and one that ignores them leaves inflows unexplained later.

Generational wealth. Inheritances from decades ago may have little documentation. The analyst decides what is reasonable; the agent should not mark these as failures automatically.

Market expectations vary; our brief on wealth management compliance in Singapore shows one market's emphasis. The overall stance on compliance agents is in AI compliance agents.

Where Pryme fits

In Pryme, every fact in the timeline arrives through the Pryme Context Protocol with its provenance: company filings as signed, the customer's and relationship manager's accounts as declared, press as external. The narrative therefore shows its own evidential strength. Establishing source of wealth is outside the agent's role; it drafts at depth 2 and an analyst decides.