Source of funds is about one specific sum: where this money came from. The agent reads the explanation and documents, traces the funds from origin to your account, checks that amounts, dates and names line up, and drafts a finding with gaps listed. An analyst decides whether the source is adequately explained, and any acceptance outside policy needs senior approval.
What is a source of funds review, and who owns it?
Source of funds is the origin of the particular funds used in a transaction or to fund an account: a property sale, an inheritance, a bonus, the sale of a business, a loan. It is distinct from source of wealth, which is how the customer built their overall wealth; the two reviews are often done together, and reviewing source of wealth evidence covers the other half. Under a risk-based approach, firms check source of funds for higher-risk customers, large or unusual transactions, and as part of enhanced due diligence.
KYC or EDD analysts review the evidence, under the MLRO's policy, with senior approval for high-risk cases. The work is document-heavy: completion statements, probate documents, sale agreements, bank statements, payslips. An agent reads these quickly and checks consistency. Whether an AI agent can verify source of funds and source of wealth sets out the boundary.
What does the agent prepare, step by step?
- Agent, depth 1: reads the explanation. The customer's stated source, amount and timing, recorded as declared.
- Agent, depth 1: lists the evidence your policy requires for that kind of source, such as a completion statement for a property sale or a grant of probate and estate accounts for an inheritance.
- Agent, depth 2: drafts the request for missing evidence, using approved wording. A person or an approved chase process sends it.
- Agent, depth 1: reads each document. Extracts names, dates, amounts, accounts and issuers, and records each document's provenance.
- Agent, depth 1: traces the funds. From the origin, such as the sale proceeds, through any intermediate accounts, to the payment into your account. Each step needs a document.
- Agent, depth 1: checks consistency. Amounts reconcile, dates are in order, names match the customer or explained parties, and the documents are consistent with each other and with the customer profile.
- Agent, depth 2: drafts the finding. The source as evidenced, the trace, any gaps or inconsistencies, and anything that looks altered.
- Person, analyst: decides whether the source of funds is adequately explained.
- Person, senior approver: approves exceptions and high-risk cases under policy.
| Step | The agent does | Depth | Stays with a person |
|---|---|---|---|
| Explanation | Records as declared | 1 | None |
| Requirements | Lists evidence needed | 1 | Setting the policy |
| Request | Drafts with approved wording | 2 | Sending |
| Documents | Extracts and records provenance | 1 | Judging authenticity |
| Trace | Links origin to arrival | 1 | None |
| Consistency | Reconciles amounts, dates, names | 1 | None |
| Finding | Drafts with gaps | 2 | Analyst decides |
| Exceptions | Nothing | None | Senior approver |
Which controls apply, and what evidence does each leave?
The trace is the evidence. A finding that does not show how the money got from origin to your account is incomplete. Evidence: the trace, step by step, with documents.
Provenance graded. A completion statement from a regulated conveyancer, sent directly, is stronger than a scan the customer emailed. Evidence: provenance on each document.
Inconsistencies listed, not resolved. If the sale price on the completion statement differs from the amount received, the agent lists it. Explanations such as fees and mortgage redemption are checked against documents. Evidence: the inconsistency list and how each was resolved.
Signs of alteration flagged. Fonts, alignment, totals that do not add up, metadata. The agent flags; the analyst judges. Evidence: flags with the analyst's decision.
Decision named. Evidence: the analyst's decision and any senior approval.
What should you measure?
- Cases closed without further requests, a sign that requirements are clear to customers.
- Inconsistencies found by the agent and confirmed by analysts.
- Altered documents detected.
- Time from evidence received to decision.
- Requests sent per case, since several rounds of requests usually mean the first one was unclear about what was needed.
- QA findings on source of funds files, for example in Consumer Duty outcome monitoring where delays harm customers.
Where does it usually go wrong?
Accepting the explanation as the evidence. "From the sale of my house" is a claim. The evidence is the documents that show it.
Breaking the trace. Funds pass through an account at another bank before arriving. Without that account's statement, the trace has a gap. The finding should say so.
Over-asking. Requesting the same evidence for a modest, well-explained transfer as for a large unexplained one frustrates customers and wastes analyst time. The requirements should scale with risk.
Third-party funds. Money arriving from a parent, partner or company needs its own explanation and, often, the third party's source of funds. The agent should treat each contributor as a separate trace and show how the parts add up to the total.
Customer delays treated as refusal. Gathering documents takes time. Distinguish a customer who cannot provide evidence from one who has not yet done so.
Currency and timing gaps glossed over. Sale proceeds received abroad in one currency arrive months later in another, at a different amount. The agent should show the conversion and the time gap explicitly, and say what happened to the money in between, rather than matching approximate figures.
The overall stance on compliance agents is in AI compliance agents.
Questions people ask
Can the agent accept a source of funds as verified?
No. It can show the trace is complete and consistent. Deciding the source is adequately explained, especially in a high-risk case, is the analyst's judgement.
How much evidence is enough?
Enough to trace the specific funds to a legitimate origin, proportionate to the risk. Your policy should set expectations per source type and risk level.
Can open banking data replace statements?
Where the customer consents and your process supports it, account data from the bank itself is stronger than an uploaded statement because it comes from the issuer. The trace logic is the same.