AML onboarding is where accountancy practices most often stall a new client, and it is also where automation projects most often overreach. Both problems come from the same confusion: treating onboarding as one task, when it is really two — a collection and evidencing task, which automates well, and a risk decision, which must not.
Getting that line in the right place is not a productivity question. It is a supervision one.
Who supervises you, and why it matters here
Anti-money-laundering supervision for accountancy sits in one of two places. Practices supervised by a designated professional body — ICAEW, ACCA, AAT and CIOT among them — answer to that body. Accountancy service providers not supervised by a professional body must register with HMRC, which then acts as supervisor.
Either way, the obligation attaches to the firm and the responsible person within it. It does not attach to a piece of software, and it cannot be delegated to one. Any automation decision has to be made with that in mind: the question is never "can the system do this?" but "if this is later examined, does the responsible person still own the decision, with evidence of how it was reached?"
The part that automates well
Most of the elapsed time in onboarding is not decision-making. It is waiting.
Collecting identity and verification documents. A structured request that states exactly what is needed, accepts uploads, and confirms receipt. Most delay here is a client not knowing precisely what to send.
Chasing what is outstanding. Same discipline as any record chase: specific, sequenced, and it stops the instant a document arrives.
Running the checks. Screening and verification tooling produces a result. The result is an input, not a conclusion.
Collecting the surrounding facts. Beneficial ownership, source of funds information, the structure of the entity, the nature of the business. Structured collection beats an unstructured email thread every time.
Evidencing the sequence. This is the underrated one. A system that records what was requested, when, what arrived, what the check returned and who reviewed it produces an audit trail as a by-product. Practices that do this by hand tend to have complete files and incomplete records of how they got there.
The part that must not
The risk decision stays with a qualified person. Whether to accept a client, whether enhanced due diligence applies, whether an explanation of source of funds is adequate, whether something about the arrangement does not add up — these require professional judgement, and the responsible person owns them.
This is not a cautious framing. It is the actual regulatory position, and the failure mode is specific: a practice that automates the decision does not usually discover the problem when the automation is wrong. It discovers it when the file is examined, and finds it cannot show who decided or why.
A green result is not a decision. Screening tools return outcomes with a confidence attached. Treating a clear result as automatic acceptance converts a tool's threshold into your firm's risk appetite — set by a vendor, undocumented, and applied uniformly to clients who are not uniform.
Suspicion is not a workflow state. If something looks wrong, the obligations that follow are personal and legal. No system routes around that.
Where the gate should sit
A structure that works in practice:
- Automated collection. Specific requests, uploads, receipts.
- Automated chasing. Sequenced, stopping on arrival.
- Automated checks. Run them, record the outputs.
- Assemble the file. Everything gathered, the sequence evidenced, presented in one place.
- Human decision — the gate. A qualified person reviews and decides. The system records who, when, and on what basis.
- Automated everything after. Engagement letter, client setup, welcome sequence, first data request.
Steps 1 to 4 are where the days go. Step 5 takes minutes when the file is properly assembled — which is precisely the argument for automating around it rather than through it. The practice gets the speed without moving the decision.
What this does to the client experience
Worth saying, because it is the commercial case rather than the compliance one.
A new client's first experience of a practice is usually onboarding, and it is frequently the worst part of the relationship: a vague request, a wait, a chase in the other direction, and no visibility into what is happening or why. Clients do not know AML obligations are the cause. They conclude the firm is disorganised.
Structured collection with clear requests and visible progress fixes that without touching the regulatory position at all. The checks are identical. The difference is that the client can see the process working.
Three failures worth designing against
Practices that automate onboarding tend to hit the same three problems, and all of them are avoidable at the design stage rather than after.
The file is complete but the reasoning is not. Every document is present; nothing records why the client was accepted. This passes a casual look and fails a serious one. If the system captures the decision and its basis at the point it is made, the problem never arises.
The automation runs ahead of the gate. Engagement letters issued, client set up in the ledger, welcome sequence sent — before anyone has reviewed the file. The client is now onboarded and the decision has become awkward to reverse. Nothing downstream of the gate should fire until the gate is passed, and that has to be enforced by the system rather than by convention.
Periodic review is forgotten. Onboarding is treated as a one-off, and the file ages. Circumstances change, and a check performed three years ago is evidence of what was true three years ago. Scheduling the review is exactly the kind of remembering software is good at.
The short version
Automate collection, chasing, checks and evidencing — that is where the elapsed time lives, and none of it requires judgement. Assemble the file and stop. The acceptance decision belongs to a qualified person, recorded with its basis, because that is where the obligation sits and no system can hold it for you.
We cover onboarding alongside record chasing and MTD on our AI for accountants page, and the intake side under services.