AML document processing for NI law firms — manual vs automated | Somvilla
Northern Ireland solicitors spend hours every week processing AML identity documents. Here's what automating that actually looks like.
Every NI law firm that handles conveyancing, company formation, or client funds has the same Monday morning task: stack of passports, utility bills, and bank statements, a trainee working through them one by one, and a compliance file that needs to be right before the matter can open.
It works. It’s just slow, error-prone, and an expensive use of a qualified person’s time.
Here’s an honest look at what AML document processing actually requires, how most small NI firms currently handle it, and what the automated version looks like in practice.
What AML compliance requires from NI solicitors in 2026
The Money Laundering, Terrorist Financing and Transfer of Funds Regulations 2017 (as amended) require solicitors to verify client identity before acting. In practice, for a private client or a company formation matter, that means:
- Collecting a valid government-issued photo ID (passport, driving licence)
- Collecting proof of address dated within 3 months (utility bill, bank statement)
- Recording what was checked, when, and by whom
- Keeping that record for 5 years after the matter closes
- For higher-risk matters: enhanced due diligence, which often means source of funds documentation as well
The Law Society of Northern Ireland’s AML guidance adds the requirement to document your risk assessment for each client. That assessment has to be reviewable — “we checked a passport” is not sufficient without recording the document number, expiry, and the name it was matched against.
None of this is new. What has changed is the frequency and scrutiny. The LSNI’s AML supervision work has intensified since 2023, and the expectation is that your compliance records are complete and retrievable, not reconstructed from memory when someone asks.
How most small NI firms currently handle it
The typical process at a firm with under 20 fee earners looks something like this:
A client sends through scanned documents by email, or hands them over in reception. Someone prints them, or saves them to a matter folder with a filename that made sense at the time. A fee earner — often the trainee or a paralegal — opens each document, reads it, and types the relevant details into a compliance form or spreadsheet: document type, number, expiry date, name as it appears, address as it appears.
That data then gets filed in the matter, sometimes cross-referenced against a central AML register, sometimes not. If the matter goes to audit, someone has to find those records and confirm they’re complete.
The failure modes are predictable. Documents get saved in the wrong folder. The compliance form gets filled in from memory two days after the documents were reviewed. Expiry dates get transposed. The central register falls out of sync with the matter files. Someone leaves the firm and their naming convention goes with them.
None of this is negligence — it’s what happens when you ask humans to do repetitive data transcription across dozens of matters a week. The process isn’t broken because your staff are careless. It’s broken because it was designed for a lower volume of work than you’re now handling.
What the automated version looks like
An automated AML document intake pipeline has four stages.
Document intake. The client uploads documents through a web form, or emails them to a monitored address. Either way, documents land in a processing queue rather than someone’s inbox. No manual downloading, renaming, or routing required.
Extraction. An AI model reads each document and pulls out the structured fields: document type, document number, full name as printed, date of birth, expiry date, issuing country, address lines, date on the document. For a passport, that’s reading the machine-readable zone and the biographical data page. For a utility bill, it’s extracting the name, address, and statement date. This takes seconds per document.
Confidence scoring. The system assigns a confidence score to each extracted field. Fields extracted cleanly from a high-resolution document score high. Fields from a blurry scan, or where the AI had to infer rather than read directly, score lower. Anything below a threshold — say, 85% confidence — gets flagged for human review.
Human review queue. A fee earner sees a pre-filled compliance record. The documents are there. The extracted fields are there. Fields flagged for low confidence are highlighted. The fee earner’s job is to verify the flagged fields, not to re-enter everything. They confirm the record is correct, sign off, and the matter opens.
The output is a structured compliance record attached to the matter, with a full audit trail: when each document was received, what was extracted, who reviewed and confirmed it, and when.
What stays human and what gets automated
This is the part that matters most for a profession where professional judgement carries legal weight.
What gets automated: reading the document, extracting the data, checking that an expiry date hasn’t passed, formatting the compliance record, filing it against the matter, updating the central AML register. All of this is data entry and clerical organisation. None of it requires a solicitor.
What stays human: deciding whether the documents are sufficient for the risk level of the matter. Applying your judgement about whether a client’s explanation for their source of funds is credible. Deciding whether enhanced due diligence is warranted. Escalating concerns to your MLRO. None of these decisions are touched by the automated system.
The reassurance that most fee earners need is this: the system doesn’t make AML decisions. It does the reading and writing so that you can focus on the assessment. Your professional judgement is applied to a pre-organised, complete set of information rather than to a pile of PDFs in someone’s inbox.
For a firm doing 30 new client onboardings a month, the difference is roughly 3–4 hours of paralegal time recovered per week. More importantly, the compliance records are consistently structured, consistently complete, and consistently retrievable — which is what your LSNI supervisor will want to see.
ClearDoc AI is built for exactly this kind of document processing work. It handles identity documents, proof of address, and source of funds documents, with a confidence-scored extraction and human review queue. Fixed price, delivered in 7–10 days.