Conveyancing Client Onboarding With AI: ID, Forms and Reminders

Build onboarding as a fixed sequence: send one checklist on day one, collect forms and documents through a single route, remind on a set cadence, and let AI draft the checklist, the questions and the reminders. A named person must verify identity and money-laundering evidence, record the result and approve anything that states a check has passed.

Files stall in the first fortnight for dull reasons: a client sends a photo of one side of a passport, a form is half completed, a utility bill is too old, nobody chases on Thursday. Those are exactly the jobs AI does well. The jobs it should not do, deciding that an identity is genuine or that money is clean, are the ones the firm answers for.

Follow me on Instagram@sagnikteaches

What AI does in onboarding and what it never does

Split the work before you automate any of it, and write the split on one page that every fee earner and case handler can see.

Connect on LinkedInSagnik Bhattacharya

Notice that the right-hand column never shrinks. AI reduces the time a person spends on the left-hand work, so they can spend more on the decisions. Keep your anti-money-laundering policy as the authority, and check what your regulator and insurer currently expect about using AI near those checks before changing a step.

Subscribe on YouTube@sagnikteaches

The six-step sequence from instruction to ready

This sequence assumes a purchase with a mortgage, which is the busiest case. Sales and remortgages need shorter versions of the same steps.

1. Day 0: acknowledge and ask. A message confirms the instruction, names the case handler and fee earner, and attaches the checklist and the link to upload.
2. Day 0 to 1: identity and funds request. The client gets a plain explanation of why the firm needs ID and evidence of where the money comes from, and exactly what counts.
3. Day 1 to 3: forms. The client completes the information and contents forms with prompts that explain each question.
4. Day 2 to 6: reminders. The cadence below runs, with a human call at day four.
5. Day 3 to 7: checks. A person verifies identity documents, reviews funds evidence and records the outcome.
6. Day 7: ready. A status message tells the client what happens next. The case handler ticks "onboarding complete" only when every item has a named checker.

The target timings are examples, not promises. Clients who are slow to reply need the full cadence, and some matters, such as a purchase with a gifted deposit, always need more steps. Treat the sequence as a skeleton the firm tunes after measuring its own files.

A worked example: a first-time buyer on a leasehold flat

Illustrative. A client instructs the firm on Monday to buy a leasehold flat with a mortgage and a deposit partly gifted by a parent. The case handler runs the checklist builder with the matter type, and gets a list of nine items: two forms of identity, proof of address, a mortgage offer, deposit source evidence, a signed declaration from the person giving the gift, the property information form, the fittings and contents form, and a note on any leasehold queries the client already has.

By Wednesday the client has uploaded a passport photo page, a bank statement and the information form with four blank fields. The completeness prompt, run on the checklist and a list of what arrived (not on the documents themselves), returns: "Missing: signed gift declaration, proof of address, fittings and contents form. Information form has blanks at questions 6, 9, 14 and 22. Bank statement dated 11 weeks ago: check against the firm's age limit." The case handler sends a single message listing five items rather than five messages. By Friday the gift declaration arrives, and a fee earner does the identity and funds review on Monday. Ready, in this example, on day seven.

Write the checklist once, by matter type

A good checklist is the cheapest and highest-impact piece of onboarding. Draft one for each matter type the firm handles: purchase with mortgage, cash purchase, sale, sale and purchase together, remortgage, transfer of equity, and new-build purchase. Ask AI to produce the first draft from the firm's own internal list, then have a fee earner correct it, because a missing item costs days.

You are helping a conveyancing firm write a client onboarding checklist.
Matter type: [purchase with mortgage / sale / remortgage / ...]
Our internal list of what we always need is below.
Produce a client-facing checklist in plain English, grouped under:
"Identity", "Money", "Forms", "About the property".
For each item give: what we need, why we need it in one sentence, and
what a good copy looks like (for example "all four corners visible").
Do not give legal advice. Do not mention fees. Use short sentences.

Internal list:
[paste list]

The sample output for the "Identity" group might read: "Passport photo page. We need this to confirm who you are, because we are required to check identity before we act. A good copy shows the whole page, edge to edge, with every line readable and no glare." The fee earner's edit is usually to swap the generic "bills" for the firm's actual accepted proof of address list, and to add the age limit on documents. Once corrected, the checklist lives in the firm's document management system and the case-management template, so no one drafts it from scratch again.

Review the checklist every quarter against files that stalled. If three files in a row were held up by a missing document that was not on the list, add it. The general guide to automating client onboarding covers templates and contracts for other businesses, and is a useful reference for the wider structure.

Turning forms into questions clients can answer

Property forms are written for professionals. Clients hesitate, skip questions or answer them wrongly. Use AI to produce plain-language prompts that sit beside the official form, never in place of it. Give it the form's questions (not a client's answers), and ask for a one-line explanation of what each question is really asking and an example of a good answer.

Below are questions from a property information form. For each one, write:
1. What this is asking, in one plain sentence.
2. An example of a complete answer for an invented flat, labelled "example".
3. A warning if a wrong answer could cause a problem later.
Do not suggest how the client should answer for their own property.
Do not change the question wording; I will show your text beside it.

[paste questions]

Sample output for a question about disputes with neighbours might be: "What this is asking: whether you have ever argued with a neighbour about boundaries, noise or access, even if it was settled. Example: 'No disputes.' or 'In 2022 we agreed to share the cost of repairing the fence; settled in writing.' Warning: leaving out a past dispute can cause problems for the buyer and for you." The fee earner checks that the explanation matches what the question means in the firm's jurisdiction, and removes anything that drifts into advice.

Checking returned forms for gaps

Once a form comes back, run a completeness check that looks at structure, not substance: blank fields, "see attached" with nothing attached, dates in the future, a signature missing. Use the firm's approved system for this rather than pasting a form with names and addresses into a general chat tool. Where your system cannot do it, a case handler can check blanks against the checklist by eye in two minutes, and AI is not worth the risk for that step.

Keep the output as a list, such as "Questions 6, 9, 14 and 22 are blank. Question 17 says 'see attached' but no attachment is on the file." The case handler then writes to the client. The AI should not fill gaps, guess an answer or assume "not applicable".

Reminder cadence that clients tolerate

Reminders work when they are specific, short and escalate to a person. A client who gets "Please send your documents" every day stops opening them. A client who gets "We are waiting for your proof of address, so we cannot ask the lender to release funds" on day four acts.

Draft the reminders in your own voice once and store them as templates. This prompt produces the day-two and day-six versions from your tone sample:

Here are two short messages in the tone our firm uses with clients:
[paste two real, anonymised examples]

Write a day-2 reminder and a day-6 reminder for a client who still owes:
[list]. Day 2 is friendly and practical. Day 6 explains what is held up
because of the missing items, calmly, without legal threats.
Keep each under 90 words. Use "[first name]" and "[your name]".
Do not promise a completion date.

A sample day-six output: "Hi [first name], we are still waiting for your proof of address and the fittings and contents form. Until we have them we cannot complete our checks, which means we cannot yet request the lender's funds. If something is getting in the way, reply with a word or call me and I will help. [your name]." The fee earner would change "I will help" to a specific person and number the firm uses.

A reminder mistake and the rule that prevents it

Realistic failure: the automation sends a day-six message to a client who phoned on day five to say the document was in the post. The client is irritated, and the case handler learns about it from a complaint. The rule: automated reminders pause the moment a person logs a conversation on the file, and the case handler restarts them. If your system cannot pause by conversation, use scheduled drafts that a person sends, not automatic sends.

Identity and money-laundering steps that stay human

These checks are legal obligations on the firm. The standard answer to "can AI do this for us?" is that AI can help the person who does it, and the person remains responsible. Concretely:

• Collection: AI-drafted instructions tell the client what to send and how to photograph it. Fine.
• Technical checks: an approved electronic identity provider may perform checks the firm's policy accepts. The person reviewing the result records it and decides whether the match is adequate.
• Judgement: a face against a photo, a discrepancy between documents, a source-of-funds story that does not add up. A trained person decides, and escalates to the firm's money-laundering reporting officer where required.
• Record: who checked, what they saw, the date, the outcome and any follow-up. A free-text note is not enough on its own; use a structured record in the case system.

Ask a simple test question of every AI feature in the process: "If this is wrong, does a person see it before the client or the lender relies on it?" If the answer is no, the feature does not belong in identity or funds work. For deepfake and impersonation risk in particular, the fraud controls in the call-back procedure tutorial apply from the first contact, because a synthetic voice or forged document can appear at onboarding as easily as at completion.

Source-of-funds summaries: help with the paperwork, not the judgement

A case handler can use AI inside an approved system to turn a pile of bank statements into a neutral list of deposits, dates and amounts, which is slow to do by hand. The summary does not say whether the money is acceptable. A person reads the original statements, compares the list, and writes the assessment in the file record.

Illustrative failure: a summary lists a $38,000 deposit as "salary and savings" because the statement label said "transfer". The fee earner, reading the original, sees the transfer came from an account in a third party's name. The funds story changes, and the file needs a gift declaration and further questions. The lesson is to have the AI quote the label exactly as it appears and never characterise it.

Clients who cannot follow the standard route

A sequence built around an upload link fails for some clients, and the failures are predictable. Plan for them in the template instead of improvising on the day.

• The client without a smartphone or reliable internet. Offer a paper checklist and a drop-off or post option, and have a person copy documents into the file. The checklist wording stays the same; only the route changes.
• The client who finds forms hard. Offer a fifteen-minute call where the case handler goes through the form question by question, using the plain-language prompts as a script. Note on the file that the call took place.
• Joint buyers with different timetables. Each person needs their own identity and funds evidence. Send two checklists with each person's name at the top, and track completion per person, not per file. A file where one buyer is complete and one is not should show amber, never green.
• Clients who write in another language. AI can translate your checklist and reminders, which helps, but have a person who reads that language check the key sentences if the firm has one. Documents the client sends still go to the fee earner in the original, and any translation of a document the firm relies on follows the firm's own rules.

Sale and purchase on the same day

Illustrative. A client selling one house and buying another has two matters that depend on each other, and onboarding doubles the paperwork. Run one identity check for the person and link it to both matters, but keep two checklists: the sale needs the seller's forms, deeds access and mortgage redemption details, while the purchase needs deposit evidence and the lender's requirements. Ask AI to produce a combined timeline from both checklists, with the dependencies marked, such as "purchase mortgage offer needed before exchange on the sale". The case handler checks the dependencies against the actual chain and sends the client a single message each week summarising both files, so the client is not chased by two people for two things.

Measuring whether onboarding got better

Pick three numbers and track them for ten files before and after the change. Time from instruction to "onboarding complete". Number of messages sent to the client per file. Number of files where the first search or enquiry was delayed by missing onboarding items. If time falls from twelve days to eight, messages per file stay the same or fall, and delayed starts drop, the sequence is working. If messages rise sharply, the reminders are too eager.

Also collect two pieces of qualitative feedback: ask three clients after completion what was confusing, and ask the case handlers which template they edit most. The most edited template is your next improvement. For firms choosing software to run this sequence, the tutorial on cutting admin on each transaction and the one on AI for ID checks, searches and chasing help you decide which tool should do what.

A one-page onboarding policy the team can follow

Write the rules down so they survive staff changes. A page is enough:

• Which tools may be used for onboarding, and for what.
• That no client identity document goes into any tool the firm has not approved.
• That AI-drafted messages are reviewed by the case handler before sending, apart from approved templates.
• That identity, funds and risk decisions are made and recorded by a named person.
• That reminders pause when a person speaks to the client.
• That any suspicion of fraud or impersonation goes straight to the fee earner and the reporting officer.
• How and when the checklist, forms and templates are reviewed.

Once it exists, add a ten-minute walk-through to the induction for every new case handler, and ask them to run one mock onboarding with a colleague playing the client. It will find the awkward sentence in the template that no amount of reading catches.

Onboarding questions conveyancing teams ask

Can AI verify a client's identity?

It can help you collect documents and spot missing pages or expired dates, but the decision that an identity is verified stays with a named, trained person using your approved checking method. Record who checked, what they saw and when. Never let an AI tool mark a check as passed by itself.

How often should we chase a client for documents?

A workable pattern is a message after two working days, a second after four with a phone call from the case handler, and a third at six that explains what is blocked. Stop automated messages once a person has spoken to the client, so they are never chased for something already sent.

Can we put client documents into a general AI chat tool?

Not unless the firm has approved that tool, understands where it processes data and has the right contract terms. Use AI on the templates and checklists, which contain no client data, and keep identity documents inside your case-management or onboarding system.

What should the AI never write to a client?

Anything that gives legal advice on their specific matter, states that a check has passed, quotes figures it was not given, or promises a completion date. Drafts should ask questions and list what is needed. A fee earner approves anything that goes beyond that.

Further reads

• Conveyancing AI Software: What a Fee Earner Should Test in a Demo — Test onboarding features before buying software.
• AI for UK Conveyancers: Reports on Title and the Checks You Keep — What comes after onboarding: reports and checks you keep.
• How Small Law Firms Use AI to Answer Client Status Questions — Keep clients informed once the file is moving.
• How Small Law Firms Use AI to Draft Letters and Routine Documents — Draft routine client letters with review steps.
• AI for Small Law Firms: What to Automate First — Decide whether onboarding is your first automation.
• AI Readiness Checklist for Accountants, Solicitors, Consultants — Check the firm is ready before adding tools.
• Which AML Client Checks Can an Accountant Automate With AI? — A step-by-step map of client due diligence in an accounting practice: what to automate, what AI may only draft, and what must stay with a person.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: general conveyancing practice and anti-money-laundering good practice; check your regulator's current guidance and your insurer's requirements.