Fracas Digital • Aug 1, 2026 • 8 min read
AI Client Onboarding for Accountants: The UK Practice Guide
Most of the repetitive work in new client onboarding for UK accountancy practices can now run on automated workflows. Intake form processing, document extraction, ID verification, engagement letter pre-population from firm templates, and account provisioning in practice software can all be handled without someone driving each step manually. The two steps that remain human are AML sign-off and the final engagement letter review. Both carry regulatory weight and neither should run unsupervised.
Every accounting practice knows the delay. A new client agrees to proceed, the proposal goes out, and then two weeks pass before they are actually in the system. Somewhere in between: an AML questionnaire sitting in someone's drafts folder, an engagement letter that went back and forth three times, a Xero account that nobody set up yet. For practices onboarding 20 or 30 new clients a year, that delay costs real partner time and occasionally costs the client relationship.
Most of the delay is not professional work. It is templated repetition wrapped in manual admin, and that is where agents come in.
What onboarding actually involves in a UK practice
A new client typically moves through five stages before they are properly set up. The real delays cluster in stages two and three.
Enquiry and scope confirmation. A phone call or contact form, followed by email exchanges to agree on services and fees. Most practices handle this with a mix of calendaring tools and personal email, which means details fall through the gaps.
Identity verification and AML due diligence. Under the Money Laundering, Terrorist Financing and Transfer of Funds (Information on the Payer) Regulations 2017, ICAEW-supervised practices must carry out customer due diligence before establishing a business relationship. That means verifying client identity using reliable, independent source documents, understanding the nature and purpose of the engagement, and for higher-risk clients, completing enhanced due diligence. Somebody has to collect the documents and record the outcome. Most practices estimate this takes 30 to 60 minutes per new client, not counting chasing.
Engagement letter. A formal agreement covering scope, fees, responsibilities, and the basis on which the firm acts. ICAEW professional standards require one before any work starts.
Account setup. Client reference numbers, service codes, billing schedules, Companies House filing dates, and Xero or practice management system records.
Service handover. Whoever is running the accounts gets briefed on the client and engagement scope, plus any flags from the onboarding conversation.
An AI agent is software that carries out a defined sequence across the systems you already use, without someone driving it step by step. For onboarding, that means three or four automation points working in sequence.
Where AI fits into the workflow
Intake form processing. When a new client completes an onboarding questionnaire covering business type, trading history, current accountant, and expected services, an agent reads the response and creates a draft client record in your practice management system, with any missing information flagged before the file is even opened. Nobody types the same details into three different places.
Document extraction and filing. A client uploads their most recent tax return, bank statements, Companies House certificates, and ID documents. The agent extracts key data fields, including UTR number, registered address, company number, and most recent year's turnover, and places them in the right fields. For practices running Xero, that data can pre-populate the Xero client record via the API, so the account is live before the engagement letter is counter-signed.
AML document assembly. The agent checks what has been uploaded against the firm's CDD checklist and chases anything missing. Once the file is complete, it assembles the AML package in the format the MLRO reviews. The CDD decision stays with the MLRO, exactly as the regulations require. But the assembly work, which typically runs 30 to 45 minutes per client, can be reduced to a few minutes of agent processing plus the time for the MLRO to review a complete, pre-assembled file.
Engagement letter pre-population. Based on the agreed scope and fee schedule from the proposal, the agent drafts the engagement letter from the firm's templates. Fee, period of engagement, named services, signatory details: all pre-filled. The partner checks and adjusts, then sends via DocuSign or Adobe Sign. After counter-signature, a triggered workflow provisions the client record in practice software and sets the first filing deadlines. The welcome email with portal access goes out automatically.
That sequence currently takes most practices one to three weeks. With a properly built workflow, the time from completed questionnaire to signed engagement letter can get down to two or three days. Most of the remaining time is waiting for the client to counter-sign.
AML and CDD: the step most automation guides skip
Every US-facing SaaS tool in this space treats onboarding as a document collection problem. Upload documents and fill in fields. For UK supervised practices, that framing misses the regulatory context entirely.
The Money Laundering Regulations 2017 require accountancy firms supervised by professional bodies to conduct customer due diligence before establishing a business relationship. That includes identifying the client and verifying their identity using reliable, independent source documents; understanding the nature and purpose of the relationship; and for companies, identifying and verifying the beneficial owner.
Where agents help is with the checking work: confirming that the correct documents are present and flagging inconsistencies where names or addresses don't align across them. They can also assemble the AML file for the MLRO to review. What they cannot do is make the CDD decision. ICAEW's guidance on AI in accountancy practice is clear on this. The professional remains responsible for the output, regardless of what produced the first draft.
Build the workflow so the human sign-off is a genuine gate, not a formality on a pre-approved file. One practical point: AI document extraction works well on structured documents such as bank statements and Companies House filings, but is weaker on poor-quality scans or unusual formats. Build in a check that flags low-confidence extractions for manual review rather than passing them through automatically.
The engagement letter stage
ICAEW members are required to issue an engagement letter for every new client before work begins. The letter covers services, fees, applicable professional standards, liability, and data protection obligations.
The content is largely standard. The fee schedule, named client, period of engagement, specific services, and signatory name can all be pre-populated from the data the practice already holds after the proposal stage. What requires judgement is anything that deviates from the firm's standard terms, which a partner should be reading regardless.
The practical workflow: the agent generates the letter from the template, the partner reads and adjusts, sends via e-signature tool, and the countersigned copy files automatically. No printing, no chasing the post.
One boundary worth stating clearly: never let the agent draft client-specific legal language or vary liability caps without human review. Template population is safe; writing novel contractual terms is not.
MTD and why digital-first intake matters now
Making Tax Digital for Income Tax applies from April 2026 to sole traders and landlords with qualifying income above £50,000. The threshold drops to £30,000 in April 2027 and £20,000 in April 2028.
MTD-mandated clients need to submit quarterly updates via compatible software. That means every new client in scope needs to be set up in Xero, Sage, or QuickBooks from day one, not onboarded onto paper-based processes and migrated later.
An AI workflow that extracts client data at intake and provisions the Xero account automatically ensures that MTD-mandated clients are correctly set up in the first week. Onboarding them manually and retrofitting MTD compliance six months later adds hours you will not be able to bill. For practices onboarding 20 or 30 new clients a year, that is reason enough to look at this properly before the April 2027 threshold expansion.
What this looked like in practice
Fracas built a client intake and onboarding workflow for a mid-size UK accountancy practice that was spending around 12 hours per new client on the steps above. The setup: a structured intake form feeding into their practice management system, an agent that processed uploaded documents and extracted the key entity data, then routed the assembled AML file to the MLRO.
After six weeks of supervised running, onboarding time per new client dropped from 12 hours to under three. The MLRO review step remained exactly as it was, with the same person making the same decision from a complete, pre-assembled file rather than chasing documents to assemble it themselves. The change was in everything around that decision.
The practice runs IRIS for accounts production. The agent does not write directly into IRIS, which would require integration work specific to their configuration, but stages data in a pre-formatted handover document that the practice manager transfers in 15 minutes rather than two hours.
That result is not unusual for IRIS-based practices. Most see between 60% and 80% of their onboarding admin disappear in the first few months, with the remainder being professional judgement calls that should not be automated.
If you want to understand what the broader range of agent automation looks like for a UK practice, our guide to AI agents for UK accountancy practices covers the wider picture. A scoping conversation on what a client intake workflow would look like for your firm starts at our AI agents service page.
Frequently asked questions
What can AI do in client onboarding for accountants?
AI agents handle intake form processing, document extraction and filing, AML document assembly for MLRO review, engagement letter pre-population from firm templates, Xero or practice software provisioning, and chasing clients for missing documents. The two things that must stay with a qualified human are the AML CDD decision and the final engagement letter approval.
Is it legal to use AI for AML checks in a UK accounting firm?
AI can assist with document assembly and consistency checking. The CDD decision under the Money Laundering Regulations 2017 rests with the supervised practice, not the software. Using AI to assemble the file before MLRO review is fine. Using it to approve the file without a human making the decision is not.
How much does AI client onboarding automation cost for a UK firm?
A custom workflow covering intake form processing, document extraction, engagement letter pre-population, and practice software provisioning typically costs between £4,000 and £15,000 to build, depending on integration complexity. SaaS portal tools such as Canopy or Karbon offer lighter versions at per-user monthly costs. The numbers are hard to argue with: at 10 hours saved per client and 20 clients a year, a £10,000 build pays for itself inside the first year at any reasonable billing rate.
Does AI onboarding work with IRIS or Sage?
Xero has a well-documented API that agents can write to directly. IRIS and Sage offer more limited external write access. The common approach is an agent that stages data in a formatted handover document that a practice manager pastes in, cutting the transfer time from two hours to around 15 minutes. Direct write integration is possible for some IRIS modules but typically requires scoping with their API team first.