How to Automate Client Onboarding in a Small Accounting Practice

Map every step from "yes" to the first piece of work, then automate in this order: proposal, engagement letter and payment in one signing flow; an intake form that creates the client in your practice software; identity checks with a human sign-off; document requests with automatic chasers; and a welcome sequence. AI drafts and checks; people make risk decisions.

The order matters because each stage feeds the next. Practices that start with a clever welcome email end up with a nice first impression and the same manual scramble behind it. Start at the signature, because until a client has signed and set up payment, everything after it is at risk of wasted effort.

Follow me on Instagram@sagnikteaches

Stage 1: Map what happens now (one to two hours)

List every step, who does it, how long it takes and where it waits. Here is the map from an illustrative four-person accountancy practice taking on about eight new clients a month:

Connect on LinkedInSagnik Bhattacharya

Two numbers to note before you automate anything: 3.5 hours of staff time per client and 23 days from acceptance to first job. Those are your baseline.

Subscribe on YouTube@codingliquids

Stage 2: One signing flow for proposal, letter and payment (1-2 days)

A proposal tool that combines the proposal, engagement letter, e-signature and payment set-up removes four steps from the map. Ignition is one widely used example: the client reviews, signs and gives payment details in one session, the letter comes from your templates, and it connects to Xero, QuickBooks, Xero Practice Manager, Karbon and Zapier. Several practice management systems offer their own versions; check yours before buying another tool.

Set-up time goes on the templates: your service list with prices, the engagement letter with a schedule per service, and your standard terms. The drafting side of those letters, including where AI helps and where it must not, is covered in AI engagement letters for accountants.

On the stage 1 map, the four steps this replaces took 85 minutes per client: 30 for the proposal, 25 for the letter, 15 chasing the signature and 15 on payment details. With templates in place, choosing services and checking the proposal before it goes takes about 15. That's 70 minutes saved per client, a little over nine hours a month at eight new clients, before counting the days no longer spent waiting for a printer and a scanner at the client's end.

Plan for the client who wants to change the terms. A new client's finance director asks to remove the limitation of liability clause before signing. The signing flow can't negotiate, so that request leaves it: the partner decides, amends that client's letter from a template variant, and re-sends through the same tool. Keep a record of which clients have non-standard terms, because an automation that later re-issues "the standard letter" at renewal will quietly undo the agreed change.

Stage 3: An intake form that creates the client record (half a day)

Once the letter is signed, the client gets one form. Whether it is Microsoft Forms, Google Forms, a form tool such as Jotform, or your practice software's own, the rule is the same: one form, and the answers create the client record automatically, so nobody retypes them. A filled-in field list for a limited-company client:

INTAKE FORM: limited company client
Company name, registration number, registered address
Trading address (if different)
Year end; date of incorporation
Directors: name, date of birth, home address, email, phone
Shareholders over [threshold]%: name, holding
Main contact for day-to-day queries
Accounting software in use (none / Xero / QuickBooks / other)
Payroll: yes/no, number of employees, pay frequency
Previous accountant: name, email (for professional clearance)
How you'd like us to contact you: email / phone / portal

An automation (Zapier or Make, or your practice software's own) then creates the client and contacts in the practice management system and a draft organisation in the accounting software. On Zapier, the form submission is the trigger and costs no task; each record created is one task, so a client with two directors might use four or five tasks.

It's worth adding up the whole flow before choosing a plan. Per client, with illustrative counts: about 5 tasks for the records, 1 to send the ID link, 3 for the authority, clearance and document requests, up to 3 chasers, and 2 for the first job and welcome email. Call it 14 tasks a client, or about 112 a month at eight clients. That rules out Zapier's Free plan, which allows 100 tasks a month and only two-step Zaps, and sits comfortably inside the 750 on Professional, with room for a busy month.

This is also the first good place for AI. Before anyone relies on the answers, an assistant on a business plan can check them for gaps. The prompt, and an illustrative reply:

Below are a new client's intake form answers. Check for: missing
required fields, dates that don't make sense, a trading address with
no postal code, directors listed without dates of birth,
and any shareholder percentages that don't add up to 100. List only
problems, each with the field name. Don't correct anything yourself.

Illustrative output: "1. Directors: second director has no date of birth. 2. Shareholders: holdings total 90%. 3. Year end given as 31 February, which isn't a valid date. 4. Previous accountant: email missing."

That list becomes the text of a single follow-up email, sent once rather than in four separate chasers. Note what the assistant didn't catch: the trading address was a residential flat, which for this client's line of business was a question for the partner, not an error. Checks for sense and judgement stay with people.

Stage 4: Identity checks with a person signing off (2-3 hours)

Most practices now use an electronic identity verification provider: the client gets a link, photographs their ID and takes a selfie, and the provider returns a result. Providers charge per check, so compare per-check prices and what each check includes. Connect the provider so the link goes out automatically when the intake form is complete.

The step that must not be automated is the decision. The verification result and the documents go to the partner responsible, who completes the client risk assessment your anti-money-laundering rules require and records the decision. Build the workflow so nothing after this stage can start until that sign-off is recorded. A realistic failure to design against: an automation that treats "ID check completed" as "client approved" and sends the welcome email and first-job setup the moment a check returns, including for a client whose check came back with a mismatch flag nobody had read.

AI can make the partner's review quicker without making it for them. Given the verification result, the intake answers and the uploaded proof of address, an assistant on a business plan can list differences for the partner to look at. An illustrative summary:

"Name: passport gives full first name; intake form uses a shortened version. Address: utility bill shows the same street and postal code as the intake form but no flat number. Date of birth: matches. Verification provider: pass, with a note that the selfie was retaken twice. Nothing here decides the risk assessment."

Each item takes the partner seconds to resolve: the shortened name is normal, the missing flat number gets one question to the client. What the summary can't know is whether this client's business, funding or relationships raise a risk, which is exactly why the decision field stays with the partner.

Stage 5: Authorities, prior-year information and chasers (half a day)

After sign-off, the automation sends the requests you always make: authorities to act, the professional clearance email to the previous accountant, and the list of documents needed for the first job. Each request gets a due date and a chaser sequence, for example a reminder at 5 days, another at 10, and a task for a person at 15 days if things are still missing.

AI earns its place here in two ways. It drafts chaser wording that varies with what is actually missing, instead of a generic "please send your documents", and it can sort and label what arrives. The difference in a second chaser, before:

"Just a reminder that we're still waiting for some documents from you. Please send them as soon as you can."

And after, drafted from the checklist of what has and hasn't arrived:

"We have your authority form and last year's accounts, thank you. Two things are still missing before we can start: bank statements for the business account from April to June, and an invite to your accounting software (the steps are in the link below). The statements are the one holding things up."

The second tells the client they've done most of it, names exactly what's left, and says which item matters most. Review the wording once as a template; the variable parts come from the checklist, not from the AI's guesswork. Chasing missing client records with AI covers the chasers in depth, and whether AI can sort and file incoming documents covers the filing side.

Stage 6: The welcome and the first job (two hours)

When the last required item arrives, the automation creates the first job from a template in your practice software, assigns it and sends the welcome email. A filled-in version, illustrative:

Subject: Welcome to [practice name], what happens next

Hi [first name],

Thanks for sending everything through. You're all set up.

Your first job with us is your company accounts for the year ended
31 March. [Manager name] is your day-to-day contact, and will be in touch by
Friday 17 October with any questions on the records you've sent.

A few useful things:
- Deadlines: we'll email you at least four weeks before anything
  is due from you.
- Queries: reply to this email or call [Manager name] on [number].
- Documents: please upload to [portal] rather than email, so they
  go straight to your file.

We're glad to have you on board.
[Partner name]

Draft the template once with AI if you like, then fix it. The personal details come from the client record, so the email is accurate without anyone typing it.

Who does what once it's automated

It helps to write down, stage by stage, what the automation does, what AI does and what a person must still do. Pin this next to the workflow so nobody quietly hands a decision to software:

The last cell is deliberate. However smooth the automation, a short call or personal email from the person who will do the work is what makes a new client feel looked after.

Adapting the flow for sole-trader clients

The same flow works for sole traders with fewer fields and one less check. The intake form drops the company number, shareholders and directors, and asks instead for the trade, start date, whether the client is registered for any business taxes, and which bank accounts they use for the business. There is one person to verify rather than several. In practice the flow for a sole trader often completes in under a week, and the most common delay moves from identity checks to the bank statements for the first year. Set a separate chaser sequence for statements, with its own wording, rather than folding them into the general document list where they get lost.

The practice after a month on the new flow

For the illustrative four-person practice, the new flow looked like this after its first month:

Before switching everyone over, the practice ran four test clients through the whole flow, each built to break something:

1. A company with two directors. Expect two contacts, two ID links and one client record. First run: correct.
2. A director who is already a client through another company. Expect the existing contact to be linked, not duplicated. First run: duplicated, which led to the email search in the list below.
3. A sole trader. Expect the shorter form and no shareholder questions. First run: correct.
4. A client who abandons the form halfway. Expect a reminder at 3 days and a task for a person at 7. First run: nothing happened, because the form tool only sends completed submissions. The fix was a reminder triggered from the signed letter instead.

Two failures from four tests is a normal first result, and both are far cheaper to find with made-up clients than with real ones.

Setting it up took about 20 hours over four weeks. At 18 hours saved a month, the practice recovered that within the second month, and the shorter wait to first job meant fees were billed earlier too.

What breaks, and how you'll notice

• Duplicate records. A director who already runs another company with you gets created twice. Add a search for existing contacts by email before creating new ones.
• The wrong contact gets the letter. In one illustrative case, the automation sent the engagement letter to the email of whoever filled in the enquiry form, who was the client's bookkeeper, not a director. Always send letters to a named director confirmed on the proposal.
• Clients stall with no end state. A chaser sequence that just stops leaves half-onboarded clients in limbo. The final step should always create a task for a person.
• Documents arrive by email anyway. Clients reply with attachments. Decide whether someone files them or the automation does, or they'll sit in an inbox.
• AI steps drift. If you change the intake form, update the gap-check prompt, or it will check for fields that no longer exist.

Check the flow monthly with three numbers: days from acceptance to first job, the share of new clients complete within 14 days, and how many onboarding steps someone did by hand. If the first number creeps up, find the stage where clients are waiting. For the general version of this workflow outside accountancy, see automating client onboarding with AI, and if you are rolling AI out across a bookkeeping team at the same time, rolling out AI without losing control covers the wider change.

Onboarding automation questions from practice owners

Which tool should run the whole onboarding flow?

Usually your practice management system, with a proposal tool in front of it and an automation platform joining the gaps. Avoid building the flow inside a tool nobody opens daily. If your practice management system has its own onboarding templates, test those first; only add Zapier or Make for the steps it cannot do.

Can AI decide whether a client passes our anti-money-laundering checks?

No. AI can summarise documents, compare names and addresses, and flag inconsistencies for a person to look at. The risk assessment and the decision to accept a client should be made and recorded by the person responsible under your rules. If you are unsure what your obligations require, ask your professional body or compliance adviser.

How long does setting this up take?

For a small practice, roughly 15 to 25 hours spread over three or four weeks: a couple of hours mapping, a day on the proposal and letter templates, and the rest on the intake form, automations and testing. Test with two friendly clients before switching everyone over.

What if a client won't use online forms?

Keep a manual route. A member of staff can complete the intake form on the client's behalf during a call, which still creates the record and starts the same automations. The aim is one consistent flow inside the practice, not forcing every client to use every screen.

Further reads

• What an AI Implementation Looks Like in a Small Accounting Firm — How onboarding fits into a wider AI set-up in a small firm.
• AI Document Processing: Turn PDFs and Scans Into Usable Data — Turning the documents clients send into usable data.
• ChatGPT Prompts for Bookkeepers: Client Queries, Chasers, Notes — Prompts for the client queries that follow onboarding.
• How Much Does AI Cost a Small Accounting Firm? — What the tools in this workflow add to the practice's AI bill.
• How Small Accounting Firms Use AI Day to Day: Real Examples — Other day-to-day uses once onboarding runs itself.
• Is It Safe for an Accountant to Use ChatGPT With Client Data? — The data rules behind the AI steps in stages 3 and 5.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Ignition engagement letter product page; Zapier pricing and task-counting help; Karbon AI feature page; Microsoft 365 business plan pages. Checked September 2026. The practice, times and figures in the example are illustrative.