Yes, for the parts that change from client to client: the schedule of services, scope and exclusions, a plain-English summary and change-of-scope letters, drafted from your approved template and service list. The standard terms, liability wording and regulatory clauses should come from your professional body's template or your solicitor, and AI shouldn't rewrite them. A partner approves every letter.
Most of the time an accountant spends on an engagement letter isn't on the terms, which rarely change. It goes on the schedules: working out what this client gets, what they don't, what they must provide and by when. That is exactly the part AI drafts well, provided it drafts from your own descriptions of each service rather than from its general idea of what accountants do.
Which parts of the letter AI should touch
| Section | AI's role | Why |
|---|---|---|
| Standard terms of business | None. Paste as approved | Legal wording; small changes can shift responsibility |
| Limitation of liability, regulatory and data-protection clauses | None | Set by your professional body's template, your insurer or your solicitor |
| Schedule of services | Drafts from your service catalogue | Varies per client; mostly selection and assembly |
| Exclusions and out-of-scope list | Drafts, tailored to the client's situation | Where scope disputes start; worth being specific |
| Client responsibilities and deadlines | Drafts from the catalogue and the client's year end | Dates and duties are easy to assemble, easy to get wrong |
| Cover letter and plain-English summary | Drafts | Clients read this part |
| Fees | Copies from the approved proposal only | Never let AI calculate or suggest a fee in the letter |
If your letters already come from a proposal tool such as Ignition, which builds the letter from templates and collects signature and payment in one step, AI's job narrows to writing and tightening the service descriptions inside those templates. The signing and onboarding side is covered in automating client onboarding in a small accounting practice.
Build the service catalogue before writing a prompt
AI can only be as precise as what you give it. The single most useful thing you can build is a catalogue: one entry per service, written once, approved by a partner. A filled-in entry, illustrative:
SERVICE: Annual accounts, limited company (small)
Includes:
- Preparation of statutory accounts from records you provide
- Filing of accounts with the company registry
- Preparation of the company tax return and computations
- One meeting or call to go through the accounts
Excludes:
- Bookkeeping or correcting records beyond [X] adjustments
- Audit or assurance work
- Tax planning advice, unless agreed separately
- Responding to tax authority enquiries (quoted separately)
Client responsibilities:
- Records complete and sent within [3] months of year end
- Bank statements for all business accounts
- Directors confirm accuracy before we file
Our deadlines (given timely records): draft within [6] weeks
Typical out-of-scope requests: dividend paperwork, director's
personal tax return, grant applicationsTen to fifteen of these cover most small practices. Writing them takes three to five hours, and it is worth doing even without AI: most scope disputes trace back to a service nobody defined.
Where to run it
- A shared Project in ChatGPT Business or Claude Team: upload the catalogue, your letter template and two or three approved past letters (client details removed), and write the instructions once. Both plans don't train on your content by default. Don't build a custom GPT for this; OpenAI is retiring them, and they stop running on 11 December 2026.
- Copilot in Word, if your practice runs on Microsoft 365 with Copilot licences: open the letter template and type / to point Copilot at the catalogue and a past letter in your files. Drafting in Copilot in Word from your own files shows the method.
- Your proposal tool's templates, with AI used once to draft and tighten each service description before you load it.
Whichever you choose, keep the working prompts in one place so the whole practice uses the same ones; a shared prompt library explains how.
Prompt 1: the schedule of services
Using the service catalogue in this Project, draft the schedule of
services for the client below. Include only the services listed.
For each: what's included, what's excluded, the client's
responsibilities and our deadlines, adjusted to their year end.
Copy catalogue wording; don't add services, advice or promises that
aren't in the catalogue. Flag anything you had to assume.
Client: limited company, year end 31 March, 1 director, payroll for
3 staff monthly. Services: annual accounts (small company), company
tax return, monthly payroll.Illustrative output: a clear three-part schedule. Records due by 30 June; draft accounts within six weeks of complete records; payroll processed by the 25th of each month from timesheets received by the 20th. It also added "we will advise on tax-efficient remuneration strategies" under the accounts service, and flagged one assumption: "Assumed payroll includes year-end employee forms."
Two fixes. The remuneration line had to go: it wasn't in the catalogue, and it quietly committed the practice to advice it hadn't priced. The flagged assumption was right to flag; payroll year-end forms were in the catalogue entry, so the partner confirmed it. The instruction "don't add services, advice or promises" reduces this kind of addition, but review is what catches it.
Prompt 2: exclusions that fit this client
Generic exclusions don't prevent scope creep. Specific ones do. Ask for exclusions based on what this kind of client typically asks for:
From the catalogue's "typical out-of-scope requests" and the client
details above, list 5 to 7 things this client is likely to ask us
for that are NOT included, each in one plain sentence, with "we can
quote for this separately" where the catalogue offers it.Illustrative output: "Your personal tax return as a director is not included; we can quote for it separately. Dividend vouchers and board minutes are not included. Work on grant or loan applications is not included. Corrections to payroll caused by late or changed timesheets after the 20th may be charged..."
A partner checked each against the catalogue, kept six and removed one that contradicted a fee already agreed on the proposal. The result is the section clients read when they later ask "isn't that part of what we pay you for?"
One exclusion the prompt won't think of unless you ask: prior-year figures for a client arriving from another accountant. Take a small design studio switching practices in its fourth year. Its first accounts with you start from balances you didn't prepare. Add a line to the client details ("new client; previous accountant prepared the prior-year accounts") and the draft should produce something like: "We will use the closing balances in your previous accountant's final accounts as our starting point and won't re-examine them unless you ask us to, which we can quote for." If an error in those balances surfaces in year one, that sentence is the difference between a chargeable piece of work and an argument.
Sole traders need their own version, not a trimmed company letter. For a self-employed landscape gardener, the main service is the personal tax return and business accounts, with no registry filing and no directors. The client responsibilities that matter are different too: keeping a mileage and vehicle-use record, separating personal and business spending, and telling the practice about any new income source. Put these in a sole-trader catalogue section, and the schedule drafts correctly first time instead of needing company wording stripped out.
Prompt 3: the plain-English summary
Put a short summary at the front of the letter. Clients read it; many don't read the schedules. A before-and-after, illustrative:
Before (from the schedule): "The Client shall be responsible for the provision of complete and accurate accounting records within three calendar months of the accounting reference date, failing which the Practice shall not be liable for any penalties arising from late submission."
After: "Please send us your complete records within three months of your year end (by 30 June). If they arrive later, we'll still do our best, but we can't be responsible for late-filing penalties."
The summary doesn't replace the schedule, and it should say so in one line: "This summary is for convenience; the schedules and terms below set out what we've agreed." How accountants use AI to explain tax in plain English has more prompts in the same style.
Prompt 4: the change-of-scope letter
Scope changes mid-year are where practices lose money, because the extra work starts before anyone writes it down. A filled-in example, drafted by AI from the catalogue and the original letter, for a client who asks for bookkeeping to be added in month five:
Subject: Update to our engagement, adding monthly bookkeeping
Dear [director name],
Thank you for asking us to take on your bookkeeping. This letter
updates the engagement letter dated [date] and adds one service.
From 1 September we will:
- record your sales, purchases and bank transactions monthly in
your accounting software, from documents you upload by the
10th of the following month
- reconcile your business bank accounts monthly
- send you a short monthly summary of the figures
Not included: chasing your customers for payment, or paying your
suppliers.
The monthly fee for this service is [fee from approved quote],
starting with the September invoice. All other terms of our
original letter continue unchanged.
Please confirm by signing below or replying to this email.
[Partner name]AI gets you this draft in a minute. What it can't do is know the fee, so the prompt should always pull the fee from the approved quote and leave a placeholder if none exists.
The speed matters because of what the delay costs. With illustrative figures: the bookkeeping starts in month five after a phone call, and the letter follows in month eight. At four hours a month and a $60 internal cost, that is 12 hours, or $720, of work done before anything was agreed in writing, and the first invoice for it is now a negotiation rather than a formality. A minute's draft, reviewed and sent the week the client asks, removes most of that risk.
Prompt 5: check the letter against the proposal
Before a letter goes out, have AI compare it with the accepted proposal. Mismatches here cause most early disputes:
Compare (A) the accepted proposal and (B) the draft engagement
letter. List every difference in: services, fees, payment terms,
start date, client entity name, and deadlines. Quote both versions
for each difference. Say "no differences" for any category that
matches.Illustrative output: "Entity name: A says '[Company] Limited', B says '[Company] Holdings Limited'. Fees: A shows payroll at [fee] a month for up to 5 employees; B shows [fee] a month with no employee limit. Start date: A says 1 August, B says 1 September. Services, payment terms, deadlines: no differences."
All three were real problems. The entity mismatch was the serious one: the holding company is a different client, and a letter addressed to it doesn't engage the trading company at all. The missing employee limit would have committed the practice to payroll for any number of staff at the same fee. Keep this prompt's results short and factual; its only job is to surface differences for a person to resolve.
Where AI-drafted letters go wrong
- The wrong client type. A partnership's letter drafted from the limited-company template ends up with "directors" who are actually partners, and company filing duties that don't apply. Keep a separate catalogue section and template per client type, and state the type in the first line of every prompt.
- Dates that don't follow the year end. AI assembles deadlines well from a year end it is given, and invents them when it isn't. If the prompt lacks the year end, the draft still contains confident dates. Make the year end a required field.
- Old wording coming back. Past letters in the Project are useful examples until the catalogue changes. A practice that raised its bookkeeping adjustment limit found the old limit reappearing, because a two-year-old letter in the Project still said it. Replace example letters whenever the catalogue changes.
- Personal data in the wrong place. Directors' home addresses and dates of birth belong in the practice system, not in the Project where every user can see them. Draft with placeholders and merge personal details at the end, or keep the drafting inside a tool that respects client-level permissions. In practice that means the AI only ever sees tokens such as [DIRECTOR_NAME], [REGISTERED_ADDRESS] and [YEAR_END], and Word's mail merge or the proposal tool's merge fields fill them from the client record. A quick check before sending: search the final letter for "[" to catch any placeholder that didn't fill.
Most of these trace back to the inputs rather than the model. When a draft is wrong, ask first whether the catalogue, the template or the example letters told it to be.
Keeping the catalogue current
Give the catalogue a version number and an owner, usually the partner who signs most letters. Review it once a year before the re-issue season, and whenever a service, fee basis or professional-body template changes. When you update it, update the Project in the same sitting: upload the new catalogue, delete the old one, and swap any example letters that no longer match. A catalogue in two versions is worse than one out-of-date catalogue, because the assistant may quote either.
Check the update took with two test questions in a fresh chat inside the Project, each about something you've just changed. After raising the bookkeeping adjustment limit and moving the records deadline, for example: "What is the adjustment limit in the annual accounts service?" and "When are records due for a client with a 31 December year end?" If either answer gives the old figure, an old file is still in the Project, most often an example letter. Find it and replace it before anyone drafts a real letter.
The partner's review, filled in
A review checklist for each letter, completed for the payroll client above (illustrative):
ENGAGEMENT LETTER REVIEW, [client], [date]
[x] Client entity name matches registry record and proposal
[x] Services match the accepted proposal (3 of 3)
[x] Nothing in the schedule that isn't in the catalogue
(removed: "advise on remuneration strategies")
[x] Fees copied from proposal, not calculated
[x] Deadlines match client's year end (records due 30 June)
[x] Standard terms and liability clause unchanged from template
(compared with template version 2026-03)
[x] Exclusions reviewed (6 kept, 1 removed)
[x] Signatory is a director confirmed on the proposal
Reviewed by: [partner initials] Time: 6 minutesThe line about standard terms being unchanged is the one to protect. A realistic mistake: a manager asks the assistant to "tidy up the whole letter so it reads better", and it rewrites the limitation of liability paragraph into friendlier language that no longer caps anything. Nobody notices for a year. Compare the terms against the approved template every time, or better, keep them as a locked section the AI never receives. Mentioning AI use in client contracts covers the related clause about your own AI use.
What it saves in an annual re-issue
An illustrative five-person practice re-issues around 120 engagement letters a year, most when services or fees change. Before, each took about 40 minutes: find last year's letter, update the services, check the dates, fix the formatting. That is 80 hours. With the catalogue, a shared Project and the five prompts, each letter took about 15 minutes including the partner's six-minute review: 30 hours. Building the catalogue and prompts took about 8 hours, once.
For re-issues, one extra prompt helps: paste last year's schedule and this year's services, and ask for a list of what changed and a draft of only the changed sections. That keeps unchanged wording identical, which is exactly what you want in a document clients may compare year to year.
An illustrative reply for the payroll client in its second year: "Changed: payroll now covers up to 8 employees (was 5), fee per approved quote. New: quarterly management accounts. Unchanged: annual accounts, company tax return." It looked complete and wasn't. The practice had also shortened its records deadline from three months to two in the catalogue, and the assistant reported the accounts service as unchanged because it compared service names, not deadlines. The fix was to spell out the comparison: "compare each service's inclusions, exclusions, client responsibilities, deadlines and fee basis separately". On the next re-issue the deadline change came up for every client it affected.
Engagement letter questions practices ask
Can I let clients sign an AI-drafted letter without a partner reading it?
No. The letter sets out what the practice is responsible for, so someone accountable should read every letter before it goes out, even when AI only drafted the schedules. The review gets quicker once the service catalogue is stable, often a few minutes per letter, but it should never be skipped.
Which AI tool is best for engagement letters?
The one that can hold your templates and approved examples in a shared, business-grade workspace: a Project in ChatGPT Business or Claude Team, or Copilot in Word if your templates live in Microsoft 365. The tool matters less than the inputs. A good service catalogue and clause library in any of them beats a clever prompt with nothing behind it.
Do I need to tell clients AI helped draft their letter?
Many practices add a general line about AI use to their terms rather than flagging individual documents. What matters most is that a named person is responsible for the letter's content. If your professional body has guidance on disclosure, follow it, and if a client asks directly, answer plainly.
Further reads
- How to Catch Made-Up Figures in AI-Drafted Proposals — Checking fees and figures in AI drafts before they go out.
- 7 AI Mistakes That Put an Accounting Firm's Client Trust at Risk — The AI errors that most damage client trust in a practice.
- How to Write Business Proposals Faster With AI — The proposal that comes before the letter, drafted faster.
- Claude vs ChatGPT for Business Writing, Proposals and Reports — Which assistant handles careful business writing better.
- Turning Hours Saved by AI Into Advisory Work Clients Pay For — Turning hours saved on paperwork into advisory work.
- Custom GPT vs Claude Project vs Gemini Gem: Which Should You Use? — Where to keep your letter templates and examples.
- Rolling Out AI in a Bookkeeping Practice Without Losing Control — A staged AI rollout for bookkeeping practices: control points, an inventory of AI already in your software, a pilot, sampling rates and client wording.
- What an AI Implementation Looks Like in a Small Accounting Firm — A seven-person practice followed through 12 weeks of AI implementation: time audit, tools switched on, records chasing, costs and what went wrong.
- How Small Tax Practices Use AI Through the Busy Season — A season-long plan for tax preparers: what to build eight weeks out, how AI handles intake, chasers and the inbox, and where preparer review stays in charge.
- Is It Safe for an Accountant to Use ChatGPT With Client Data? — Which ChatGPT plans an accounting practice can defend for client data, what never goes in, and the engagement-letter wording to add.
- How to Choose an Accountant Who Uses AI Well — What good AI use inside an accountancy practice looks like, eight questions for the first meeting, and how a café compared three firms.
- 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; Microsoft support page on drafting with Copilot in Word; OpenAI help article on custom GPT retirement; OpenAI and Anthropic business plan pages. Checked September 2026. Prompts, outputs and figures are illustrative.