More deadlines are coming. Your headcount is not.
AI pays off in an accountancy practice in three places: preparing workpapers, drafting client letters in your house style, and chasing the records that have not turned up. The AI does the repetitive first pass, a qualified person reviews and signs off. We build these systems for UK firms on top of the practice software you already run.
Rather build it yourself? We will show you where to start, free.
What we would build
- 01Workpaper first-pass engine
- 02Records chaser
- 03House-style drafting library
Where the week goes
The work that eats a practice is not the judgement work
Workpaper prep
Transcription and tie-out dressed up as professional work. Hours of re-keying figures from bank statements, invoices and receipts before any actual judgement happens.
Records chasing
Who owes what, by when. Vague "please send your records" emails, then chasing the chasers. From April 2026, MTD quarterly updates multiply the deadlines per client.
Client letters
Year-end letters, query lists, plain-English explainers of a tax position. Drafted from a blank page when last year’s version plus this year’s numbers would do.
Meeting write-ups
Calls that need transcribing, summarising and turning into file notes. The notes that protect the firm are exactly the ones nobody has time to write.
What we would build
What we would build for your practice
Three systems, built in payoff order. Each one keeps a qualified person on every sign-off, because nothing carrying your firm’s name should leave the building unread.
Workpaper first-pass engine
The routine ninety per cent of workpaper prep done before a reviewer opens the file.
- Extracts dates, amounts, counterparties and VAT from statements, invoices and receipts, including scans and photos.
- Proposes nominal codes from the client’s own history. Where it is unsure, it says so and adds the item to a query list rather than guessing.
- Ties totals to control accounts and flags anything that does not agree, with a plain-English note. Every figure links back to its source document.
- Drafts lead schedules in your template. The reviewer works through flags and evidence, not a haystack.
Records chaser
Chasing is a state-tracking problem, not a writing problem. So we build it as one.
- Every client has a live record of what is outstanding: statements, payroll data, approvals, signed engagement letters.
- Reminders go out on a schedule and name the specific missing item, never "your records".
- The moment something arrives, chasing stops. Nobody gets nagged for a thing they sent last week.
- Non-responders escalate to a human after a set number of attempts, so the difficult conversations happen earlier, not never.
House-style drafting library
Client correspondence drafted from your firm’s own past letters, not the open internet.
- Your templates, past letters and tone rules go in. New letters come out sounding like your practice, because they are built from it.
- Query lists turned from a reviewer’s terse bullets into a structured email a client can act on.
- Tone per client: the sole trader who texts you and the FD who wants formality do not get the same letter.
- The rule on day one: AI drafts, a person sends. The read takes seconds because you are editing, not writing.
Do it yourself
You could build this yourselves. Here is how to start.
A practice can build the first version of all three systems itself, with off-the-shelf tools and a few quiet evenings. This is where we would start.
The guides are free and they do not hold anything back. If you get partway and want it finished fast, or built properly first time, that is the other reason this page exists.
How the free route works- 01Track where the hours go for one week: workpaper prep, chasing, letters, write-ups. The biggest number is your first project.
- 02Set up a business-tier AI account that contractually excludes your data from training, and write a one-page AI policy before any client data goes near it.
- 03Build the chasing list first. One spreadsheet, one row per outstanding item per client, and a weekly reminder that names the specific missing thing instead of "your records".
- 04Draft your next client letter from your past letters: paste two or three good ones in as examples, then this year’s facts. Compare it with a blank-page draft and you will see the whole idea.
- 05Keep one rule from day one: AI drafts, a qualified person signs off. Nothing carrying the firm’s name leaves the building unread.
Straight answers
Questions we get from this industry
- Is client data safe with AI tools?
- It can be, but not by default. We use business-tier tools that contractually exclude your data from model training, put a data processing agreement in place, and keep personal data out of consumer chatbots entirely. You also get a one-page AI policy and a line for your engagement letters, so you can answer the question when a client asks it.
- Will AI-drafted letters sound generic?
- Only if they are drafted from the open internet. We build the system to draft from your firm’s own past letters, templates and tone rules, so the output sounds like your practice because it is made from your practice. A person still reads everything before it goes out.
- Which tasks should never go to AI?
- Anything that is judgement on treatment: revenue recognition, provisions, going concern, anything near a materiality threshold. The AI assembles the facts that inform those calls. The call itself is the job, and it stays with a qualified person.
- Do we need to replace our practice software?
- No. The systems we build are connective: they join your inbox, your practice management system and an AI model into one flow. The wrong move is a big-bang platform migration in the name of AI. We build on what already works.
- What does Making Tax Digital actually change?
- From 6 April 2026, sole traders and landlords with income over £50,000 must keep digital records and submit quarterly updates. That means more filing events per client with the same headcount, and the work that expands is exactly what automation handles well: collecting records, categorising transactions and chasing the clients who have not sent anything.
Tell us where the hours go
Thirty minutes, no pitch deck. We will tell you which bottleneck in your practice to automate first, and if it is not worth building, we will say so.
Or email dc@operosus.com and tell us what is eating your week.