AI for accountancy practices

The problem

AI is very good at reading and drafting. It turns a messy pile of papers into a tidy first draft, and it notices what is missing. It is poor at judgement, and it will quote last year's tax rates with complete confidence.

So the useful work in a practice is the reading and drafting around each job: the chase, the summary, the query list, the first look at a set of accounts. The decision stays with someone accountable for it.

Everything we build for practices follows one rule. The AI drafts, and a person in the firm approves before anything reaches a client or HMRC.

What we build

Client email drafts

Records chases, deadline reminders and replies to routine questions, written from the client's own record and waiting in an approval queue.

HMRC and Companies House letters

What each letter says, what it asks for, by when, and which client it belongs to, filed against the client.

Query lists from a pile of papers

Bank statements, receipts and last year's file in; a list of gaps and questions out, ready for the senior to check.

Accounts pre-review

This year set against last, with every movement above your threshold and every unusual balance listed, so the reviewer starts with the questions.

A staff assistant that cites its source

Questions about procedures and tax answered only from the firm's own manual and a source you maintain, with the page it came from shown.

Work your clients need

Each client's figures read for the tax, cash, funding and protection conversations they suggest, each with the figure that raised it. This is what the Opportunity Engine does inside Accountin.

Proof

Six agents, every one drafting for approval

Buzz OS, the system Andy is building for their own firm, has six AI agents that chase information, nudge new clients, triage the inbox and pre-review jobs. All six are in testing, and every one puts its work in an approval queue.

os.buzzaccounting.co.uk — job planner

Six agents, every one drafting for approval

Case study

The practice operating system

Twenty-plus modules, six AI agents and full HMRC Making Tax Digital filing, built to replace the separate tools an accountancy firm usually runs on. In testing ahead of ISO 27001 certification.

Read the case study →

Straight answers

Questions about ai for accountancy practices.

Will AI send anything to my clients?

Not unless you decide it should. Every message waits in an approval queue, and a person presses send.

Where does our client data go?

It is processed in the UK or EU, it is not used to train anyone's model, and the data processing position is put in writing before anything is built.

Can it answer tax questions?

Only from a source the firm maintains, and it shows where each answer came from. A model answering from memory quotes out-of-date rates, so we never build it that way.

Where should a small practice start?

With the dullest job the team repeats most often, where a mistake is cheap and easy to spot. Records chasing is the usual first choice.

Tell us what's eating your week.

Tell us the job that keeps not getting done. We'll tell you if software can take it off you, and what it costs.

AI for accountancy practices that drafts for a person to approve