Where the time really goes
In a small practice, the skilled work often isn't what fills the week. It's everything that has to happen before anyone can get to it: chasing clients for records, opening attachments and keying figures in, answering the same deadline question for the tenth time, and working out where a job has got to.
It gets worse around deadlines. Work queues up waiting for one missing statement. A client emails the partner, the bookkeeper and the general inbox about the same thing, and three people half-answer it. A job sits between preparer and reviewer because nobody noticed it was ready. The work is rarely hard. Keeping track of it across the team is.
What it looks like in practice
Here's a made-up example. A bookkeeping practice in December, and a client called Ade who still hasn't sent everything for their tax return.
- Ade gets a clear, specific reminder. Not a generic "please send your records", but a friendly message in your practice's wording that lists exactly what's still outstanding, based on the checklist for Ade's job, with a link to upload it.
- It reads what arrives. When Ade uploads statements and a few receipts, the assistant reads them, pulls out the key figures and dates into a review sheet, and ticks off what has arrived. Anything it couldn't read clearly is flagged, not guessed.
- It tells the right person. Once everything is in, the job moves to ready for preparation and the bookkeeper responsible is told. If something is still missing after a week, Ade gets another polite nudge.
- Questions go to a professional. When Ade replies asking whether a new van can be claimed, the assistant doesn't answer. It acknowledges the question, says who will reply, and passes it to the accountant with the context attached.
Your team still checks every figure and gives every piece of advice. They just stop spending December chasing paperwork and retyping it.
Where AI isn't the answer
Accountancy carries professional and legal responsibility, and your clients trust you with sensitive financial information. We'd keep AI away from:
- Tax and accounting advice. An assistant can draft a routine reply about deadlines or what's needed, for a team member to approve. Advice on a client's own position comes from a qualified person.
- Anti-money-laundering and ID decisions. Software can send requests and track what has come back, but client due diligence and the decisions that follow belong to your team and your AML procedures.
- Signing off accounts or returns. AI can speed up preparation. Review and sign-off stay with the person professionally responsible.
- Client data in consumer AI tools. Bank statements, payslips and tax records belong only in tools set up properly for business use, with a data processing agreement in place and your professional body's confidentiality rules in mind.
How to get started
- List the jobs your team does before the real accounting work can start, and roughly how long each one takes.
- Pick the most repetitive. For many practices it's chasing records or keying in figures from documents.
- Start there, with one assistant, built around your practice management and accounting software. Prove it saves time before going further.
If you'd like help working out which one to start with, that's exactly what our AI Health Check is for. It's from £2,250 depending on team size, and you can start with a free AI plan first.
Common questions
- Will clients get answers about their tax from an AI?
- Not if we set it up. An assistant can handle routine messages, such as confirming what's been received, reminding clients what's outstanding and explaining your process, in wording you have approved. Questions about a client's own tax or accounts go to a qualified member of your team.
- Can AI help with our AML and ID checks?
- Only with the admin around them. It can send requests, remind clients and track what has arrived. The checks themselves, and any decision about whether to take on or keep a client, stay with your team and your AML procedures.
- How accurate is AI at reading receipts and statements?
- Good on clear documents, less reliable on poor scans, handwriting and unusual layouts. That's why we build it to flag anything it isn't sure about and to put figures into a review step, not straight into the books. Your team stays the check.
- Is client financial data safe?
- We only use business-grade AI services whose terms say your data is not used to train their models, with a data processing agreement in place. We tell you which services we use and where data is processed, so you can keep your own records and privacy notice up to date.