AI automation for accountants and bookkeepers

Most accountancy practices do not have a software problem. Xero and QuickBooks do their job. The problem is everything around them: chasing clients for records, tracking who has sent what, and the fact that January arrives every year and every year it hurts.

By Rashid, founder of Layerform System Ltd, based in London.

The work that is not accounting

Ask anyone in a practice where their week goes and the answers are rarely about accounts.

• Chasing clients. For records, for signatures, for the one bank statement missing from an otherwise complete set.
• Tracking who has sent what. Usually in a spreadsheet, usually maintained by one person, usually slightly out of date.
• Handling what arrives. Photographed receipts, PDFs, a folder of scans, an email saying "see attached", all needing sorting before any work can start.
• Onboarding a new client. The same sequence of engagement letter, ID checks, software access and questions, done by hand every time.
• Knowing where every job stands. Which is hard to answer quickly when the answer lives across a practice management tool, an inbox and someone's memory.
• Answering "is my return done yet?" Repeatedly, because clients have no way to see for themselves.

What is genuinely worth automating

• The chase. Every client automatically followed up on a schedule, naming the specific records still outstanding, until they arrive. This alone changes what January looks like.
• Reading what clients send. Working out what a document is, which client and period it belongs to, and filing it. AI earns its place here, because client records arrive in every format imaginable.
• Completeness checks. Comparing what has arrived against what the job needs, and listing what is missing, without a person opening the folder to look.
• Onboarding. One sequence that runs itself: engagement letter out, checks tracked, software access set up, questions asked in order.
• A live job board for the practice. Every job, its stage, what it is waiting on, and who owns it, updated by the work itself rather than by someone maintaining a spreadsheet.
• Client-facing status. So clients can see where their return is without emailing to ask.

Note that only one item on that list really needs AI. The rest is ordinary automation, which is cheaper and more reliable. Anyone selling you AI for the whole list is overcharging you. There is more on that distinction in the guide to business process automation.

What stays with you

The professional judgement, the review, and the final numbers. Nothing goes to a client or to HMRC without a person approving it. The value here is in getting the admin out of the way so there is more time for the work that actually needs a qualified person, not in letting software make accounting decisions.

Start with January

If you only do one thing, automate the chase. Self assessment season concentrates a year's worth of chasing into a few weeks, so it is where the manual effort is most visible and the return is easiest to see. Get the follow-ups running on a schedule with the specific outstanding items named, and the tracking spreadsheet stops needing a human.

It is also a good first project because it is small. You will know within one season whether it did what I said it would.

To be straight about it: I have not yet built a system for an accountancy practice. What I have built repeatedly is the same underlying problem, a business running on spreadsheets, chasing and memory, turned into something their team relies on. If you would rather work with someone who has done ten practices, that is a fair thing to want, and I would rather you knew where I stand.