Baevich

Accounting firm automation. What a small practice should automate first

By Andrew Baevich

Most small accounting practices get the biggest return from automating the work around the accounting, not the accounting itself. Client document collection, new enquiries and onboarding, reminders, billing and the partners' weekly numbers eat hours every week and follow clear rules. Bank feeds and reconciliation are already handled well by your accounting software. Start with the firm's admin, keep judgement work with people, and have a person approve anything that goes to a client.

This guide is for practices of roughly five to thirty people. It covers what to automate, in what order, what to leave alone, and a thirty-day plan you can start before the busy season.

Two kinds of automation, and why the order matters

There are two layers of automation in an accounting firm.

The first is ledger automation. Bank feeds, transaction categorisation, reconciliation and recurring journals. Most modern accounting platforms already do a lot of this, and your team probably uses it.

The second is firm automation. Everything that happens between the client and the work. Chasing documents, answering new enquiries, sending engagement letters, booking meetings, reminding clients about deadlines, invoicing and collecting fees, and reporting to the partners. This layer is usually still manual, spread across inboxes and spreadsheets, and it is where most of the lost hours sit.

That is why we start with the second layer. It is less glamorous, but the hours come back faster and the risk is lower.

What to automate first, in order

1. Client document collection and chasing

If you only fix one thing, fix this. Every season the same pattern repeats. A list of what each client needs to send, a first request, then rounds of reminders, then someone checking what arrived and what is still missing.

An automated flow sends each client a personalised list, tracks what has been uploaded, sends polite reminders on a schedule you set, and gives the team one view of who is complete and who is not. An AI agent can also read what arrived, check it against the list and flag gaps, for example a missing bank statement month. A person reviews the flags before anything goes back to the client.

2. New enquiries and onboarding

Prospective clients often contact several firms at once. The one that answers first, clearly and with a next step, tends to win the work.

Route every enquiry into one place. Have an AI agent sort it by service and size, draft a reply with the right next step and offer a meeting time. A partner or manager approves the reply. Once the client says yes, the same flow can prepare the engagement letter from your template, request identity documents where your rules require them and set up the client in your practice management system.

3. Deadlines and reminders

Clients forget deadlines, and the firm pays for it in rushed work. Automated reminders tied to each client's deadlines, in plain language, cut the last-minute rush. In the UK the self assessment online deadline of 31 January is the obvious one. In the US the April filing deadline plays the same role. The reminder text is written once and approved, then sent on schedule.

4. Billing and collections

Fees should be invoiced when the work is done or on a fixed schedule, not when a partner finds time. Late payment reminders should go out automatically in a tone you approve. This is one of the quickest ways to improve cash flow in a small practice, and it removes an awkward task from the partners' week.

5. The partners' weekly numbers

Pick five numbers that matter to you. New enquiries, new clients signed, work in progress, fees invoiced and cash collected is a good start. Have them collected automatically from your systems into one short weekly report. If you ever plan to sell the practice, a buyer will ask for exactly this kind of history, and our guide on what to fix when a business will not sell explains why.

6. Client updates during the busy season

When the season is at its peak, clients still want to know where their return or accounts stand. A short automated status update, drafted by an AI agent from your practice management system and approved by the manager, can save a lot of phone calls.

What to leave with people

Automation should not make professional judgements. Tax positions, advice, reviewing the final figures and anything that needs professional scepticism stay with qualified staff. The same goes for difficult client conversations, fee negotiations and anything where a mistake would be expensive.

A good rule is simple. The machine prepares, a person decides. Every message that goes to a client is approved by someone in the firm, at least until you have months of evidence that a particular message type is always right.

Also check your professional body's guidance and your data protection obligations before client data goes into any AI tool, and choose tools that do not train on your data.

A thirty-day plan before the busy season

Week one. Map how a typical client moves through the firm, from first contact to final invoice, and note who does each step. Pick the two steps that cost the most hours. For most practices that is document collection and chasing fees.

Week two. Write the messages. Request lists, reminder texts, the engagement letter template and the late payment reminders. Have a partner approve each one.

Week three. Set up the flows for those two steps with a small group of clients. Keep a person approving every outgoing message.

Week four. Review what worked, fix what did not, and roll it out to everyone. Start the weekly numbers report so you can see the effect.

This is a realistic pace for a small firm. Trying to automate everything at once before the season usually means nothing is finished when the season starts.

What this looks like in practice

In practice the pattern is simple. AI agents prepare the repeat work around the client work, a person approves everything that goes out, and partners get their time back. Law firms follow a similar order, covered in our guide on law firm automation.


Andrew Baevich

Andrew Baevich runs a small team at baevich.com that sets up AI automations for law and accounting firms. He founded and ran CyberFox, and has spent more than ten years building and running service businesses.

How this article was made. Written with AI assistance from our own notes and work, and fact-checked before publishing.