Bookkeeping automation

What gets automated, and what does not

Bank feeds arrive already matched to invoices and bills where the reference is clean. Recurring transactions are coded from history. Receipts are captured, read and coded from the supplier name. Sales from e-commerce and payment platforms are summarised and posted. That covers most of a small business's transactions in a typical month.

What is left is the interesting part: the transfer that looks like income, the director's personal spend on the business card, the supplier whose invoices vary in nature, the deposit with no reference. A tool is judged on how it surfaces these: does it hide them under a confident guess, or hand them to a bookkeeper with the reasoning shown. The exceptions queue is the product.

What to look for

Rules that belong to the firm, not the client file, so a rule built once applies across the client base. A review workflow the whole team can follow, with who approved what and when. An audit trail that a reviewer or an auditor can read. Clear confidence signals on automated coding, so the bookkeeper reviews the doubtful entries and trusts the rest. And coverage of the document types clients actually send: phone photos of receipts, supplier statements, PDFs from a portal.

Integration depth matters more here than in most categories. The tool has to read the ledger's chart of accounts, tracking categories and tax codes, and write back cleanly. A tool that posts to a suspense account for the bookkeeper to move later has automated nothing.

The margin question

Firms that price bookkeeping as a fixed monthly fee depend on this category to protect the margin. The maths is simple: a tool that automates most entries but hides the rest costs more than it saves, because the hidden errors surface at year end as rework nobody billed for. A tool that automates a little less but shows its exceptions clearly protects the fee.

Measure it. Hours per client per month before and after, and the year-end adjustments per client. Those two numbers say whether the automation is working better than any dashboard the vendor shows.

Standardising the process

The return comes from doing the same thing for every client: the same rules, the same weekly review cadence, the same escalation for exceptions the bookkeeper cannot resolve. Automation tools reward firms that standardise and punish firms that let each bookkeeper run their own method, because every personal method is a set of rules nobody else can see.

Write the process down once, build the rules in the tool once, and train to the process rather than to the tool. Tools change; the process is the firm's.

Pricing and how firms recover it

Vendors price per client organisation per month, per transaction volume, or per document processed. The directory records what is published. It does not say what a firm should charge its clients for bookkeeping, and nothing here is fee advice. The practice decision is whether the tool cost sits inside the monthly bookkeeping fee (usual) or is passed through as software (less usual, and harder to explain).

What to avoid

Judging a tool on its automation percentage. Rules that live in one bookkeeper's head. Letting exceptions age until year end. Tools that post to suspense. And automating a client whose source documents never arrive, which is a collection problem the document category solves, not an automation one.