The Arithmetic of Automation

On a Thursday morning in January, someone at an accounting practice in the Midlands opened Excel and did something unusual: they calculated what manual transaction coding actually costs.

Not the headline number. Not the software subscription or the salary line. The full cost: the time, the interruptions, the rework, the things left undone because the day runs out. They tracked it for a month. Then they sent the spreadsheet to their managing partner with a two-word subject line: "We're haemorrhaging."

The figure was larger than anyone expected. And the uncomfortable truth is this: most practices have never done the calculation at all.

The baseline cost that nobody disputes

Start with the simple version. A bookkeeper in England earns, on average, £28,593 per year. That is roughly £13 per hour before you add employer National Insurance, pension contributions, desk space, software licences, or the cost of someone more senior checking the work.

How long does it take to code a transaction manually? The industry does not publish a single authoritative figure because the answer depends on context. A Tesco purchase with a clear description codes faster than an ambiguous international wire transfer with no reference. But bookkeepers we spoke to estimate between one and two minutes per transaction for routine work, assuming the chart of accounts is familiar and the description is legible.

Call it 1.5 minutes on average. That is £0.33 per transaction in direct labour cost. Before supervision. Before corrections.

For a practice managing 30 clients averaging 300 transactions each, that is 9,000 transactions per month. At 1.5 minutes per transaction, you are looking at 225 hours of bookkeeper time just on coding. That is roughly 1.4 full-time equivalents doing nothing but assigning categories to bank feeds.

But here is what the baseline leaves out.

What the baseline leaves out

The figure above assumes the transaction codes correctly the first time, that the description is unambiguous, that the client responds immediately when clarification is needed, and that no one makes mistakes.

None of these assumptions hold reliably in practice.

Then there is the opportunity cost. Advisory work generates higher revenue per employee, higher average fees per client, and better profit margins than compliance-only work. But advisory requires time. Time that is currently spent checking whether a Screwfix receipt belongs in repairs or capital expenditure.

The uncomfortable arithmetic: if a bookkeeper spends 25 hours per week coding transactions at £13 per hour, that is £325 in labour cost producing work billed at compliance rates. The same person spending those hours on advisory work might generate £600-800 in billable value. The difference compounds weekly.

The arithmetic

Consider a realistic scenario. Not an extreme case. Just a practice with normal volume and normal problems.

That is £160 per client per month just for transaction coding and its immediate consequences. Annually, £57,864 for work that produces no strategic insight, generates no advisory fees, and creates no competitive advantage.

Against this, automated coding tools typically cost between £5 and £50 per client per month depending on transaction volume. Even the higher end is £1,500 monthly for 30 clients. The arithmetic writes itself.

But that is not the whole story.

The case for doing it manually anyway

There are legitimate reasons to code transactions by hand, and pretending otherwise would be dishonest.

For very small volumes, the overhead of learning new software outweighs the time saved. A sole trader with 50 transactions per month spends roughly 75 minutes coding them manually. The breakeven point for automation is measured in months, not weeks, once you account for setup time and the learning curve.

Some businesses have genuinely unusual transactions that pattern-matching systems struggle with. A property developer with complex joint venture accounting, or a business operating across multiple VAT jurisdictions, may find that automated tools create more work through false positives than they save through correct classifications.

And there is a subtler point: coding transactions manually forces you to look at every line. That scrutiny sometimes catches problems that an automated system would miss. An unexpected large payment. A duplicate invoice. A supplier charging VAT when they should not. These things matter.

The question is not whether automation is always right. It is whether most practices are making a conscious decision based on the actual numbers, or simply continuing with manual processes because that is what they have always done.

What this means for the profession

The shift is already happening, though not uniformly. Practices that have embraced technology-enabled workflows report using compliance automation as a catalyst to move clients onto year-round advisory relationships, turning what was once a January scramble into a continuous service model.

But the broader implication is uncomfortable: if manual transaction coding is economically indefensible at scale, then the profession's relationship with compliance work is changing faster than many practitioners realise. The work that once filled the gaps between advisory engagements is being compressed into minutes instead of hours. The firms that recognise this early have time to build advisory capabilities before the compression accelerates. The ones that do not may find themselves defending hourly rates for work that clients increasingly believe should be instantaneous.

There is a parallel in legal services. Twenty years ago, junior solicitors spent billable hours manually reviewing contracts for standard clauses. Software now does that work in seconds. The firms that adapted early built new service lines around the capacity that freed up. The ones that insisted clients should keep paying for manual review lost those clients to competitors who offered flat-fee automation.

The same dynamic is unfolding in accounting, though more quietly. The practices making the shift are not talking about it loudly because there is no competitive advantage in advertising that you have stopped doing something inefficiently. But the arithmetic remains the same for everyone.

That spreadsheet from the Midlands practice, the one with the two-word subject line? The managing partner read it on a Friday evening. By Monday morning, they had signed up for a trial of three different automation tools.

Not because they believed automation was perfect. Because they had finally calculated what manual was actually costing.