Bookkeeping Automation Cost Per Client UK: Subscription or Build

·

Bookkeeping automation cost per client in the UK splits into two different bills: per-client software, published at roughly £3 to £24 per client per month, and a built workflow, priced per workflow at £2,000 to £5,000 with £100 to £600 a month to run. The first rises every time you win a client; the second does not, which is why the decision is arithmetic rather than preference.

That framing matters because almost everything published on this question prices a different purchase. Search it and you get guides to what a business pays a bookkeeper, or listicles of tools. What a practice spends automating the work it already does for clients is barely covered.

What does bookkeeping automation cost per client in the UK?

There are three prices in this category and they are not alternatives to each other.

Per-client software is a subscription priced by how many clients you carry. Document capture is the clearest example. One UK comparison puts Dext at from £2.50 per client per month at the small end and £16.50 to £17.50 per client at a ten-client minimum, with Sage AutoEntry priced per document from £15.45 (Compare the Cloud, checked 17 November 2026). A competing vendor publishes Dext practice pricing at £391 a month for 50 clients excluding VAT, roughly £7.80 per client, rising to about £894 at 150 clients (Receiptflow, checked 17 November 2026).

Whole-stack software is what all of it adds up to. One UK buyer's guide estimates a mid-sized practice running practice management, a ledger and document capture at around £40 to £80 per client per month before staff cost (Accounts-OS, checked 17 November 2026) — an estimate rather than a survey, and the number worth testing against your own invoices.

A built workflow is priced per sequence, not per client. At Ihsan Ops a single well-scoped workflow is £2,000 to £5,000 over two to four weeks, with £100 to £600 per workflow per month to run at production volume. Practice Autopilot, a custom CRM and workflow build, starts at £4,000. Which workflow to build first is what the AI Opportunity Audit settles, from £1,000.

Why every per-client figure you can find was published by a vendor

The per-client prices above come from a comparison site, from a vendor selling a flat-fee alternative to the product it is pricing, and from a buyer's guide run by a software consultancy. Several disagree with each other by a factor of five for the same product. The capture vendors' own UK plan builders were not reachable when this post was written, so none of these figures could be confirmed at source. That is the honest state of the public data.

The per-invoice benchmarks that dominate the wider results are worse for a UK reader: US accounts-payable vendors quoting dollars, typically $15 to $16 to process an invoice manually against $3 to $5 automated (Centime, checked 17 November 2026). Useful as a shape, useless as a budget.

So treat all of it as directional, and get the one number that is not: your own invoice. Add up every per-client line you pay this month and divide by your client count. That figure is what a build has to beat.

What you are actually automating in a bookkeeping practice

Four units of work, and only two of them are usually worth a build.

  1. Records in. Receipts, bank statements, sales invoices. Capture is already solved by software; the chase is not, and the chase is where the hours go. Vendor estimates put document chasing at three to six hours a week for a practice with 50 to 100 clients, rising in self assessment season — marketing figures with no disclosed method, which is why you count your own for a fortnight instead.
  2. Queries out. The unidentified transaction, the missing invoice, the personal expense in the business account. Each one is a round trip, and round trips multiply by client count rather than by complexity.
  3. Coding exceptions. The entries the rules engine will not take. This needs a bookkeeper, every time.
  4. The sign-off. Someone qualified says the ledger is right. Never automate the opinion; automate the asking.

Items one and two are sequences: same shape, dozens of times a month, no judgement in the middle. Items three and four are judgement, and a practice that automates those has not removed work — it has added a step that fails quietly and then does the work by hand anyway.

There is also a cheaper move than any build, and it costs an afternoon: list every per-client subscription you pay for and check which duplicates something already inside your ledger or practice management licence. Cancelling a duplicate saves money today at no build cost.

Subscription per client or one built workflow: the two bills compared

The honest comparison is not "rip out the capture tool and build one". Optical capture, bank feeds and ledger integrations are commodity software, cheaper to rent than to rebuild. The per-client bill worth questioning is the next subscription — the chasing tool, the query tracker, the onboarding portal, each sold per client per month to solve one sequence.

What you are comparingAnother per-client subscriptionOne built workflow
Billing shapePer client per month, for as long as you use itOne-off build, then a flat monthly run cost
Published UK figureRoughly £3–£24 per client per month£2,000–£5,000 per workflow, £100–£600 per month to run
When you win 20 clientsThe bill rises with themUsually unchanged; run cost may rise at volume
When you lose a clientThe bill fallsThe build cost is already spent
What one unit buysThe vendor's whole feature set, for every clientOne defined sequence, nothing else
Who fixes itThe vendor, on the vendor's roadmapYou, or the supplier who built it
Where it runsThe vendor's platformYour stack, around the software you keep
What you hold at the endA subscriptionThe workflow

Those columns are not the same unit, and a supplier who pretends otherwise is selling. A subscription does many things for every client; a build removes one sequence. Practices do not choose between them at the level of the firm — they choose sequence by sequence.

The break-even, and how to work out your own

A chart comparing annual cost against client count. The per-client subscription line rises steadily from zero, reaching about fourteen thousand four hundred pounds a year at a hundred and twenty clients, at ten pounds per client per month. The built workflow line is flat at five thousand four hundred pounds for year one, being a three thousand pound build plus two hundred pounds a month to run. The two lines cross at about forty-five clients, which is the year-one break-even point. Below forty-five clients the subscription is cheaper; above it the build is.Annual cost against client count. Subscription at £10 per client per month, build at £3,000 plus £200 a month.£0£4k£8k£12k£16k0306090120clients on monthly bookkeepingbreak-even, about 45 clientsper-client subscriptionone built workflow, year oneBelow the crossing, renting is cheaper. Above it, the gap widens every year.

Four steps, none of which needs a vendor's statistics.

  1. Count the sequence for a fortnight. Not an estimate. How many times did the chase run, how many messages did each client take, how many minutes went into it.
  2. Price the hour honestly. If the work is done by an administrator, the right figure is loaded cost, not charge-out. From 1 April 2026 the National Living Wage for workers aged 21 and over is £12.71 an hour (GOV.UK minimum wage rates, figure confirmed 17 November 2026), and employer National Insurance runs at 15% above a £5,000 secondary threshold for 2026/27, against which eligible employers can set up to £10,500 of Employment Allowance (GOV.UK rates and thresholds for employers 2026 to 2027, figure confirmed 17 November 2026). So a part-time administrator at the statutory minimum costs roughly £12.71 to £13.90 an hour in wage and NI alone. If a qualified bookkeeper is doing the chasing, the number is far higher: one 2025 survey of UK bookkeepers reported a median hourly rate of £33.36 (VJM Global, checked 17 November 2026).
  3. Divide. Build fee plus twelve months of running cost, against twelve months of the hours you counted at the rate you just set. Then cut the hours saved, because the exceptions stay with you.
  4. Compare against the subscription, not against zero. The question is never "is £3,000 a lot". It is "is £3,000 plus £200 a month cheaper than £10 a client a month, at the client count I expect in two years".
ClientsAnother subscription at £10 per client per monthOne build: £3,000 plus £200 a month
20 clients£2,400 a year£5,400 in year one, £2,400 a year after
45 clients£5,400 a year£5,400 in year one, £2,400 a year after
100 clients£12,000 a year£5,400 in year one, £2,400 a year after
Three years at 100 clients£36,000£10,200

Those are sums on the figures above, not quotes. Change the per-client price to your own and the crossing point moves, which is the entire point of doing it on paper first.

What Making Tax Digital does to the per-client maths

It multiplies coordination events rather than complexity. Making Tax Digital for Income Tax began on 6 April 2026 for qualifying income above £50,000, with the threshold falling to £30,000 from April 2027 and £20,000 from April 2028 (GOV.UK, get ready for Making Tax Digital, confirmed 17 November 2026; the same three steps are set out by the Low Incomes Tax Reform Group, checked 17 November 2026).

For a practice the consequence is structural. An affected client moves from one annual gathering of records to several, so every cost that scales per client per period scales with it — the subscription line included. That is why the crossing point above moves left for most practices over the next two years rather than right; the per-client MTD arithmetic is worked through separately. None of this is tax advice: the thresholds are HMRC's, and your own adviser confirms which clients are in scope and when.

When another subscription is the right answer

Four cases, stated because a supplier who cannot name them has not looked.

  • Below the crossing point. With 15 clients and a ten-minute chase, a per-client tool at £10 is cheaper than any build, and will be for years.
  • The work is genuinely commodity. Optical capture, bank feeds, open banking connections — rent these. Rebuilding them spends £4,000 reproducing £8 a month.
  • The process is about to change. Mid-migration between ledgers, or a client mix turning over, means anything built now is built on a moving target.
  • Nobody has measured it. An unmeasured process automated is a guess made faster.

The warning in the other direction: a build only saves money if the hours it frees are either sold to clients or not bought in the first place. Hours that simply become less pressure are a real improvement and a worse investment — decide which you are buying before a supplier decides for you.

Frequently asked questions

How much does bookkeeping automation cost per client in the UK? Per-client software for document capture and chasing is published at roughly £3 to £24 per client per month, depending on tier, volume and which vendor published the figure. A built workflow is priced per workflow instead: £2,000 to £5,000 at Ihsan Ops, plus £100 to £600 a month to run, spread across every client you have.

Is it cheaper to add another per-client subscription or to build a workflow? It depends entirely on client count. A per-client subscription costs more every time you win work; a build is a fixed sum that does not care how many clients you have. At £10 per client per month, a £3,000 build with £200 a month to run breaks even at about 45 clients in year one, and about 28 clients across three years.

Should a bookkeeping practice replace Dext or AutoEntry with a custom build? Usually not. Optical capture, bank feeds and ledger integrations are commodity software that is cheaper to rent than to rebuild, and replacing them adds risk with no upside. The per-client bill worth questioning is the next subscription — the chasing tool, the query tracker, the onboarding portal — not the capture tool you already rely on.

What does Making Tax Digital for Income Tax change about the per-client maths? It multiplies coordination events rather than complexity. Quarterly updates began on 6 April 2026 for qualifying income above £50,000, with £30,000 from April 2027 and £20,000 from April 2028. Each affected client moves from one annual cycle to several, so any cost that scales per client per period scales with it.

When is bookkeeping automation the wrong spend for a practice? When the records arrive in a different shape every month, when the client list is small enough that the chase is a ten-minute job, when the underlying process is about to change, or when nobody has measured where the hours actually go. Automating an unmeasured process buys you a faster version of a guess.

Where to start

Count one sequence for a fortnight, add up your per-client lines, and put both numbers against the table above. If the build still looks like the answer, the next step is scoping which sequence to build first — that is the AI Opportunity Audit, from £1,000, and it is also where the answer sometimes turns out to be "cancel two subscriptions and change one process". If deadlines rather than records are the bottleneck, Deadline Autopilot is the packaged route at £1,800 setup plus £50 a month. If several sequences are leaking at once, it is a process automation engagement.

Book a 30-minute discovery call at cal.com/ihsanopsai/30min and bring your subscription list.