AI Bookkeeping for Accounting Firms: Pricing It as a Service Line

Price an AI-assisted bookkeeping service as a fixed monthly fee in bands of transaction volume, set from your measured review time plus the per-client tool cost and a margin, with an overage rate above the top band. Per-transaction fees alone fall to pennies once software does the coding, while your real cost sits in review and exceptions.

The change AI brings is where the work sits. Capture, matching and first-pass coding become cheap and fast. What remains is review, judgement, chasing missing records and fixing the odd transaction the software got wrong. A price that pretends the hours have vanished will undercharge; a price that still bills the old hours will lose clients to a competitor who passes on a sensible share. The way through is to price the outcome and cost it from what you measure.

Follow me on Instagram@sagnikteaches

Why hourly and per-item prices stop working

Many practices started pricing bookkeeping by the hour or by the transaction. Both models weaken when AI does the data entry.

Connect on LinkedInSagnik Bhattacharya

• Hourly. The faster you get, the less you earn from the same client, and the client sees the hours fall and expects the price to follow. You also bill for review time at a rate that now reflects checking, not typing.
• Per transaction. A rate that made sense when a person keyed each line, say a dollar, is hard to defend when software reads a receipt for pennies. The client can see the logic and starts to negotiate.
• Fixed fee, unbanded. Simple to sell but unfair when a client sends 80 transactions in one month and 600 in the next.

Banded fixed fees solve most of this. The client gets a predictable bill; you get a price tied to volume and a rule for the exceptions. The sections below build the bands from your own numbers.

Subscribe on YouTube@sagnikteaches

Step one: cost each client's tools

Start with the cheapest part to know: what the tools cost you per client per month. Two common stacks for a practice are shown with their published prices as fetched on 1 November 2026. Prices change, so check the page before you set yours.

Two cautions. First, the clients' own accounting software is separate: Xero's list prices in dollars are Early $27, Growing $59 and Established $97 a month, and QuickBooks Online's four main plans run $38, $85, $140 and $340 (it also lists a limited free plan and a $20 Lite plan), but many clients pay for these themselves, so decide whether you resell or ask them to hold the subscription. Second, the Dext page prices per client with a client-count slider and says the rate comes down as the client list grows, so build the quote for your own client count rather than relying on the headline figure.

Add up the tool stack per client. A practice using Dext Advanced ($19.20) plus AI Assist ($7) has a base of $26.20 per client per month before any accounting-platform fees. With Double Core ($10) as well, for month-end close work on a subset of clients, that subset costs $36.20. The numbers are small per client and large across a book, which is why the next step matters.

Step two: measure the review time that is left

Guessing review time is the easiest way to misprice the line. Measure it for two months before you finalise bands.

1. Pick ten clients across your range: a quiet sole trader, a busy shop, a client with a messy history.
2. Time each month's review from the moment the data is captured to the moment the books are signed off. Record two numbers: review minutes and exception minutes (fixing wrongly coded items, chasing missing documents).
3. Record the transaction count for the month.
4. Calculate minutes per 100 transactions for each client, then take the median.

Here is an illustrative log for five of the ten clients. The figures are invented.

Notice that minutes per 100 transactions does not fall steadily with volume; it jumps where the paperwork is messy. That is why bands based on volume alone can undercharge messy clients, and why it is worth adding a second variable such as the number of bank accounts or the supplier count. Our guide to chasing missing client records with AI shows how to cut the exception minutes, which is where the money leaks.

Step three: build the bands

With a median of about 30 minutes per 100 transactions and a cost rate for your reviewer, the base fee follows. Say a reviewer costs the practice $45 an hour, fully loaded, which is $0.75 a minute.

Price each band at its top volume, so a client at the upper edge still pays their way. For the 0 to 150 transaction band, 150 transactions at about 30 minutes per 100 is 45 minutes of review: 45 × $0.75 = $33.75 of labour, plus a tool cost of $26.20, plus a share of overhead. Suppose you add 40 per cent for overhead and risk to labour only, and a margin target of 35 per cent on the total. The arithmetic: labour $33.75 × 1.4 = $47.25; add tools $26.20 = $73.45; divide by (1 − 0.35) = $113, which you round to $115 a month. That figure covers bookkeeping review only, not tax returns, payroll or year-end work, which you price separately.

Build four bands the same way and compare with what clients pay now. The values below are illustrative.

The overage rate matters more than it looks. It stops one client's busy month from eating the margin on ten others. Write it into the engagement letter; the guide to AI engagement letters for accountants covers how to state scope clearly.

Fixed monthly fee versus per transaction: a direct comparison

Take one client, a landscaping company with 220 transactions in a typical month and 360 in the busy spring months. Compare three models over twelve months, assuming six busy months and six typical ones. All numbers are made up.

The banded model earns the most because it matches price to the work. The per-transaction model undercharges because $0.60 doesn't reflect the review time, which does not fall as fast as the cost of coding. The single fixed fee looks neat but is the one that quietly erodes margin during busy months. If clients push back on bands, a smoothed alternative is to average the year and charge the same amount each month, with an annual true-up if volume lands outside the agreed range.

What to say to clients who expect a discount

Some clients know AI exists and expect cheaper bookkeeping. Prepare a plain answer that is true. Something like: "Software now reads and codes your transactions, so the typing is gone. What you are paying for is the accountant checking the results, fixing what the software gets wrong and standing behind the books." Then show them what they now get that they did not before, which might be a monthly summary, a quicker close, or fewer questions about missing receipts.

Consider a client who runs a yoga studio, with 130 transactions a month and a stable pattern. Previously billed $140 on an hourly basis, now $115 banded. The practice shares some of the efficiency with a lower price, adds a one-page monthly report, and keeps the client. That is a considered trade: a lower fee, a better service and a client who stays. Set the $25 a month you gave up against what it would cost to win a replacement client. Our overview of what AI costs a small accounting firm helps you set the rest of the stack.

Protecting margin: the exceptions that cost most

Four patterns tend to cause the overruns, each with a pricing remedy.

• Late or bulk uploads. Remedy: a cut-off date written into the engagement letter, and a fee for catching up after it.
• Cash and unstructured receipts. Remedy: a per-batch fee for photo-and-paper clean-up, or an instruction to use the receipt app. See AI receipt capture and bank categorisation for setting that up.
• Multiple bank accounts and cards. Remedy: price per feed, for example $15 per extra account a month.
• Historic mess. Remedy: a separate one-off clean-up project, quoted after a sample, so the monthly fee starts from a clean base.

An illustrative case: a print shop with messy supplier invoices shows 90 exception minutes a month in the log. At $0.75 a minute that is $67.50 of unrecovered cost, every month. Adding a $60 exceptions line, or moving them to a cleaner capture routine, recovers it. You find this only because you measured.

A five-client portfolio check

Before you roll the model out, run it against five real clients and compare with what they pay now. Here is an illustrative result.

If the model raises a client's price by a large margin, phase it in. If it lowers a price, check you have not underpriced a messy client. A portfolio where roughly half the clients go up a little and the rest go down a little suggests the bands sit about right.

Capacity: how many clients one reviewer can carry

Pricing and capacity are the same problem seen from two sides. Once you know the review minutes per band, you know how many clients a person can handle, and whether the line can grow without hiring.

Say one reviewer has 100 productive hours a month for bookkeeping review, after meetings, training and admin. That is 6,000 minutes. Using the band assumptions above (45, 90, 150 and 240 minutes), a book of 20 band A clients uses 900 minutes, 15 band B clients use 1,350, 8 band C clients use 1,200 and 3 band D clients use 720. The total is 4,170 minutes, leaving 1,830 spare, which is room for about 20 more band B clients (at 90 minutes each) or 40 more band A clients (at 45 each). At the fees in the illustration, that book brings in 20 × $115 + 15 × $185 + 8 × $285 + 3 × $430 = $2,300 + $2,775 + $2,280 + $1,290 = $8,645 a month. The figures are invented, but the exercise tells you something real: whether your next hire should come at 46 clients or at 66.

Raising prices on existing clients without losing them

Moving old clients onto the new model is the awkward part. A short, plain note works better than a long justification. The wording below is a sample to adapt; fill the brackets with your own details.

Subject: Changes to your monthly bookkeeping fee from [month]

Hi [first name],

From [month] we are moving bookkeeping to a fixed monthly fee based on how many transactions go through your accounts. Your usual volume of about [number] a month puts you in our [band] band, at [fee] a month.

What you get: the same sign-off by a named person each month, [a one-page summary / faster close], and no surprise bills in a busy month unless you go well past the band.

If your volume changes, we will tell you before the next invoice and agree any change first.

Happy to talk it through on a call.

[your name]

Send it with at least a month's notice, and offer a call to anyone whose fee changes by more than about fifteen per cent. Clients who rise sharply are the most likely to leave, so phase their move over two or three invoices, or pair the increase with something visible such as a monthly summary.

A month-end timeline that supports the fee

A fixed fee only works if the month runs to a rhythm. A practical timeline, with invented dates, looks like this.

• Day 1 to 3: bank feeds and receipts captured automatically; the tool flags missing documents and requests them from the client.
• Day 4 to 6: first-pass coding is done by the software; reviewer checks flagged items and a random sample of ten per cent.
• Day 7: cut-off for client uploads. Anything arriving later rolls into next month or incurs the catch-up fee.
• Day 8 to 9: reconciliation, accruals and a short variance check against last month.
• Day 10: sign-off and a summary to the client.

This rhythm is what you are charging for, and it is easy to describe on a sales call. It also tells you where to add measures: the share of clients who upload before the cut-off, the percentage of items corrected in review, and the days from month-end to sign-off. If the correction rate in your sample review exceeds, say, one in twenty for a client, the software's rules need adjusting for that client before you blame the model.

A rule-tuning example

Take the tutoring centre in the log. Review shows that 12 of 140 items in a month were put in "Office supplies" when they were for teaching materials, a cost that belongs elsewhere. That is 8.6 per cent wrong, above the one-in-twenty line. The fix is a supplier rule: one supplier maps to one category. After the rule is added, next month's sample shows two wrong items out of 150, about 1.3 per cent, and review minutes fall from 25 to 18. A seven-minute saving a month on one client does not sound much, but across 40 clients it is nearly five hours a month, which is the kind of gain that makes the fee model hold.

How to check the service line is paying

After three months, review the line as a business. Calculate revenue per client, tool cost per client, review hours per client, and gross margin. Compare the margin by band; if band A has a 20 per cent margin and band C has 40, the bands are mispriced. Track the share of clients going over band, because it tells you whether the thresholds fit. Review the tool bill against the published price each quarter, since a per-client tool cost that creeps up with a client count can pass unnoticed. For the wider rollout picture, including controls and staff, see rolling out AI in a bookkeeping practice.

Pricing questions firms ask once the tools are live

Should AI savings be passed on to bookkeeping clients?

Partly, and mostly through scope rather than price cuts. Keep the fee tied to the outcome (books closed and reviewed by a set date) and the transaction band, not to your hours. If you cut fees as you get faster, you pay for the tools and the review and keep nothing for the risk you carry.

Is per-transaction pricing a good idea with AI?

Rarely as the only model. Software reads and codes most transactions cheaply, so a per-item fee falls to cents, while your costs sit in review and exceptions. Use bands of transactions inside a fixed monthly fee, with a clear overage rate above the top band.

How do I recover the software cost?

Either build it into the fee or itemise it. Tools such as Dext and Double bill per client per month, so a simple rule is to add the tool cost per client, plus a margin, to the base fee. Itemising is transparent but invites clients to compare the line against the software's website.

How many hours of review should I budget per client?

Measure it. A practical starting point is to time the review for a sample of ten clients over two months, then set the budget at the median plus a margin. Cleaner clients need less; clients who upload late or in bulk need more. Price the second group differently.

When should I stop offering a client the fixed fee?

When the client's volume or mess consistently exceeds the band you agreed, for two or three months running. Offer a move to the next band, or a clean-up project, before you absorb the extra work for free.

Further reads

• Best AI Accounting Agents: QuickBooks, Xero JAX and Sage Copilot — The accounting platforms' own AI, compared.
• How to Automate Client Onboarding in a Small Accounting Practice — Onboard new clients onto the fee bands cleanly.
• Is AI Worth It for a One-Person Bookkeeping Business? — The same sums for a one-person practice.
• Best AI Tools for Small Accounting Firms, Sorted by Task (2026) — Tools sorted by task for small firms.
• AI Readiness Checklist for Accountants, Solicitors, Consultants — A readiness check before you launch the line.
• What an AI Implementation Looks Like in a Small Accounting Firm — What an implementation looks like in a small firm.
• Accountant Offering AI-Enabled Automation: What to Ask Before Hiring — Questions, sample answers, a fee comparison and a scoring sheet for judging an accountant's AI-enabled automation pitch before you sign.
• Bookkeeping AI Reviews: How to Test a Tool on Your Own Client Files — A step-by-step test for any AI bookkeeping tool: consent, a 150-transaction set, an accuracy tally and an exception log you can compare across vendors.
• Dext Pricing for Accountants: Essentials, Advanced and AI Assist — Dext practice pricing per client, how its five-client blocks work, and a payback grid for the $7 AI Assist add-on.
• Double (Formerly Keeper): AI Bookkeeping Features and Cost per Client — What Double charges per client on Core, Plus and Scale, which AI features each tier adds, and a 40-client monthly bill worked through line by line.
• How to Price Accounting Services: Quote Fixed Fees With AI Support — A fixed monthly fee built from a scope list, an hours-times-rate check, three tiers and an AI-assisted quote worksheet, with every sum shown.
• Intuit Accountant Suite Pricing: Is Accelerate Worth It for a Small Firm? — Core is free, Accelerate is $149 a month from 20 January 2027. Break-even sums, a four-person worked example and a plan for the free beta.
• Virtual CFO Pricing: Build a Monthly Package With AI-Prepared Reports — How a firm can price a monthly virtual CFO package: three tiers, the reports in each, review time, tool costs and add-on rules.
• Catch-Up Bookkeeping With AI: Rescue Months of Untouched Books — A bookkeeper's month-by-month recovery method, with exception lists, client queries, reconciliation checks and a scoped clean-up quote.
• AI in Accounting Examples: Six Worked Jobs for a Small Practice — Six jobs a small accounting practice can give to AI today, each with a worked example, a sample output and the check a person still makes.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Dext partner pricing page and Double pricing page (fetched 1 November 2026); Xero and QuickBooks Online pricing pages (USD list prices checked 1 November 2026). Tool prices change; confirm on the vendor's page.