What to Charge for AI-Enabled Accounting Services

Most firms can build the automation long before they can answer the only question that decides whether it pays: what do we charge for this now?

What should an accounting firm charge for AI-enabled services?

An AI-enabled bookkeeping package typically prices at $900 to $1,600 a month, against $450 to $900 for compliance-only work. A full AI controller service runs $2,000 to $3,800. The price moves because the deliverable changes from a monthly close to a monthly decision, not because the software is expensive.

That is the whole pricing problem in one paragraph, and almost every firm gets it backwards.

The mistake: pricing the tool instead of the outcome

When a firm automates categorization and reconciliation, the instinct is to pass the saving on. Keep the $650 fee, deliver it with less labor, book the margin. It feels safe and it is the worst available option, because you have just improved the profitability of a service whose price is falling for structural reasons. Software vendors are coming for that fee directly. So is every competitor who bought the same tools.

The firms that come out ahead do something less obvious. They use the freed capacity to change what the client receives, and then they charge for the new thing.

The client does not care that a reconciliation agent runs at 2am. They care that they now get a cash position on the first of the month instead of a tax return in April.

The four tiers, and what each is actually for

This is the ladder that works in practice. The prices are bands, not quotes, and the margins assume you have actually automated delivery rather than just bought a license.

Illustrative model based on the assumptions shown. Not a guarantee of results. Individual firm results vary.

Two things in that table are worth pausing on.

Margin peaks at Tier 3, not Tier 4. Fractional CFO work sounds like the prize, but it pulls partner hours back into delivery, and partner hours are the most expensive thing you own. Tier 3 is where automation carries the most weight relative to human judgment. That is where the business is.

Tier 1 is not a failure state. Some clients should stay there permanently. The error is spending your automation budget trying to make Tier 1 profitable instead of using it to move people to Tier 2.

Those are the prices you charge. For what the automation behind them costs to build and run, see AI implementation cost for accounting firms.

Price the jump, not the increment

A firm moving a client from $650 to $1,150 often tries to justify it line by line. Do not. The client will price-check each line and you will lose.

Present it as a different service with a different name, a different deliverable, and a different meeting. The old package was a monthly close. The new one is a monthly review where someone tells them what the numbers mean and what to do next. That is a defensible 77 percent price increase. "The same thing but with AI" is not.

Practically, that means the review call is not optional. It is the part the client can see, and it is why they accept the price.

What the margin actually does

Here is the same client under three treatments. This is the arithmetic behind every claim above.

Assumes a $42 fully loaded delivery hour. Illustrative model based on the assumptions shown. Not a guarantee of results. Individual firm results vary.

Note that delivery hours go up in the third column, from 3.1 to 4.4. That is the review call and the analysis behind it. You are deliberately spending some of the reclaimed time back on the client, because that spend is what justifies the fee. Automation alone buys 23 points of margin. Repackaging buys 77 percent more revenue and another 5 points on top.

What happens to revenue during the switch

The tier table shows where you land. It does not show the quarter in between, which is where firms lose their nerve. Here is the same client, the same work, at three points in the transition.

Derived from the delivery hours and the $42 loaded rate published above, valued at the $165 realized advisory rate. Illustrative model based on the assumptions shown. Not a guarantee of results. Individual firm results vary.

Read the middle column carefully, because it is the one nobody plans for. Automate the work and change nothing else, and the fee falls from $650 to $269. Not because you did worse work, but because you billed fewer hours for it. Realization does not move at all. It cannot, because the hourly model ties your revenue to your inefficiency, and you just removed some.

That column is not hypothetical. It is the default outcome of automating without renegotiating, and a firm that spends four months building before it opens a single pricing conversation will live in it.

The third column is the whole point. Same client, fewer hours than you started with, and the fee is up 77 percent, because you are no longer selling the hours. Realization above 100 percent looks wrong to anyone raised on timesheets. It is what a fixed fee does once delivery is cheaper than the standard rate assumes, and it is the only column where the automation pays for itself.

The sequencing follows from that. Renegotiate before you finish building, not after. The conversation is easier while the client can still see the old service, and impossible once they have quietly enjoyed the faster one for two quarters.

The two failure cases

The client who will not move off hourly. Some will not, particularly ones with procurement or an audit committee. You have three options and only one is bad. Keep them hourly and keep the freed hours as margin on other work, which is fine. Move them to a fixed fee at a discount to their trailing twelve month spend, which buys the transition. Or automate their work and keep billing the old hours, which is the bad one, and it is fraud rather than pricing.

The engagement where the agent gets good faster than the contract can change. This is the more common failure and it is a happy problem read wrongly. Delivery drops from 7.5 hours to 3.1 in six weeks while the fee is locked for eleven months. Margin looks excellent, and next year's renewal arrives with the client knowing exactly how little time you now spend. Fix it in the contract, not in the renewal: agree the fixed fee against the deliverable and the response time, never against an hour count, and say so in writing at the start.

A note on scope, because these two pages answer different questions. This page prices what your firm charges its clients. What the automation itself costs to build and run is a separate number, and we publish those bands in AI implementation cost for accounting firms.

The migration math on a real client list

A twelve person firm with 200 clients does not need a growth strategy to make this work. It needs a migration target.

Move 25 percent of Tier 1 clients up to Tier 2, and 15 percent of Tier 2 up to Tier 3, inside twelve months. On a list weighted toward compliance work, that is roughly 30 to 40 upgrade conversations, not 200. Each one is with someone who already trusts you and already pays you.

This is why the model works for small firms. You are not competing for new logos against firms with bigger marketing budgets. You are having a different conversation with a client who already answers your calls.

Three ways firms get this wrong

Discounting the first cohort to prove the concept. The early clients become your reference price. If the first five pay $800 for what you intend to sell at $1,400, you have set the market rate for your own practice.

Announcing the price change to everyone at once. Migrate in cohorts. Start with clients whose books are cleanest, because those engagements will hit target margin fastest and give you the confidence to hold the price with harder ones.

Selling the automation. Clients do not buy agents. They buy knowing whether they can make payroll in March. Lead with the second thing.

Before you set a price

Work out your actual loaded delivery cost per hour first. Most firms guess, and the guess is usually 20 to 30 percent low because it omits software, supervision, review time and the hours nobody logs. Every margin figure above is meaningless if that input is wrong.

If you want the delivery side built rather than assembled from tools, that is what we do for accounting firms, and the free assessment is a scoping conversation, not a demo.