The tool can tell you what work like this has actually cost you. It cannot tell you what to charge, and no professional body will tell you either.
Key takeaways
AI helps with the part of fixed-fee pricing that is arithmetic and pattern-matching — what engagements of this shape have actually consumed in your firm, where they overran, which clients generated the most unbilled follow-up — and it is useless for the part that is judgment, which is what you are willing to be paid for carrying that risk. Practices that get value out of it feed it their own history. Practices that ask a general model what to charge for a compilation get a plausible number produced by a system that has never seen their file.
The honest framing is that a fixed fee is a price for a defined scope plus a margin for variance. The tool improves your estimate of the variance. You still choose the margin.
This surprises newly independent practitioners. Professional regulators do not set fees. CPA Alberta says so directly on its fee mediation page: the Chartered Professional Accountants Act, the bylaws and the Rules of Professional Conduct do not allow it to decide how much a member may charge for services, and what the regulator offers instead is a mediation process for resolving disagreements about fees after the fact. The rules that do bite are about how you conduct yourself and what you disclosed, not about the rate card.
For a law firm in Ontario the framing is different but the conclusion is similar. Rule 3.6-1 of the Law Society of Ontario’s Rules of Professional Conduct requires that a fee or disbursement be fair and reasonable and have been disclosed in a timely fashion. That is a constraint on conduct and communication, not a published tariff.
So the discipline has to come from inside the firm, and that is exactly where a data tool is useful.
Four inputs, all of which you already have if you record time at all. Historical effort by engagement type and by staff grade, so the estimate is built from your mix rather than from an assumed one. Overrun history, which is the difference between what you quoted and what it took, broken down by cause. Client-driven rework — late records, missing documents, mid-year scope changes — because that is the variance a fixed fee has to absorb. And the boundary itself: what is included, what is extra.
Practice-management platforms increasingly package this. Karbon, for example, describes its Time & Budgets feature as tracking time and managing budgets in one place, which is the raw material any pricing model needs. The point is not the specific vendor; it is that a fixed fee estimated without recorded effort data is an opinion, and one estimated with it is a forecast you can defend to a client and revisit next year.
If you quote a headline fee and then add obligatory charges the client cannot avoid, you have a problem beyond client relations. Subsection 74.01(1.1) of the Competition Act states, for greater certainty, that making a representation of a price that is not attainable due to fixed obligatory charges or fees is a false or misleading representation — with a carve-out only for amounts imposed by or under an Act of Parliament or a provincial legislature. Sales tax is outside the rule; your own mandatory technology fee, filing fee or administration charge is not.
The practical translation for a practice: whatever a client cannot decline should be inside the quoted fixed fee. Genuinely optional services can be priced separately, but they have to be genuinely optional. Our sister firm’s explainer on drip pricing rules sets out how that plays out in Ontario, and its wider guide to advertising and marketing law covers the rest of the advertising rules.
Illustrative only, with no figures, because your figures are the only ones that matter. A three-partner practice prices year-end compilation work by memory and last year’s invoice. Realisation on the smallest clients is poor and nobody can say exactly why.
The rebuild starts by grouping the book into four bands by the thing that actually drives effort — the state of the client’s bookkeeping — rather than by revenue. For each band the firm pulls recorded time for the last two cycles, separates the base engagement from the rework, and looks at what triggered the rework. The pattern that emerges is almost always the same: a small number of clients whose records arrive incomplete consume a disproportionate share of senior time.
Three decisions follow, and none of them are made by the model. The base fee per band is set from the base effort. The rework is priced explicitly as a documented add-on with a stated trigger, disclosed in the engagement letter rather than sprung in an invoice. And the worst band is given a choice: a bookkeeping clean-up scope at a separate fee, or a higher band price. The tool’s contribution was the segmentation and the effort analysis. The partners decided everything that mattered.
It does not set the fee, it does not decide whether to take the engagement, and it does not sign anything. It also should not be given your client list and asked to recommend price increases per client without a partner reviewing every line — a fee change is a client conversation, and the model has no view of the relationship, the referral source or the work you hope to win next year.
And it does not fix a collection problem. If the difficulty is not the price but getting paid after the fact, that is a different exercise; our sister firm’s guide to suing a customer for an unpaid invoice in Ontario is the starting point, and the boring preventive answer is a deposit or progress billing written into the engagement terms.
Can AI benchmark my fees against other firms?
Not from data it has. Whatever number it produces is a pattern from public text, not a survey of Canadian practices, and quoting it to a client would be presenting an invented figure as a fact. Where real benchmarking data exists it comes from a named published source, and you should be able to name it.
Should the fixed fee include the scope-creep allowance or price it separately?
Either can work, but only one can be undisclosed, and that one is neither. Whatever the client cannot avoid belongs in the quoted price.
Does this apply to the accounting inside my own business?
No — this article is about a practice pricing work for a book of clients. Running the finance function of a single company is a different problem and it lives on the accounting automation page.
A 30-minute call is enough to tell you whether AI pays for itself here.
