Under time-based billing, the hours AI saves mostly come off the invoice or disappear into unbilled work. Under a well-scoped fixed fee, they become margin.
We see it in accounting firms, deal advisory boutiques and law firms alike: most have bought AI licences and trained their people over the past year or two, but the way they quote and bill their clients hasn’t followed. The engagement letter that goes out today looks very much like the one that went out before anyone in the firm had opened a chatbot.
When we raise this with the firms we work with, the conversation tends to go one of two ways. Some partners are convinced the billable hour will be gone within a few years, while others expect very little to change, because in their eyes clients pay for experience and judgement, and they don’t see AI replacing either of those.
Our view sits somewhere in between. Time-based billing will remain the right model for a large part of professional work, but on the repeatable part of the business, the work a firm has done a hundred times and could describe on a single page, a fixed fee is the simplest way to turn the hours AI saves into margin without renegotiating your rate every year.
Who keeps the time AI saves?
Getting AI into the building has turned out to be the easy part. Organisation-wide AI use in professional services almost doubled in a year, from 22% to 40%, according to a Thomson Reuters survey of more than 1,500 professionals. European accounting firms followed the same curve, from 8% to 42%, according to Wolters Kluwer.
Turning that into results is another matter. Only 18% of professionals in that survey say their organisation tracks what its AI tools return. Among Europe’s leading lawyers, 82% say their practice has a clear AI plan, but only 55% have a strategy for making money from it, according to Thomson Reuters.
Meanwhile, the time savings are real. In a 2025 Thomson Reuters survey, professionals expected AI to save them around five hours a week, roughly six working weeks a year. Their clients are not standing still either: AI use among corporate M&A executives more than doubled in 2025, to 45%, according to Bain.
So the question most firms haven’t answered is who keeps those six weeks, the firm or the client?
Why time-based billing works against the firm
The reasoning behind this is simple. Under hourly billing, revenue is billed hours multiplied by rate, so every billable hour AI takes out of an engagement also comes off the invoice. The only exception is work the firm was already writing off, such as an overrun, a partner trimming the bill or a correction nobody wants to charge for. Hours AI saves there were never on the invoice, so they become margin. In other words, under hourly billing the only savings a firm keeps are on work it wasn’t charging for anyway.
Let’s take a simple illustration. A financial due diligence data pack used to take 40 hours at €200 an hour, which comes to €8,000. With AI-assisted extraction and first-draft analysis, the same deliverable now takes 25 hours. Billed by the hour, the invoice drops to €5,000, so while the firm has paid for the tools and the training, it earns €3,000 less for the same deliverable. You can run the same sum on a year-end closing that goes from 30 hours to 20, or a standard contract review that goes from 12 hours to 7, and the outcome looks roughly the same.
It is no surprise that many firms reach for the most obvious answer, which is to raise their rates. In the US, 61% of accounting firms have raised fees to cover technology investment in general, according to the AICPA, while 40% have not yet worked out how to track the efficiencies it brings.
How far you raise them matters. If you raise the rate until the invoice is back where it was, the firm keeps the entire gain while the client pays the same for work they know is partly done by AI. We don’t think that is fair to clients, and we doubt it lasts: they use AI in their own teams, they assume you do too, and the first competitor who passes part of the gain on will have a very easy conversation with them. A more modest increase, where the client still ends up paying less than before, is a fairer deal and certainly better than doing nothing.
The trouble is that a rate increase only solves the problem once. Your revenue is still tied to the hours, so every next step in efficiency comes off the invoice again, and you find yourself back at the table asking for a higher rate. Explaining to a client why your hourly rate keeps going up while the work keeps getting faster is not a conversation anyone looks forward to, and with hourly billing it is one you will have to have over and over.
A fixed fee settles the question on the deliverable instead of the hour. You agree once on a price that is fair to both sides, typically below what the client used to pay, and from then on every hour the team saves stays with the firm. Think back to the hours that were written off before: under hourly billing, those were the only saved hours that turned into margin. Under a fixed fee, every saved hour behaves that way, because none of them comes off the invoice any more. That is the heart of our argument: a fixed fee lets the firm keep the gain and still give the client a better price.
There is one condition. Saved time only becomes margin if someone decides where it goes. Left alone, it tends to disappear into extra polishing and longer reviews, so what makes a fixed fee profitable in the end is how far the people doing the work have actually changed the way they work.
A fixed fee is not a permanent shield either. Clients will benchmark the price at renewal, and competitors will undercut it as AI makes the work cheaper for everyone. What a fixed fee buys is a lead: the firm keeps the gain until the market price catches up, and the firms that stay ahead are those that keep improving faster than the price falls. As the chart shows, it also earns less per hour than a full rate rise. That is the price of a deal the client will still accept next year.
With a fixed fee, the client pays less and the firm earns more for every hour worked
One due diligence data pack that AI cuts from 40 hours to 25. Billed at €200 an hour before AI; fixed fee set at €7,000.
What the client pays
€8,000
€3,000 less
€5,000
€1,000 less
€7,000
Before AI
40 hours
Hourly billing
with AI,
25 hours
Fixed fee
with AI,
25 hours
What the firm earns per hour worked
€200
No change
€200
+40%
€280
Before AI
40 hours
Hourly billing
with AI,
25 hours
Fixed fee
with AI,
25 hours
Illustration: Merlin Innovation. Hypothetical example.
Where fixed fees work first
Obviously, you can’t move your entire company to a fixed-fee structure overnight. The place to start is repeatable, well-scoped work with a clear deliverable and enough history behind it that you already know roughly how much effort goes into those deliverables.
Accounting and tax: year-end closing and the corporate tax return for an SME client with stable activity.
M&A and transaction advisory: a standard financial due diligence data pack, or a first-draft information memorandum built from an agreed set of inputs.
Legal: standard contract review against an agreed playbook, or a recurring regulatory filing.
What these have in common is that the scope can be written down, the effort is reasonably predictable, and AI changes that effort in a way you can measure.
Each sector starts from a different place. Many accounting firms already charge fixed monthly fees for recurring bookkeeping and compliance, so the move is mostly a matter of extending that logic to project work such as the year-end closing. Transaction advisers are close for a different reason: a success fee is already the nearest thing to value-based pricing that most professional firms have, and every hour AI saves on a deal that closes goes straight into the margin on that fee. What remains exposed is the retainer and the time-billed diligence and valuation work behind it, and that is where a fixed, scoped deliverable makes sense. Law firms lean most heavily on the hour, so they have the furthest to go, but also the most to gain.
Open-ended work is a different matter. Disputes, complex negotiations, investigations and contested transactions carry too much uncertainty for a single fixed price, and forcing one usually ends with either the firm or the client feeling short-changed. For that kind of work, we would split the engagement into stages instead of pricing it in one go. The first stage, such as an initial assessment of a dispute or a first round of due diligence, can usually be scoped well enough to carry a fixed fee. Once it is done, both sides know far more about what lies ahead, and the next stage can either get its own fixed fee or stay on hourly billing with an agreed ceiling, so the client always knows the maximum they will pay.
A fixed fee is safest where the scope fits on a page and you have priced the work before
Which fee model to use, by how well the work can be scoped and how much history you have
Pricing it right
Some early pilots will be underpriced. That is the cost of learning, and most firms already hold the data to keep it small; it simply hasn’t been used for this purpose yet. If we were sitting down with your team to price a first fixed-fee engagement, this is roughly how we would plan the next six weeks.
Weeks 1 and 2: get the history out of the timesheets. Choose the engagement type and pull every past instance of it from your time records, whether those hours were booked against a closing file, a deal or a contract. The narratives in time entries are usually messy, which makes this a good first job for AI itself: a language model can sort them into phases and task types in a matter of days, where doing it by hand would keep someone busy for weeks. A person who knows the work still needs to check the result, but you end up with a first honest view of where the hours actually go.
Week 3: break the engagement into phases and study the spread. With the entries sorted, look at how long each phase took across past engagements, and pay as much attention to the range as to the average, because the average hides exactly the difficult files that would hurt you under a fixed price.
Week 4: estimate the effort with AI in the workflow. Go through the phases with the people who deliver them and agree how much each one changes with the tools you now have. First-pass document review and data extraction tend to shrink considerably, while client meetings, judgement calls and sign-off barely move. This is also the moment you find out whether your teams genuinely use the tools, and it is worth being honest about it, because an estimate built on adoption that hasn’t happened yet will cost you money.
Weeks 5 and 6: write the scope down and build in protection. Set out the assumptions the price rests on, such as the number of entities, documents, contracts or adjustments it covers, add a sensible buffer, and agree a change mechanism with the client up front, so that anything outside the scope is priced as a change instead of being absorbed by the team.
Selling it to clients
It is tempting to assume clients mainly want a lower price, but at least among large companies the evidence suggests otherwise. In the same Thomson Reuters research, just 4% of general counsel at large organisations named cost or billing as their main complaint about their law firms. What clients tend to value more is knowing the budget up front, not being surprised by the invoice, and paying for the outcome rather than for an army of junior staff brought to every meeting.
The same logic holds across sectors. A finance director commissioning year-end work wants to know the cost before the year closes, and a private equity buyer commissioning a diligence pack wants a number that fits the deal budget. Both will prefer a fixed fee with clear assumptions over an hourly estimate with a wide range.
We would not launch this across the whole firm in one go. Start with a few clients you trust, offer a fixed or capped fee on one engagement type with the scope assumptions spelled out, and review the outcome together afterwards. The pilot tests your estimate and your client’s reaction at the same time, and it gives you something concrete to point to when you take the idea further.
A word on the rules
One caveat before you start: each of these professions works within its own rules on fees, and those rules leave some firms more room than others. In Belgium, lawyers can work for a fixed fee as long as it is agreed at the start and not tied solely to the outcome of a dispute, as set out in the Judicial Code and by the Orde van Vlaamse Balies. Accountants and tax advisers record their fee in an engagement letter that the Law of 17 March 2019 makes mandatory. Before you change how you charge, check what applies to your firm with your professional body or your own legal adviser.
What we recommend
Choose one repeatable engagement type per practice. Look for work with a clear deliverable and enough history to price from, such as a closing file, a diligence pack or a contract review, and resist the urge to start with ten at once.
Build the estimate from your own time data. Use your timesheet history, sorted by phase with AI where that helps, to model the effort with AI in the workflow, and settle the scope assumptions, the buffer and the change mechanism before you quote anything.
Make adoption the condition for the fixed fee. A fixed fee only turns into margin if the team genuinely works differently, so redesign the workflow for the chosen engagement, train the people who deliver it, and measure the hours per engagement before and after.
Pilot with a few trusted clients, then review. Offer the fixed or capped fee to a small group, compare the actual effort with your estimate after a handful of engagements, and adjust the price before deciding whether to extend it to more clients.
Our answer
Under hourly billing, the time AI saves either ends up with the client or quietly disappears into work nobody bills, whether the firm planned it that way or not. Under a well-scoped fixed fee, it stays with the firm that actually changed how it works, and some of it can still go back to the client as a better price.
Want to talk through what AI changes in your engagements?
Tell us where you are with AI. We’ll tell you what we’d do next.
Sources and further reading
Thomson Reuters Institute, 2026 AI in Professional Services Report, 9 February 2026. Survey of 1,500+ respondents in 27 countries. Supports the 40% (2026) and 22% (2025) adoption figures and the 18% ROI-tracking figure. thomsonreuters.com
Thomson Reuters Institute, 2026-2027 State of European Legal Markets, with IE University Law School, 5 November 2026. Supports the 82%, 55% and 4% figures. thomsonreuters.com
Thomson Reuters, The AI Adoption Reality Check (press release on the Future of Professionals Report 2025), 26 June 2025. Supports the five-hours-a-week expected saving. thomsonreuters.com
Thomson Reuters, Future of Professionals Report 2026 (press release), June 2026. Survey of more than 1,800 professionals in 62 countries. Supports the 91% figure. thomsonreuters.com
Wolters Kluwer, AI adoption in Europe: from hesitation to acceleration (Future Ready Accountant 2025), 8 October 2025. Supports the 8% to 42% and 35% figures. wolterskluwer.com
Bain & Company, M&A Capability for a New Era (Global M&A Report 2026), 27 January 2026. Bain M&A Practitioners 2026 Outlook Survey of 303 corporate M&A executives at companies with revenue above $100 million, November 2025. Supports the 45% and one-third figures. bain.com
AICPA PCPS and CPA.com, 2025 National MAP Survey: Executive Summary, 2025 (copy hosted by the Indiana CPA Society). Supports the 61% and 40% figures. incpas.org
Belgian legislation, Law of 21 June 2006 inserting Article 446ter into the Judicial Code, Belgian Official Gazette (via openjustice.be). Supports the ban on fees tied solely to the outcome of a dispute. openjustice.be
Orde van Vlaamse Balies, Deontologie-advies 536. Supports the freedom to charge a fixed fee, agreed at the start. ordevanvlaamsebalies.be
Belgian legislation, Law of 17 March 2019 on the professions of accountant and tax adviser, Justel. Supports the mandatory engagement letter. ejustice.just.fgov.be
Commissie voor Boekhoudkundige Normen, Article 3:65 Code of Companies and Associations. Supports the fixed-amount rule for statutory auditors. cbn-cnc.be
European Union, Regulation (EU) No 537/2014, EUR-Lex, 16 April 2014. Supports the extra fee limits for public-interest entities. eur-lex.europa.eu
This paper reflects our general perspective on AI and fee models in professional services as of November 2026. It is general information and does not constitute legal, tax or financial advice.

