A managing partner in a fifteen-person accountancy practice opened a job file last month and saw the same pattern she had seen all year. The corporation tax return that used to take twelve hours of senior time now took three. Her bookkeeper had built a workflow that pulled the trial balance, ran the disallowables, drafted the computation, and queued it for review. The work was good. The client had no idea how long it had taken. And the partner was sitting with the question every service-firm owner is sitting with. What do I charge for that?
She had two bad options. Bill the old twelve hours, hope the client never notices, and live with the discomfort. Or bill three hours, watch the fee drop by seventy-five percent, and lose the margin she needed to fund the tooling. Neither answer works. The model is wrong.
This piece sits at the strategic level. Sister posts cover vendor pricing, hybrid AI pricing, and pricing a new offer. The question here is which of the four pricing models fits your work, and how to move off hourly without breaking the client base.
What does each model mean, and why does hourly break first?
Service firms have four credible pricing models. Hourly bills the client for time recorded. Value-based pricing charges for the outcome the client receives. Flat fee charges a fixed price for a defined scope. Retainer charges a recurring monthly fee for access or a defined book of work. Hourly breaks first under AI productivity because the gain is real, the client sees it, and the per-hour rate cannot rise indefinitely.
The mathematics is unforgiving. If a return used to take twelve hours at £150 an hour, the client paid £1,800. If the same return now takes three hours, hourly billing collapses the fee to £450. To hold the fee constant the firm would need to charge £600 an hour, which sits well above the published rates of senior partners at much larger firms. Clients notice. Procurement teams notice. The rate ceiling is real.
The deeper problem is that hourly billing tells the client they are buying time. Once they believe that, every conversation becomes a negotiation about hours, and under AI productivity that conversation only travels downwards. Ron Baker’s work on value pricing in accountancy, echoed by ICAEW guidance, has made this argument for two decades. AI has made it operationally urgent rather than philosophically interesting.
When does value-based pricing fit, and when does it not?
Value-based pricing fits work where the client can name the outcome they want and roughly what it is worth to them. That covers advisory work, deal work, strategic projects, tax structuring, transaction support, recovery cases, and high-end consultancy. It does not fit small recurring compliance work because the client does not value the outcome differentially. A VAT return is a VAT return. They want it filed, on time, correctly, for a known fee.
The discipline value pricing demands is harder than firms expect. It requires the firm to know what the client values, not what the work costs. That means upfront conversations about the situation, the size of the prize, the cost of getting it wrong, and the alternatives the client has. Many UK firms attempt value pricing, get uncomfortable at the conversation, and quietly fall back to cost-plus dressed up as a fixed quote. The cost-plus floor leaks the productivity gain straight back to the client.
Where value pricing works under AI productivity, the lift is meaningful. BCG analysis of professional services suggests firms that hold their fees while AI compresses delivery cost can see gross margin expand by ten to twenty points over two years. Maister’s framing of the professional service firm, written before AI, still applies. Charge for the outcome, not the input.
When does flat fee work, and what is the trap?
Flat fee is the cleanest model for productised service work, work that comes in a defined package, runs on a repeatable workflow, and produces a known deliverable. Annual accounts for a limited company, a discovery engagement, a fixed-scope marketing audit, a defined legal review. The fee is set, the scope is written down, the change-control process is explicit, and AI productivity becomes margin rather than a discount conversation.
Two disciplines hold flat fee together. The first is scope. Every flat-fee engagement needs a written statement of what is in, what is out, and what triggers a variation. Without that, the client asks for one more thing, then another, then the engagement bleeds the productivity gain across out-of-scope requests. The Harvard Business Review good-better-best framing helps here. Offer three tiers, make the boundaries explicit, let the client self-select.
The second discipline is pricing against value and sustainable cost, including the AI tooling, the prompt and workflow IP, the quality review layer, and the partner judgement. Drop any of those costs from the calculation and the model looks fat. Hold them all and the margin is earned. The Law Society’s pricing guidance for solicitor firms makes the same point in a regulated context. Fixed fees work where scope can be specified and where the firm captures the productivity gain rather than discounting it away.
When does the retainer model compound, and when does it break?
Retainers compound when AI lifts the depth and turnaround of a fixed scope. The client pays a known monthly fee for a defined book of work or advisory relationship. AI inside the workflow lets the firm fit more value into the same fee without consuming more partner time. The client feels the service improving, and the renewal conversation runs itself. UK fractional CFO and CTO retainers sit between two and eight thousand pounds per month.
The retainer breaks when scope is open-ended and AI productivity floods the request volume. If the retainer says “ad-hoc advice and support as needed”, the client will figure out within three months that they can ask for more, and the firm will respond because the relationship matters. Within six months the senior team is burned out, the margin has evaporated, and the renewal conversation goes badly. The fix is scope. A retainer needs a written response SLA, a defined attendance pattern, a clear escalation tier, and an explicit boundary around inclusions.
McKinsey’s 2024 State of AI work and PwC’s UK CEO survey both flag the retainer compression risk. Firms that adopt AI inside an open-scope retainer without re-papering the engagement letter tend to lose margin within a year. Firms that hold scope discipline and refresh terms on renewal tend to gain it.
How do you pick the model, and how do you transition off hourly?
Three tests pick the model. The work shape test asks whether the work is variable, defined, productised, or relationship-led. Variable advisory leans value-based. Defined transactional work leans flat fee. Productised recurring work leans flat fee with annual review. Ongoing relationship work leans retainer. The client buying behaviour test asks whether the client buys outcomes, packages, time, or access. The AI productivity test asks where AI compresses delivery cost the hardest.
The transition off hourly typically takes twelve to twenty-four months and runs through the renewals cycle, not a single board meeting. Start with the segment where scope is tightest and AI productivity strongest, usually a productised compliance line or a defined advisory product. Re-paper that segment at the next renewal, holding the fee constant and converting the engagement letter from time-based to fixed. Run the new model alongside hourly for the rest. At twelve months, review which segment converted cleanly and re-paper the next.
Deloitte’s 2024 Human Capital Trends work on AI adoption in professional services suggests staged transitions retain more revenue than wholesale flips. The conversation with each client matters more than the model. Done well, the client hears the firm investing in capability, holding quality, offering a predictable fee. Done poorly, they hear extraction. The difference is preparation, scope clarity, and partner confidence that the new fee is fair. That confidence is the one thing AI productivity will not give you. You build it yourself.
