The Three Shifts.
AI compresses the time the work takes. If a firm’s revenue is tied to that time, revenue compresses with it. The Three Shifts decide which firms hold their margin and which lose it.
Everyone adopted AI. Almost no one repriced.
Nearly every organization has adopted AI somewhere. Almost none see it in their earnings. Both groups bought the same models. The gap isn’t the tooling. It’s the question they asked.
Most asked how to do the existing work faster. The 6% asked what the work is worth now that AI exists, and what to charge for it. The first question saves hours. The second changes the business.
88%
6%
Adopted AI
See an effect on earnings
Two firms, the same tools, opposite trajectories.
Clio tracked law firms over four years. The ones that moved to modern delivery and pricing (fixed fees, continuous advisory, outcome-based work) nearly doubled their revenue, on only 50% more clients. The ones that stayed hourly lost half (Clio’s Legal Trends Report, four years of tracking).
Same tools, same day, opposite outcomes. One firm changed what it sells; the other did the same work faster. The Three Shifts are the difference between them.
Adopting AI is not the shift. Changing what you sell is.
Three shifts, three ladders.
Pricing, delivery, and a productization framework, one shift each. Every shift moves work from a position AI compresses to one it makes more valuable.
HoursOutcomes
Billable hourRetainerProject feeOutcome
AI compresses the hours and the revenue with them. Clients know it. PwC’s Chief AI Officer Dan Priest says clients hear about the savings and ask for “their fair share of those efficiencies.” On an hourly model, that share comes straight out of your margin.
Name the result, price it, deliver it. Price the value, not the time.
59% of law firms now use flat fees alongside or instead of hourly billing (Clio Legal Trends 2025). McKinsey runs roughly a quarter of its global fees on outcome-based structures, per UK managing partner Michael Birshan in November 2025.
Convert one defined offering (contract review, the monthly close, a fixed-scope analysis) to a deliverable fee. Not the whole book.
ReactiveProactive
ProjectStandbyWatchFirst call
Every engagement starts at zero, and AI shrinks the per-project fee every quarter. Between engagements you’re dark, so the client gets to decide when the relationship is worth paying for. The model that resets to zero loses ground to the one that compounds.
Go always-on, then call first. Sell the visibility you produce, not the hours you log. No RFP, no comparison shopping.
Firms running continuous client accounting earn 30% or more in monthly recurring revenue than compliance-only competitors (2024 CPA.com/AICPA Client Advisory Services Benchmark Survey). McKinsey’s internal tool Lilli redirected its consultants’ time from reactive research to proactive insight, per McKinsey’s own Lilli rollout report.
Pick one client-relevant signal AI can watch, and sell the visibility on a monthly retainer. Calling first follows once you’re already watching.
The firm that calls first has no competitor on that engagement. The client didn’t go to market. Nobody else has the same finding.
BespokeScalable
Custom serviceNamed offeringRepeatable methodologyLicensed IP
AI can’t codify what was never written down. When the method lives in a partner’s head, the offering walks out the door when they do, and a named offering still hits a ceiling, because that partner is the bottleneck.
The productization framework is the climb: build the method, train the team, keep the judgment. The partner who codifies becomes the architect of something that serves 50 clients instead of 5.
93% of accounting firms say they offer advisory; almost none have systematized it well enough that the work survives the senior person being unavailable (Future Ready Accountant Report). You can’t train a junior to deliver advisory that lives only in a partner’s head, and you can’t price a service you’ve never scoped.
The first decision: keep the line, or not.
Every ladder on this page rebuilds a line the firm intends to keep. Some lines can’t be kept: the client can get the deliverable without the firm at all. Repricing a line like that changes nothing, and the decision about it comes before any ladder.
A firm has four choices with an exposed line.
Defend
Keep the line by selling what software can’t deliver: a person who knows the client, answers for the outcome, and signs their name to the work. The costliest of the four, and it buys the most time.
Repackage
Sell the same work in a shape a software company can’t match: continuous instead of one-off, bundled instead of itemized, priced on an outcome you answer for. The only one of the four the ladders below serve.
Migrate
Move the revenue to an adjacent line the client already budgets for, through the relationship the exposed line built.
Harvest
Keep the line for the cash it still produces, stop investing in it, and move the freed partner time to a line with a future. The partners have to make that call together and on the record, or nothing actually changes for a year.
Five tests, run in order, sort a line into one of the four. Each test either settles the question or narrows what’s left.
Economics
Can the firm deliver this at a cost that survives the software company’s price, counting its funding and its willingness to lose money? If no, stop here: the line harvests, whatever the other tests say.
Accountability
Does someone have to answer for this work? Signed, licensed, insured, or fired if it’s wrong. If yes, the software company has a ceiling it can’t price through, and Defend stays on the table.
Substitution
If the client bought this from the software company tomorrow, what breaks for them that they’d call you about? A real answer keeps Defend and Repackage alive. “Nothing, really” removes both.
Relationship
If this line disappeared, would the client still take your call? If yes, Migrate is open. If no, the line was the relationship, and the choice is between Repackage and Harvest.
Complaint
Is the client’s complaint about the deliverable itself, or about how it’s delivered and priced? A complaint about shape points to Repackage. A complaint about the deliverable points to Migrate or Harvest.
When two qualify, take the cheaper commitment. In rising order of cost: Harvest, Migrate, Repackage, Defend. A partner who can honestly reach Defend usually should. A partner who reaches for Defend without passing tests 2 and 3 is protecting a habit.
Run on real service lines, one from each vertical with a published AI Exposure Index, the tests come out like this:
- A law firm’s e-discovery practice survives on cost and accountability, and the client’s complaint is the hourly bill, so the line repackages into fixed-fee discovery.
- An accounting firm’s tax preparation can’t match flat-fee, software-prepared returns on cost, and the problem doesn’t improve with skill, so the line harvests.
- An agency’s paid media buying has no license protecting it and a small account loses nothing by leaving, but the client relationship holds, so the revenue migrates to work the client already budgets for.
- A consulting firm’s board advisory is a senior advisor’s judgment, answered for in person; no company in the census sells that counsel as a product, so the line defends.
The ladders above are for the lines a firm keeps.
Scale got cheaper. Depth got more powerful.
PwC, EY, KPMG, and Deloitte have collectively invested more than $10 billion in AI since 2023, and for the first time they can serve the mid-market profitably. That sounds like the end of the story for everyone below them. It isn’t.
Their model is built for scale and standardization; mid-market firms are built for depth. AI makes scale cheaper, which is the Big Four’s game, and depth more powerful, which is yours. The structural limit of a standardized model is the opening for the firm that knows the client’s situation and picks up the phone.
One shift, one offering.
Don’t try to move the whole book. Switching everything at once produces partner disagreement and stalled implementation. Pick the shift where the pressure is highest and the work is most defined, and convert one offering.
The free assessment shows where your firm stands across the shifts. The Workshop names and prices 2–3 offerings tightly enough to put in front of named clients for a real yes or no.
Questions
Hours to outcomes (pricing), reactive to proactive (delivery), and bespoke to scalable (the productization framework). Three shifts, one ladder each. Each moves the firm from work that compresses under AI to work that holds or grows margin.
No. The shift is in what you sell, not which AI you adopt. McKinsey found 88% of organizations have adopted AI but only 6% see a measurable earnings effect (2025 State of AI). Adoption alone doesn’t move margin.
Usually pricing. Convert one hourly-billed service to a fixed or outcome fee. It’s the most defined work and the fastest to prove. 59% of law firms already use flat fees, and flat-fee billables grew 34% from 2016 to 2025 (Clio Legal Trends 2025).
Margin compresses. Clio tracked law firms over four years: the ones that moved to modern models nearly doubled revenue, while hourly firms lost about half (Clio Legal Trends). The work AI speeds up is the work clients stop paying full price for.
Defend, Repackage, Migrate, and Harvest: the four things a professional services firm can do with a service line that AI has exposed. Defend keeps the line by selling the accountability software can’t match. Repackage changes the shape and pricing of the work. Migrate moves the revenue to an adjacent line through the client relationship. Harvest keeps the cash, stops the investment, and moves the partners’ time to a line with a future.
Five tests, run in order: economics, accountability, substitution, relationship, and complaint. Each test settles the choice or narrows what’s left. A line that fails the economics test is Harvest regardless of the rest; a line that passes accountability keeps Defend open; a shape complaint points to Repackage.