You've already done this math. If AI lets your team finish in one hour what used to take five, the invoice shrinks by 80%. The output is the same. The client is happy. Your revenue just dropped.
Every managing partner I talk to has run some version of this calculation. Many ran it months ago. The diagnosis is not the problem.
The problem is that "change what you sell" is obvious advice and almost nobody has actually done it. Not because they disagree. Because the gap between knowing the answer and operationalizing it is enormous.
The gap is in mechanism, not insight. Every managing partner knows the answer. Almost none have operationalized it.
The gap is in mechanism, not insight.
I hear the same thing in nearly every conversation: "I know we need new services. I just don't know what they look like for my firm." That is not a knowledge gap. It is a design problem. And most of the AI advice available right now doesn't touch it.
The AI consulting market is full of adoption guidance — which tools to buy, how to train people, where to automate. That work has value. But it stops short of the revenue question. And the revenue question is the one keeping you up at night.
What "new offerings" actually looks like
When I talk with firms about this, the design process has three parts. None of them require more AI knowledge. They require commercial thinking applied to capabilities you already have.
Redesign the delivery model. If AI handles the research, the analysis, or the first draft, your people spend their time on judgment and client relationships. That's a different service than the one you sell today. A law firm that uses AI for contract analysis isn't selling faster contract review — it's selling risk assessment with attorney judgment on top. The scope changes. The staffing changes. The client experience changes. You have to design all of that deliberately.
Rebuild the pricing. Hourly billing on AI-assisted work is a race to zero. You know this. But switching to fixed-fee or value-based pricing requires knowing your actual cost of delivery under the new model. Most firms haven't measured that yet because they haven't designed the new delivery model yet. The pricing problem is downstream of the design problem.
Find the first buyer. This is where most firms stall even after they've sketched out a new service. They try to sell it to everyone. The move that works: pick one client who has a problem this new offering solves, have a specific conversation, run a pilot, get a reference. The first sale is the hardest part, and it's the part nobody talks about.
Why this hasn't happened yet
It is not inertia and it is not ignorance. A few real things are in the way.
The people who understand your clients and your market — your senior practitioners — are busy delivering work. They don't have 40 hours to design a new service line from scratch. The people selling you AI tools don't know your clients, your pricing, or your competitive position. And the firm's existing incentive structure (billable hours, utilization targets) actively discourages the experimentation required to build something new.
Those are real obstacles. But none of them are permanent. They're design constraints, and design constraints have solutions.
Efficiency has a ceiling. New revenue does not.
Every technology cycle follows the same pattern. The firms that figure out how to commercialize the new capability — how to turn it into something clients will pay for — capture the market. The firms that limit their response to internal efficiency lose ground that's very hard to recover.
BDO's Nick Kervin put a number on the efficiency-only path: 25-40% gains, max. And even those gains are fragile. Brian Solis reported that nearly 40% of AI time savings are being lost to rework. The speed is real. The durability of the speed is not.
TSIA, the industry research body for professional services, named the problem: the cannibalization dilemma. "The more efficient you become, the more revenue you risk losing" under hourly pricing. The efficiency that was supposed to save the model is the thing accelerating its decline. After that ceiling, the only path is changing the model.
You've done the margin math. The external threat compounds the internal margin math. The next step is designing what comes after it.