AI compresses the leverage pyramid your firm is built on. The judgement that survives is the firm.
A firm that sells expertise by the hour will not be saved by doing the same billable work faster. The hard question is not whether AI can draft the deck. It is what your judgement is actually for once it can.
Request a Strategic Briefing →Professional-services firms sell judgement by the hour, and the leverage pyramid, juniors doing the work that partners review, is exactly the structure AI compresses first. Most firms have answered by treating AI as a way to do the same work faster, which protects margin for a quarter and erodes the thing the client was paying for. A generic answer dressed as expert advice is the most expensive mistake a firm that sells expertise can make. The real question sits at leadership level: what the firm's judgement is for, how partners reason with AI rather than instruct it, and how the model of expertise holds. Gildoni installs the Havruta Methodology (formerly the Think Partner Methodology) into how leadership teams reason with AI. This is not a productivity tool. It is the reasoning discipline that keeps the firm worth hiring.
The pyramid that makes the money is the part AI compresses first
This is not a forecast. It is already the working reality of the firms closest to the change.
Knowledge work is where generative AI lands hardest, and professional services sits at the centre of it. The World Economic Forum's research puts AI's strongest effects in knowledge-intensive industries, and finds that of the work done in an organisation today, humans handle 47 per cent alone, technology 22 per cent, and human-machine collaboration the remaining 30 per cent, a balance the Forum expects to move sharply towards collaboration by 2030.
And the gain is not spread evenly across the pyramid. Stanford's analysis is blunt about where it concentrates: AI boosts productivity and bridges skill gaps, with the largest gains going to less experienced workers, the very people the leverage model relies on doing the billable hours. When a junior with AI can produce what used to take a senior, the economics of the pyramid stop adding up on their own.
The numbers show how fast the work is moving and how unevenly the gain falls.
of work today is done by humans alone, with 30% already human-machine collaboration. The Forum expects the balance to tip towards collaboration by 2030, hardest in knowledge-intensive work.
of organisations investing in AI report productivity gains over the past year, and 57% call those gains significant. The pressure to do the work faster is now near-universal.
AI boosts productivity and bridges skill gaps, with the biggest gains going to less experienced workers, the base of the leverage pyramid.
The pressure is real. The usual response is the mistake.
Firms and courts might stop asking "Can it write?" and instead start asking "How well, on what, and at what risk?"
Doing the same work faster is the wrong answer
Faced with the pressure, most firms reach for the obvious move: use AI to produce the existing billable work in less time. It feels prudent. It protects utilisation. It is also High-Speed Waste applied to the part of the business the client values most, faster production of work whose value was never the production.
A client never bought the document. They bought the judgement inside it. The deck, the memo, the model, the audit file, these were always the visible residue of thinking the firm was trusted to do. When AI produces the residue and the thinking is left out, the firm has automated the wrapper and quietly hollowed the product.
A generic answer dressed as expert advice is the Mirror Principle at its most dangerous. The machine returns a fluent, confident, average answer, and in a firm that sells expertise the average is precisely what the client could have got without you. If the reasoning going in was generic, the advice coming out is generic, however senior the letterhead. That is not a quality problem a better model fixes. It is a reasoning problem.
It is worth being honest about why this is hard. The disciplines that make a firm worth hiring, framing the real question, knowing what the data cannot tell you, deciding under genuine uncertainty, were never the billable line items. AI has not removed them. It has stripped away the busywork that used to hide how little of the work was actually the judgement, and left firms looking at the part they cannot outsource to the machine.
The faster pyramid is the easy part. The judgement is the question.
Here is what the productivity story leaves out. Every vendor and every analyst now says the same thing: adopt AI, compress the work, protect the margin. Almost none of them say how a firm's judgement is supposed to differentiate once the machine can produce a competent first answer to almost anything.
That is the real question, and it is not a tooling question. Put a client's hardest problem to a commodity AI and it will hand back a confident answer without asking what the firm knows that the model cannot. The firms that hold their premium will not be the ones that drafted fastest. They will be the ones whose partners reason with the machine, pushing it to argue, to find what was missed, to anchor in the engagement's real facts rather than the plausible average. The pyramid compresses either way. What rises to the top, or fails to, is the judgement.
AI squeezes the wide base of the pyramid, the production hours juniors used to bill, towards the middle. The review layer thins with it. What it cannot compress is the apex: the judgement the client was actually buying. The firm's value moves from the volume of hours to the quality of the reasoning at the top, and that is the part a generic AI answer cannot fake.
What the Havruta Methodology installs at partner level
The Havruta Methodology is that discipline. It changes the default behaviour of the machine from agreeing with the partner to reasoning with the partner, which is exactly what a firm that sells judgement needs the technology to do.
The Flip
The Flip puts the machine on the other side of the engagement. Instead of confirming the recommendation, it argues against it: where is this advice generic, what would a sharper competitor say, what is the client not being told. The partner is challenged in the room, before the client or the market does the challenging.
Ground Truth
Ground Truth keeps the reasoning anchored in the specifics of the engagement, the client's real numbers, history and constraints, rather than the plausible average an AI produces by default. Advice built on a generic answer is worse than no AI at all, because it carries the firm's name on it.
Decision Velocity
And Decision Velocity lets the firm move from question to defensible position at the speed clients now expect, compressing the work without surrendering the judgement that justifies the fee. The pyramid gets faster. The thinking stays the firm's.
The fuller account of how all of this works is on the methodology page.
What this is not
This is not a productivity tool and it is not automation software. It is not a drafting copilot, a practice-management platform, or a document-assembly engine. It is not AI training or general AI literacy. The tools that speed up the work are a separate market, and most firms already own them. This is the thinking that decides whether the faster work is still worth paying for.
It changes how the leadership team reasons about the judgement the firm sells: the engagement strategy, the recommendation, the call the client is actually paying a partner to make.
Where a leadership team starts
The methodology is installed along a ladder, and a firm's leadership enters at the rung that fits.
Most begin with the Eye-Opener Workshop, a half-day in which the partners see the shift on a live engagement, not a demo.
A leadership group embeds the practice through the Havruta programme, taking the discipline across the team so it holds under client pressure.
A single high-stakes question, how the firm prices judgement, where its expertise still differentiates, what the next three years of the model look like, can be worked through Advisory Havruta.
How a CEO and the leadership team reason with AI
For the managing partner and the leadership team specifically, the role page takes the same discipline to the seat that owns the firm's direction. A Strategic Briefing is how to decide where to begin.
Go to the CEO pageQuestions a managing partner asks about AI
Will AI replace the work my firm bills for?
It will compress the production work, the drafting, the first-pass analysis, the assembly, that has always sat at the base of the leverage pyramid. The research is clear that the gains land hardest on less experienced workers, which is exactly that base. What it does not replace is the judgement at the top, the framing of the real question and the decision under uncertainty. The risk is not that AI takes the work. It is that firms automate the wrapper and leave the judgement out.
Why is using AI to do the same work faster the wrong answer?
Because the client never bought the speed of production. They bought the judgement inside the output. Producing the same deliverable faster protects utilisation for a quarter while quietly hollowing the thing the fee was for. We call it High-Speed Waste: faster output of work whose value was never the output. The firms that hold their premium use AI to sharpen the reasoning, not just to shorten the hours.
What is the danger of a generic AI answer in a firm that sells expertise?
A commodity AI returns a fluent, confident, average answer. In a firm that sells expertise, the average is precisely what the client could have got without you. This is the Mirror Principle: if the reasoning going in was generic, the advice coming out is generic, however senior the name on the letterhead. Dressed as expert advice and billed at partner rates, that generic answer is the most expensive mistake the firm can make, because it puts the firm's name on something the machine produced for anyone.
Is this AI software or a productivity tool for the firm?
No. It is not a drafting copilot, a practice-management platform, document-assembly software, or AI training, and it does not touch your stack. Those speed up the work, and most firms already own them. This addresses the thinking underneath: how a partner reasons with AI so the advice is genuinely the firm's, anchored in the engagement's real facts, and stress-tested before the client or a competitor tests it for you.
How does a firm protect its premium as AI compresses the work?
By moving the firm's value from the volume of hours to the quality of the reasoning at the top, and by making AI a thinking partner rather than a faster pair of hands. That means making the machine argue the recommendation rather than confirm it, anchoring it in the client's real situation rather than a plausible average, and deciding at the speed clients now expect without surrendering the judgement. The pyramid compresses for everyone. The judgement is what differentiates.
Where should we start?
With a Strategic Briefing, or with the Eye-Opener Workshop, where the partners see the difference between instructing AI and reasoning with it on a live engagement rather than a demo. From there the path depends on whether you are embedding the practice across a leadership group or working a single high-stakes question about how the firm's model of expertise holds.