Not everything changes overnight.
But much more is already changing than we want to admit.
This may be the most honest way to describe what is happening to consulting, legal advisory, research, design, strategy, and other forms of knowledge work in the age of AI.
AI does not simply reduce the amount of work humans do. That is too narrow a view. The real change is more structural. AI is changing the way expertise is produced.
For a long time, many knowledge services looked like the craftsperson on the left.
One expert.
One client problem.
One carefully shaped output.
A report, a memo, a proposal, a legal opinion, a market study, a strategic recommendation.
Each output was made almost by hand. Its value came from individual judgment, accumulated experience, refinement, and the credibility of the person or institution behind it.
That world will not disappear overnight. Some problems will still require deep craftsmanship. Some clients will still pay for bespoke judgment. There will always be room for omakase, handmade work, and master craft.
But the baseline is changing.
What AI brings is not merely “productivity improvement.” It is closer to the industrialization of knowledge work.
The image on the right captures this shift more precisely.
The bowl still needs polishing.
It still needs decoration.
It still needs an experienced eye.
But the basic form no longer begins with a single craftsperson slowly shaping every curve by hand.
The base form is produced faster, more consistently, and at larger scale. Human effort moves upward: selecting, refining, correcting, styling, contextualizing, and giving the output its final meaning.
Many knowledge services are moving in the same direction.
A consulting report may no longer begin from a blank page. A legal memo may no longer begin with a junior lawyer manually reviewing hundreds of pages from scratch. A market study may no longer require days of repetitive search, extraction, and summarization as its starting point.
The first draft can now be generated, structured, reviewed, and improved much faster than before.
This does not mean consultants, lawyers, analysts, or strategists become unnecessary.
It means the center of gravity moves.
In the past, a large part of the premium was attached to the act of production itself: collecting materials, structuring slides, drafting paragraphs, arranging logic, comparing sources, and building the first version.
That work was necessary. It took time. It was billable. But in many cases, it was closer to intellectual manufacturing than to true expert judgment.
AI compresses that layer.
The new premium moves toward a different set of capabilities: defining the problem more precisely, verifying the answer, interpreting it in the client’s context, taking responsibility for risk, and turning a machine-generated draft into a decision-ready answer.
Expertise does not disappear.
Weak expertise simply becomes easier to expose.
When anyone can produce a first draft, the value of simply “having a draft” declines. The value shifts to knowing whether the draft is correct, useful, defensible, differentiated, and executable.
This is why the future of knowledge services will not be a simple contest between humans and AI. The market will become more segmented.
There will be industrialized services: fast, standardized, AI-assisted, and lower-cost.
There will be premium services: grounded in deep context, judgment, relationships, and accountability.
And there will be hybrid services: machine-generated foundations combined with human review, domain logic, expert refinement, and final advisory ownership.
The risk for professional service firms is not that AI replaces them all at once. The real risk is that clients stop paying premium prices for work that now looks like standardized production.
The danger is not extinction.
The danger is commoditization.
This change is already visible in project work.
In just one year, the atmosphere of consulting projects has changed. Clients now come in with drafts made through AI and collaboration tools. Internal discussions are already partly done. In some cases, even the first version of the PowerPoint deck has been created with AI and reported internally before consultants enter the room.
Then the question changes.
How quickly must a consultant catch up on the first day, or within the first month of a project?
And is catching up even enough?
In the past, consultants helped clients organize the problem. Now clients often arrive with a problem that has already been partially organized. But it is still incomplete. It has internal context, but lacks external perspective. It has a draft, but not enough judgment. It has a direction, but not yet an executable structure.
In this setting, the consultant’s role is not to create the first draft.
It is to quickly find the limits of the client’s draft and turn them into better questions.
My team is also changing the way we work.
We ask AI more questions. We let it conduct automated discussions. We feed it the client’s As-Is, AI-driven To-Be hypotheses, and the industry and project experience I have accumulated. We break the client’s organization down into lower levels and discuss possible issues by function, role, and decision point.
At first, AI does not understand the context. Even after context is provided, it is still not enough. So we keep asking. We feed in the client’s organization, work processes, decision structure, data, and exceptions. We listen to the AI’s response. Then humans discuss again.
In that process, the sense of project time changes.
In the past, four consultants might conduct interviews, refer to similar project materials, adjust the tone and manner, and work through the night to build the first draft. The first draft alone could take one and a half months.
Now, the same level of first draft can often be produced by one expert-level director and one manager-level consultant. If factual validation and source checking are needed, one or two junior consultants can support the work. And no one has to stay up all night.
Instead, the client becomes busier.
The client has to answer sharper questions, validate AI-generated hypotheses, and transfer internal context faster. In the past, consultants were busy producing materials. Now clients and consultants are busy validating hypotheses together.
The three months of a project are used differently.
Previously, the first one and a half months were spent drawing the first picture, and the remaining one and a half months were spent improving it. Now, within the first one and a half months, it is possible to approach the core conclusions that the project originally aimed to reach in three months. The remaining time should no longer be spent on producing the draft. It should be spent on refinement, validation, removal, and execution design.
Remove what is unnecessary.
Keep only what has an edge.
Turn the output into sentences the client can actually decide on.
Make not a report, but the material for decision-making.
The problem is that the pricing model of many professional service firms has not yet caught up with this change.
For a long time, consulting has been sold by headcount and duration. How many people for how many months? That question sat at the center of the estimate. Client procurement systems were also used to that logic.
But when AI compresses the production layer, this model becomes increasingly uncomfortable.
Clients expect faster drafts.
Consulting firms still have to explain people and duration.
The real value begins to shift from production volume to judgment, validation, and execution feasibility.
The contract structure remains in the past, while the way of working has already changed.
This may be where disruption in knowledge services begins.
There is no need to pretend that AI is not being used. The opposite is probably better. Firms should say clearly that, because they use AI, they can review more, faster. But the time saved should not simply be converted into price reduction. It should be used to sharpen the client’s core questions.
Not more drafts, but better questions.
Not more slides, but faster judgment.
Not longer staffing, but denser validation.
Just as manufacturing changed after the Industrial Revolution, knowledge industries cannot avoid a change in their production model.
This time, however, what enters the factory is not iron and machinery, but sentences, judgment, and experience.
And where humans remain within that process will define the next shape of knowledge services.