Artificial Intelligence exposes fragilities in the Legal Business Model

When Scott Kleinman, from Apollo Global Management, said that lawyers, accountants and consultants “are going to see a lot of pressure,” he was not giving a lecture about technology. He was talking about money. Margins, risk, returns, valuation and business model. Maybe that is why the statement caught my attention. It takes artificial intelligence out of that “comfortable place” of events, polished demos and enthusiastic posts, and puts the discussion where it actually becomes serious.

The article published by Valor, based on the Financial Times, gets it right when it shows that investors have started to look more carefully at professional services firms that “bill by the hour,” depend on “few assets” and may face “revenue disruption.” In a more direct reading, the financial market has started asking a question that the legal industry still avoids. Do businesses that sell knowledge actually know how to turn knowledge into an asset?

That is the conversation that matters… This is not about robots replacing lawyers. That discussion weakens the debate before it even begins. It works for headlines, for scaring people on panels and for LinkedIn posts saying that “the future has arrived.” What happens to an economic model based on hours when a relevant part of the work can be accelerated, reorganized or simply exposed by artificial intelligence?

AI does not threaten the legal profession because it is too intelligent. It threatens it because it can reveal, with great precision, where the legal profession is still not very intelligent about itself. The problem is not that a lawyer will be replaced by a machine. The problem is that a law firm may discover that part of what it called sophistication was, in practice, poor organization of its own knowledge.

I have heard a similar phrase many times in conversations with fellow lawyers. “Here, our work is very bespoke.” I understand the point. The practice of law does have a bespoke dimension. There is judgment, context, strategy, sensitivity, responsibility and risk assessment. But there is also a lot of rework treated as care, a lot of lack of process treated as customization, and a lot of delay treated as depth.

For a long time, part of the legal market protected its inefficiencies behind “Complexity.” The delivery took time because it was complex. The legal opinion had fifty pages because the topic was complex. The team needed to be large because the case was complex. Billing by the hour was inevitable because no one would dare measure something so sophisticated in any other way.

Sometimes that was true. Many times, it was just a lack of method. AI starts exactly there. It separates intellectual work from repetitive tasks. It separates legal judgment from the manual organization of information. It separates strategy from assembly. It separates institutional knowledge from individual memory. It separates value perceived by the client from internal effort converted into an invoice.

The other day, I heard a question that captures the discomfort of the moment very well. “But if AI does it faster, how do I bill for it?” That question is much better than it seems. It reveals that the discomfort is not just about technology. It is about the model. If efficiency increases, who captures that gain? Does the firm turn everything into margin? Does the client receive more speed, more quality or better predictability? Does pricing change? Does delivery change? Or do we pretend nothing happened?

The billable hour will not disappear. It would be naive to say that. In strategic, uncertain, transactional, litigation-related or highly specialized work, it will continue to make sense. But the hour has lost its “innocence.” In repetitive, predictable, document-heavy activities based on large volumes of information, billing time as if time were automatically value has become an increasingly weak thesis.

The Brazilian legal market is still very impressed by the obvious. We discovered that AI summarizes documents, helps write emails, compares clauses, supports research, organizes information and generates first drafts. Great. That is useful. But that is not strategy. It is the beginning of the beginning…

The tool stopped being a differentiator the moment everyone gained access to it. Today, the partner has AI. The junior lawyer has AI. The client has AI. The competitor has AI. The legal department has AI. And the firm that thinks it solved the problem by blocking a tool on corporate computers may need to take a closer look at its own team’s phones.

AI is already in everyone’s hands. On personal phones, in the app used “just to test.” Shadow AI is the name given to a simple reality. When leadership takes too long to create a safe path, people find an informal one. And usually, they find it fast.

Blocking a tool without offering a safe alternative is not strategy. It is pretending to have control. And pretending to have control may be one of the most expensive ways to lose control. Professionals will use AI because it solves real problems. The question is whether the firm will create a secure architecture, with rules, privacy, traceability, human review and reliable sources, or whether it will push everything into informality and then be surprised by the risk.

The mature conversation is not about allowing everything or prohibiting everything. Both extremes are bad. Prohibiting everything encourages hidden use and turns the firm into an organization trying to protect the past with a firewall. Allowing everything turns sensitive information into an “experiment.” The serious path is to build governance that allows the firm to accelerate safely. Governance is not the department of “no.” It is what allows you to say yes without acting irresponsibly.

The greatest asset of a law firm is its people, of course. But not people in isolation. The true asset is the knowledge they have produced over years or decades. Theses, legal opinions, negotiations, difficult meetings, mistakes, successes, client insights, decisions made under pressure. That asset is enormous… The problem is that, in many firms, it still lives scattered across folders, emails, old versions, personal templates and individual memories.

We call that accumulated experience. Many times, it is wasted intellectual capital.

Intellectual capital is not just about having great professionals. That is the starting point. Intellectual capital is the ability to capture, qualify, organize, reuse and expand the knowledge produced by the firm. It is making sure a thesis does not die inside a legal opinion. It is making sure a client meeting generates institutional learning. It is making sure a risk identified in a contract feeds an intelligence base. It is making the firm’s history work in favor of its future, instead of being buried in folders nobody opens.

This is where AI stops being a text tool and becomes intelligence infrastructure. If the firm does not know where its knowledge is, if it depends on duplicated files, loose versions, lost messages and the memories of whoever participated in a given case, the problem is not technological. It is managerial. AI does not fix that by itself. It only makes the problem faster, more visible and, in some cases, more dangerous.

I have also heard a phrase that sounds prudent, but worries me. “Let’s wait a little to see where the market goes.” I understand the caution. Law requires responsibility. But there is a huge difference between caution and immobility. Waiting can be sensible when you are assessing risk. Waiting becomes a problem when it is only a way to avoid touching processes, incentives, structures and models that are already showing clear signs of fatigue.

That is why I have been arguing, through the AI-Centric Law Firm Model, that AI in law cannot be treated as an isolated productivity tool. It needs to connect four central dimensions of the firm. Client relationships, legal delivery, risk analysis and knowledge management. When these dimensions do not talk to each other, AI becomes a useful but limited tool. It helps with text, but it does not change the business.

The game changes when AI turns project history into commercial intelligence, recurring demand into product, identified risk into a decision matrix and accumulated knowledge into competitive advantage. It changes when the firm stops asking “how do I use AI to write faster?” and starts asking “how do I use AI to learn better, deliver better, protect better and decide better?”. The first question improves productivity. The second changes the model.

That difference separates adoption from transformation. A firm may have licenses, training, internal policy and enthusiasm, and still operate through lost emails, disorganized folders, duplicated files and partner memory. It may have innovation on LinkedIn and operations stuck in the last century. It may have a modern discourse and a workflow that still depends on someone remembering where the final version was saved.

The question the market will ask is simple. What changed in your business model after AI? If the answer is only “now we use a tool to review texts,” maybe the problem is not technological. Maybe it is a lack of ambition. Using AI to do a little faster what was already being done before is incremental improvement. Important, of course. But insufficient to sustain a relevant competitive advantage.

The opportunity is elsewhere. Turning knowledge into a system. Experience into product. Data into decision. Risk into intelligence. Relationships into learning. Legal delivery into a value platform. This requires more than a “good prompt.” It requires leadership, method, governance, architecture and the courage to touch what makes money today, but may destroy value tomorrow.

The article talks about a “revolution in their business models.” That expression matters. We are not talking about replacing one tool with another. We are talking about reorganizing the value chain of professional services. Whoever controls data better controls intelligence better. Whoever controls knowledge better controls delivery better. Whoever controls workflow better controls margin better. Whoever controls the experience better controls the client relationship better.

This is the point where the discussion stops being about technology and becomes about strategy. AI is not an issue for “the IT people to look at.” Technology is part of the answer, but it is not the center of the question. The question is how the firm wants to compete in the coming years. How will it deliver value? How will it price? How will it protect data? How will it train talent? How will it capture knowledge? How will it transform its accumulated experience into real advantage for the client?

In practice, this requires a deep change in the way law firms think about knowledge management. For a long time, knowledge management was treated as a library, a repository, an organized folder or an intranet that almost nobody uses. Now it needs to become the engine of the business. The knowledge base stops being a dead archive and becomes living infrastructure for delivery, risk, relationships and the development of new offerings.

This also changes talent development. If AI takes over part of the repetition, young professionals need to learn earlier how to think better. Ask better questions. Review with more judgment. Understand the client’s business. Separate real risk from noise. Build reasoning, not just produce volume. Maybe this bothers those who were trained in a model where suffering through repetitive tasks was almost a rite of passage… lol

Leadership also changes. It is not enough to ask someone to “look into this AI topic.” That phrase, by the way, should be retired. AI is not a side project, a technological curiosity or an experimental initiative isolated from strategy. If it affects delivery, knowledge, risk, pricing, productivity, training and client relationships, it needs to be at the decision-making table. Otherwise, it becomes an orphan tool. And orphan tools usually become cost, risk or toys.

In Brazil’s Tax Reform, for example, this will become very clear. The client will not only need a legal opinion. The client will need data, tax classification, system parameterization, integration with systems, document reading, impact analysis, credit management, governance and execution capacity. Whoever thinks this will be solved only with a traditional legal opinion is looking at a new problem through an old lens.

In this type of context, AI is delivery infrastructure. But it only works for those who have a model. Without architecture, reliable sources, security, curation, integration and human review, AI becomes a layer of individual productivity, not a competitive advantage. It helps in the day to day, but it does not transform the business. And perhaps that is the greatest illusion of the moment. Confusing use with transformation.

We also need to talk about risk. In law, “almost right” can be a huge problem. A poorly contextualized answer can become a tax assessment, a contingency, reputational damage or a bad business decision. That is why applying AI in a legal environment requires more than enthusiasm. It requires access control, data protection, traceability, reliable sources, quality criteria, human review and clear accountability. That is not bureaucracy. It is part of the value proposition.

AI does not threaten what is truly sophisticated. It threatens what was sold as sophisticated without being so. It threatens rework treated as care. It threatens delay treated as depth. It threatens lack of process treated as customization. It threatens long delivery treated as robustness. It threatens well-paid improvisation, which may be one of the most resistant institutions in the Brazilian market.

In the end, maybe the right question is not whether AI can threaten law firms and accounting firms. The right question is which firms are so afraid of what AI might reveal. Because maybe the problem was never artificial intelligence. Maybe the problem is discovering that part of what we called sophistication was just an expensive, manual and poorly scalable way of operating.

And I make this provocation with genuine concern, not with pleasure in pointing out flaws. I believe in the legal profession. I believe in the role of the lawyer as someone who organizes complexity, protects interests, reduces risks, builds solutions and helps people and companies make better decisions. Precisely because I believe in this, I worry when I see part of the market treating AI as a “trend,” an abstract threat or a productivity tool, when it has already started to move the economic foundations of the profession.

The Brazilian legal profession has technical depth, talent and relevance. But it needs to look inward with more courage. Organize knowledge. Review processes. Train people for another kind of delivery. Build governance. Apply technology responsibly. And, above all, stop confusing tradition with “immunity.”

Everyone will have access to AI. Tools will change, become cheaper and increasingly similar. The differentiator will be the ability to integrate technology with intellectual capital, the delivery model, risk management, the client experience and the firm’s strategy. In other words, the advantage will not be in having AI. It will be in knowing what to do with it.

Everyone has AI. Almost no one has a model.

And when things become truly serious, it will not be enough to say that the firm was following the trends… Until the day that becomes a “confession of delay.”