The legal industry has spent the past few years being told that artificial intelligence will transform how lawyers work.
Companies such as Harvey have attracted enormous valuations by promising exactly that: AI can search documents, analyse contracts, conduct research and accelerate work that traditionally consumed hours of highly paid legal labour.
But there is an uncomfortable economic contradiction hiding underneath the AI revolution:
What happens when you sell software that makes lawyers dramatically faster, while your business model still depends on selling their time?
That may be one of the biggest reasons the legal industry has been slower to change than the technology itself.
The billable hour creates a peculiar AI problem
Consider a lawyer who normally spends one hour reviewing a contract and bills the client for that hour.
Now imagine AI reduces the task to 12 minutes.
From the client's perspective, this is fantastic.
The work is completed five times faster.
But from the firm's perspective, something more complicated has happened.
If the firm still bills by the hour, the economic value of that productivity improvement largely disappears.
The lawyer has saved 48 minutes.
But those 48 minutes were also 48 minutes of potential billable revenue.
Thomson Reuters has illustrated this basic economics in its analysis of generative AI and professional services.
That creates an uncomfortable incentive.
The technology says: become faster.
The pricing model says: don't become too fast.
That contradiction is difficult to overcome simply by purchasing better software.
This is where Harvey becomes interesting
Harvey has become one of the most prominent legal-AI companies in the world.
The company raised funding at a valuation reported at around $15 billion in September 2025, reflecting investor expectations that AI can fundamentally change professional legal work.
Its technology is designed to help lawyers with research, drafting, analysis and other knowledge-intensive tasks.
The underlying proposition is straightforward:
If lawyers can produce more work in less time, firms become more productive.
But productivity only becomes economically transformative when the firm has a way to capture the value created by that productivity.
If the lawyer spends eight hours producing something that previously required sixteen, but the firm continues charging based on eight hours, the client may eventually benefit.
But the firm's revenue per matter falls.
And that is where the business-model problem begins.
The real AI revolution may therefore be about pricing
Imagine two firms.
Firm A bills hourly.
A contract review costs:
10 hours × $500 = $5,000
AI reduces the work to two hours.
The firm now earns:
2 × $500 = $1,000
The client has received a massive productivity benefit.
The law firm has potentially lost $4,000 in revenue.
Firm B charges a fixed $5,000.
It originally takes 10 hours.
AI reduces the work to two.
The client still pays $5,000.
The firm keeps the economic benefit of becoming dramatically more efficient.
Suddenly AI isn't a threat to revenue.
It becomes a margin-expansion machine.
That distinction could determine who actually wins the legal-AI revolution.
Enter Arceus
This is the model behind the launch of Arceus, which is positioning itself not simply as legal software but as an AI-powered legal-services firm.
The company's proposition is strikingly different.
Clients don't simply purchase an AI subscription and then figure out how to use it.
They send legal work to the firm.
AI prepares the output.
A licensed attorney reviews and approves it.
The client receives the finished work.
And the price is agreed upfront.
The company says clients can submit work through Slack and that contract reviews typically return within four to six hours.
That changes the economic equation.
The firm is no longer primarily selling:
“How many hours did our lawyer spend?”
It is selling:
“What legal outcome do you need, and what will it cost?”
That is a fundamentally different business.
The lawyer becomes an exception handler
This may ultimately be the most important change.
In the traditional model:
Human → research → draft → review → revise → deliver
AI-powered legal services can increasingly become:
AI → draft → human attorney reviews → deliver
The lawyer isn't necessarily disappearing.
The lawyer's role changes.
Instead of spending most of the day producing first drafts, searching thousands of documents or performing repetitive research, the lawyer increasingly becomes the quality-control layer.
That could allow one lawyer to supervise vastly more legal work.
And that is exactly where the economics become interesting.
If one attorney can supervise 20 matters instead of five, the firm doesn't necessarily need to quadruple its legal workforce to quadruple its output.
Its cost structure changes.
The strange thing is that law firms could already do this
They don't need permission from AI companies.
Alternative fee arrangements have existed for years.
Thomson Reuters data has shown that alternative fee arrangements have represented roughly one-fifth of law-firm revenue for much of the past decade.
The industry therefore already knows how to charge clients without relying exclusively on hourly billing.
The problem is incentives.
Partners whose compensation is connected to billable hours may reasonably see aggressive efficiency improvements as threatening their own economics.
Why would someone voluntarily adopt a technology that reduces the number of hours attached to their work if their compensation system rewards hours?
That is not a technology problem.
It is an organisational-design problem.
AI exposes the weakness of selling time
There is a broader lesson here.
The same problem exists across professional services.
Consultants sell hours.
Accountants sell hours.
Lawyers sell hours.
Agencies sell hours.
Engineering firms sometimes sell hours.
AI attacks precisely that unit of economic measurement.
A machine doesn't care whether something takes eight hours.
If it can complete the task in eight minutes, it will.
That means AI doesn't merely automate tasks.
It can attack the pricing architecture built around those tasks.
And that is much more disruptive.
The winners may not be the firms with the best AI
This is the part many AI discussions miss.
Everyone is focusing on model quality.
Which AI is smarter?
Which legal model hallucinates less?
Which platform has better document analysis?
Which company has the best agents?
Those questions matter.
But eventually the competitive advantage may move somewhere else.
Who has redesigned the business model around the technology?
A conventional law firm with excellent AI can still remain economically constrained if it continues selling hours.
A smaller AI-native firm with adequate technology but a radically different pricing structure could potentially offer:
- fixed prices;
- faster delivery;
- predictable costs;
- high automation;
- attorney oversight; and
- higher margins.
The technology doesn't necessarily need to be revolutionary.
The business model does.
But there is a major caveat
Flat fees aren't automatically better.
Legal work can be unpredictable.
A seemingly simple contract can contain unusual provisions.
A litigation matter can explode in complexity.
Regulatory questions can change.
Clients can provide incomplete information.
And lawyers still carry professional and ethical obligations.
That's why the human-review component of models like Arceus matters.
If AI generates the work but a licensed attorney remains responsible for reviewing it, the model is closer to AI-augmented legal services than an autonomous robot lawyer.
That distinction will matter enormously as regulators and courts work out where professional responsibility sits.
The bigger threat to traditional firms isn't AI
It is AI plus a different pricing model.
A traditional firm can buy Harvey.
It can buy another legal copilot.
It can automate document review.
It can build internal AI agents.
But if it keeps charging clients exactly as it did before, the technology may primarily make its lawyers more productive without fundamentally changing the customer experience.
An AI-native competitor can approach the problem differently.
Instead of saying:
“Our lawyer will spend six hours on this.”
It says:
“We'll review the contract for $X and return it this afternoon.”
The second proposition is much closer to how consumers buy technology.
You pay for the outcome.
Not the electricity consumed by the machine producing it.
And this is why legal AI may eventually become a pricing war
Once clients realise that routine legal work can be delivered faster, predictable pricing becomes increasingly attractive.
That creates pressure on traditional firms.
The question becomes:
Why should a client pay $5,000 for ten hours of work if another provider can deliver essentially the same outcome for $2,000 in two hours?
The answer cannot simply be:
“Because that's how lawyers have always billed.”
The profession will have to demonstrate where human judgment actually adds value.
That may push lawyers toward increasingly complex work where expertise, negotiation, courtroom strategy, regulatory judgment and accountability matter more than repetitive production.
The paradox of legal AI
This brings us back to Harvey.
The enormous investment flowing into legal AI is essentially a bet that lawyers can become much more productive.
But productivity creates economic value only when someone can capture it.
Under the hourly model, much of the value risks flowing toward the client through lower labour requirements.
Under a well-designed fixed-price model, the provider can capture some of that productivity as margin while giving the client faster, predictable service.
That is why Arceus's approach is worth watching.
Not because it has magically solved legal services.
But because it is attacking the economic contradiction at the heart of legal AI.
Legal AI keeps selling time savings to people who sell time
That may be the central irony.
For decades, law firms built sophisticated businesses around measuring professional effort in six-minute increments.
Now artificial intelligence is arriving with a simple message:
Those six-minute increments don't need to exist anymore.
The firms that merely install AI may become faster versions of the old system.
The firms that redesign pricing, staffing, workflow and accountability around AI could become something else entirely.
And that may be the real disruption coming to legal services.
AI isn't merely asking lawyers to work faster.
It is forcing the legal industry to answer a much harder question:
If your technology makes an hour of legal work worth less, why should the client continue paying for the hour instead of the result?
Comments
Post a Comment