Nearly 50 million Americans go to court each year without a lawyer. Low-income Americans are especially vulnerable, with most saying they “do not get any or enough legal help” for their major civil legal problems. The US ranks 107th out of 142 countries on the accessibility and affordability of civil justice, according to the World Justice Project, and 47th out of 47 high-income countries.
These numbers should be shocking. They aren’t, mostly because they’ve been roughly this bad for decades. Meanwhile, for the last several years the AI industry has been having its own conversation about lawyers: Can ChatGPT pass the bar exam? Can AI draft a contract as well as a junior associate? Can it replace a $1,000-an-hour partner at a white-shoe firm?
These are the wrong questions. AI’s biggest economic impact won’t come from replacing lawyers or other professionals. It will come from doing the work that’s too low-end to be worth their time.
There is an economic principle that explains why: Cost and the required performance are linked. When something is expensive, the market demands very high performance to justify the price. When it’s cheap, the performance bar drops. That seems obvious. Its implications are not.
Companies are pouring resources into improving AI’s performance. OpenAI alone expects to lose $74 billion in 2028. And that’s even though an Epoch AI analysis found that GPT-5 couldn’t recoup its own R&D costs during the four months it spent as a flagship model.
But the cost-performance link suggests the real opportunity is in the opposite direction — not better AI, but cheaper AI. Because when costs fall far enough, many users’ required performance drops to something not far from where AI already is.
Economists refer to “missing markets” when there are transactions that should be happening but don’t, because no one can profitably supply a good or service at a price that buyers are willing to pay. US professional services are full of them. Those 50 million unrepresented litigants don’t want to go without a lawyer, they just can’t afford one. The problem extends far beyond law. CPAs charge $150 to $400 an hour. Small business tax preparation runs from $500 to $2,500. Millions of small business owners and freelancers make consequential tax and financial decisions with zero professional help. They overpay, miss deductions, and structure things wrong. Not out of carelessness, but because an accountant’s bill would eat whatever an accountant found. The same dynamic plays out in everything from tutoring to immigration paperwork.
In each of these markets, the spectrum of cost exists, but it’s compressed at the top. You can hire a $2,000-an-hour lawyer or a $100-an-hour lawyer. But there are no $10-an-hour lawyers. The same is true for accountants or consultants. Below the floor price of a credentialed human, there’s nothing. And all the transactions that could happen there … don’t. That’s the missing market.
There’s no need to speculate about what happens when you fill this gap. A randomized trial in Mexico gave 432 small businesses access to subsidized consulting they couldn’t otherwise afford. Employment rose 50%, and the gains persisted five years after the program ended. In India, management consulting provided to textile factories raised productivity 17% and profitability even more, with firms opening new plants within three years. In both experiments, the main barrier wasn’t that businesses rejected expert advice. It was that they couldn’t get access to it.
These trials took place in developing economies, not among US consumers. But the underlying economics are the same: When cost blocks transactions that would benefit both sides, removing the barrier allows the missing market to appear. The context differs. The economics don’t.
AI is positioned to remove that barrier at a scale no subsidy program could match. And to do it, AI doesn’t need to be brilliant. It needs to be cheap.
The encouraging news is that cheap is exactly where AI is headed. According to the Stanford AI Index, the inference cost for a system performing at GPT-3.5’s level dropped more than 280-fold between November 2022 and October 2024. Epoch AI found that inference prices across benchmarks have been falling at a median rate of roughly 50 times per year. These are among the fastest price declines in the history of technology.
The obvious objection is that offering people a chatbot instead of a professional is a consolation prize, not progress. If the comparison were AI versus a real lawyer or accountant, that objection would land. But for most unmet needs, that’s not the relevant comparison. The 92% of unmet legal needs aren’t currently being served by anyone. The choice isn’t between AI and a professional. It’s between AI and nothing.
This reframes the AI industry’s strategic challenge. The frontier labs are burning hundreds of billions of dollars on the bet that superhuman AI is within reach. Maybe it is. But the missing markets don’t require that bet to pay off. They just need AI that’s cheap enough to serve them profitably. That’s not a speculative future. It’s a trajectory already underway.
Fifty million Americans go to court without a lawyer every year. The AI industry thinks the solution to this problem is building something smart enough to replace one. The real solution might be building something cheap enough that it doesn’t have to.
A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by Gautam Mukunda
