Taking law as the example, I'm not sure I agree.
Legal is largely a distress purchase. Very few clients actually want to spend time interacting with outside corporate counsel, a divorce lawyer, a personal injury lawyer, etc.
The real value of an attorney is accountability. A person with a license and reputation who is on the hook for the advice they provide. But accountability is only worth buying if the judgment behind it is sound, and that judgment takes years to build.
The problem is that much of what makes a "good" lawyer comes from years of doing unglamorous work as a junior, financed by clients. If AI reduces the ability of firms to subsidize training the next generation of lawyers, you get fewer "good" lawyers and the whole market keeps eroding.
From what I've seen, the top AI models can now produce contracts that are on par with what I would have had to pay thousands of dollars for a decade ago, and the agents/chatbots are capable of collaborating with you to address fairly complex issues. If you have the money, you still run everything by a real attorney of course for the accountability piece, but the future is not bright for contract attorneys because people know even they're probably using AI to do half their work.
That is what AI will do, and firms will figure out how to adapt. To be fair, this trend has been going on for some time now already. Clients want attorneys with experience and judgment to be on the hook in real trials, mergers, etc. There’s only so many of those folks. The trend of the last several decades has been pretty steady overall headcount at firms despite large increases in the amount of legal work, and ever increasing demand for the limited set of seasoned attorneys.
Firms have not done a great job cultivating the next generation against the backdrop of those trends. But that’s true in a lot of white collar jobs. And AI will make that problem more acute.
But—lawyers are a transaction cost and it’s good for the economy to automate as much of it as we can!
I guess it depends on how you look at it.
I don't think anyone would argue that spending money on lawyers is an ideal use of capital. But if AI-driven automation reduces overall demand for legal services (as measured by billable hours) and/or erodes the pricing power of lawyers/law firms, it doesn't mean that the "savings" will benefit all participants in the economy.
The biggest beneficiary will be corporations, as reduced legal expenditure can boost corporate profits. But that is a boon for the wealthy shareholder class, not the worker class. "Lawyer" has been one of the highest-paying white collar jobs available so the question becomes: as all of these high-paying white collar jobs become less plentiful, how do young people not born into wealth develop opportunities to get ahead economically?
How would you know? Unless you have expertise in this area, you can't really know if you have a good contract until you have to litigate it.
> you still run everything by a real attorney of course for the accountability piece
Your attorney can make better use of AI than you can. A growing time sink for lawyers (and I assume, physicians) is having to explain to clients that "yes, I understand this is what Claude told you," and "no, that's not actually how X works in practice."
I ran multiple businesses for over 20 years, and during that time had to deal with countless contracts involving commercial transactions, leases, funding events and employment. I worked closely with outside counsel, and had to deal with numerous disputes and lawsuits.
Many contracts are boilerplate to begin with, which is why firms like Cooley and Orrick provide free document libraries and generators. They know that a huge segment of the people who might retain them aren't going to pay $10,000 for an NDA or a simple commercial contract. So they give documents away as a marketing tool, hoping that it might get them more complex work in the future.
Attorneys can add value when you have a unique wrinkle and of course when there's a dispute, but, as an example, if you need an employment contract for California, you don't need to pay someone $500/hour to draft one from scratch and frankly for many contract categories, you'd probably be taking on more risk by having a bespoke contract drafted than modifying a solid, widely-used template.
> Your attorney can make better use of AI than you can. A growing time sink for lawyers (and I assume, physicians) is having to explain to clients that "yes, I understand this is what Claude told you," and "no, that's not actually how X works in practice."
You're not wrong, but I think you're misunderstanding what I'm saying.
First, as far as contracts are concerned, the leading AIs are more than capable of producing pretty decent ones for many categories because they've been trained against all the real contracts that are out there. Your "yes, I understand this is what Claude told you but..." scenario is real, but it applies less and less than attorneys want to believe, especially when the subject is mundane.
Second, the bigger issue is that AI is fundamentally hindering the development of the next generation of lawyers. It's great that senior associates and partners can use AI to be more efficient, but the economic pressures AI creates mean that it's far harder for law firms to have clients subsidize the training of the next generation of lawyers.
So the question isn't whether an experienced attorney today can make better use of AI than you can, it's whether a new attorney today will be able to become a "good" lawyer when AI is eroding their firms' ability to have clients subsidize their training.
For contracts, I was thinking of the more complex deals where a battle-tested template is unavailable. Though even where one appears to fit, it doesn't always. One of the difficulties in professional/client relations is that there are often details that one party sees as minor or completely unremarkable, and the other party sees as changing the whole frame of analysis. I haven't been too impressed with LLMs being able to identify things like this -- they have reams of knowledge, but no practical sense.
The issue about halting the training pipeline is an interesting one. In law firms, a lot of "training" of young lawyers is really glorified bates-stamping charged out at professional rates. I think we could find that the work just gets more substantive and interesting. And yes, pyramid-shaped firms do rely on this type of stuff for revenue. But if all parties can do more sophisticated work, then AI may only nudge forward an arms race without fundamentally changing much (at least for work that has adversarial backpressure).
But this is the exception, not the rule because most people are not engaging in "complex" deals that require completely bespoke contracts. The average individual or company needs roughly the same things almost every other individual or company needs and a LegalZoom template is going to be perfectly usable for the 99%, ideally with a final pass by an experienced attorney.
What you're paying for when you do have a "complex" deal is the negotiation and structuring, and that's where the experience of the attorney is all the value.
> The issue about halting the training pipeline is an interesting one. In law firms, a lot of "training" of young lawyers is really glorified bates-stamping charged out at professional rates. I think we could find that the work just gets more substantive and interesting.
You can't run before you know how to walk. I don't know how a new associate becomes the next Clarence Darrow, Johnnie Cochran, or David Boies without thorough exposure to the often-boring basics: legal research, document drafting, document review, organizing and sorting documents, sitting in on calls and depositions, etc.
If clients aren't willing to subsidize these types of activities and firms have to look to automate more to maintain their profitability in the face of cost pressures, the development model breaks down long-term.
It might be worth distinguishing between LLMs as tools for fuzzy-searching hard facts, versus using them to craft logical arguments. [0] The risk-profile and verification difficulty are rather different.
[0] Or, more-precisely, using them to craft a pattern-fitting text artifact, which hopefully maps onto a sane concept when reinterpreted by a human.
One of the things that AI has helped with is automatic breech notifications. They are largely proforma, but depending on which generation of contract they are notifying from, subtle changes are needed.
Law is mostly highly structured text, and you know what loves highly structured and documented text? LLMs.
Legal LLMs do share the same issue with coding LLMs, they tend to default to lazy/insecure. Its not the case where you as a non-lawyer can prompt it to make a "water tight" contract from scratch.
How has crypto been massively impactful?
The point I’m making is a financial bubble can burst while also massively changing the world
I'd say it's pretty impactful.
A bubble doesn’t say anything about how useful the underlying technology is. We are in an AI bubble, in the sense that the allocated resources is completely out of line with the underlying businesses. AI companies would literally need the entire world as their TAM, using their services non stop to justify the valuations and level of expenditures we are seeing. What you describe is literally the symptoms of a bubble
That’s not true. Apple alone is valued higher than the largest three AI companies combined.
You asserted that AI companies would need "the entire world as their TAM using their services non stop" to justify a few trillion in valuation. Yet Apple has a $5 trillion valuation just selling phones and computers, mostly to consumers and not businesses. Google built a $2 trillion valuation by the end of 2021 mainly on search and ads. Our economy is very large and you don't need to take over "the entire world" in order to justify a few companies with $1 trillion valuations.
Apple and alphabet are both trading at something like 10x revenue, with very high operating margins.
On the other hand you have AI vendors: OpenAI $852 valuation is ~34x revenue, with $14B losses projected for 2026. Their infra commitment through 2030 is more than $600B (that's on the low end of numbers floating around).
Just for the infra expenditure, if we assume a conservative ROIC of 5%, 8% operating margin: they need an operating profit of $30B/y, and a revenue of $375B/y. That's 100% of the projected 2030 AI infra market.
Edit: actually, their 2026 losses seem to be even higher, the $14B number is from a January leak, but more recent projections are at >$22B losses for 2025
This is what people mean when they say it's a bubble.
I wish people would stop being so dense.
Yes they were. Educators have been talking about how access to the internet has affected their students learning since forever. We were cautioned and had dedicated classes to teach us to be critical of different sources, teaching us what Wikipedia could and could not be used for. The internet was deliberately turned off for our exams.
Crypto too was seen as culturally important, papers were written by otherwise well respected authors. We were expected to read some whitepapers about crypto. Interestingly, reading those papers only made us more sure it was a scam.
You're rewriting history to make this moment seem different, it is not.