Give Me Beer and TV!

Quentin Solt, thank you for yet another thought-provoking piece.

I spilled my beer and was rolling on the floor laughing at your outlet analogy.

Here are two (seemingly random and unrelated) facts: (1) I spent years advising telecoms on M&A deals, and (2) I have been a customer of Airalo for years.

Before you conclude that I have had a few too many beers and should be hauled off and locked in a sanitarium, this is how those two facts are relevant to your piece.

Telecom is – fundamentally – about the network. Fiber, switches, towers, frequency spectrum, lock, stock and barrel. Hard assets. Physics requires all this equipment to transmit a signal.

And yet companies providing telecom services have adopted very different business models. Some own as much of the infrastructure as they can. Others do sale-leasebacks with respect to towers and telecom equipment. Some MVNOs own little or none of the underlying network infrastructure. And some companies - surprise! - have built very substantial telecoms businesses by sitting almost entirely on top of infrastructure owned by others.

Airalo is perhaps my favorite example. It sells eSIMs that use existing operators’ networks, while competing with those same operators for the customer relationship. It owns the customer-facing product and distribution layer, but not the underlying mobile networks. And it has now become a unicorn without owning the network on which its customers’ connectivity depends.

To me, Airalo's case is a spectacular demonstration that you can build a very large business on top of someone else’s infrastructure.

That has been the state of play in telecom for decades. Regulation, changes in accounting treatment, availability and cost of capital make certain business models more or less “fashionable” for a time, and then other models emerge.

The underlying technology can therefore converge without the business models necessarily converging with it.

And that, I think, is where we disagree.

You argue that markets experiment and then tend to converge around a smaller number of dominant architectures. I agree with the first half, but not the second. Even where there are dominant architectures, there can be – and often are – multiple ways of organizing ownership, capital, risk, distribution and the customer relationship around the underlying infrastructure.

On to legal tech.

I agree that foundation models are becoming increasingly commoditised (incidentally, several of my hedge fund friends cannot wait for Anthropic's and OpenAI's IPOs to short the hell out of them for precisely this reason). I agree that intelligence will become cheaper. I also agree that clients will increasingly ask why they should pay a lawyer to perform work that technology can perform faster, cheaper and with no loss of quality.

But none of those propositions establishes that the legal profession will consequently converge on a small number of business models – still less that the companies that own the underlying intelligence will necessarily own the customer, the workflow, the regulated service, the specialist knowledge or the economics.

In fact, cheaper and more ubiquitous infrastructure may make the opposite possible: more experimentation at the layers above it.

Let me now move to the 4K/8K analogy that I regard as one of the key pieces of your argument. At first blush, it sounds incredibly convincing. But here is why I am not fully convinced.

The assumption hiding in your analogy is that technological progress principally makes an existing product better, and that once the existing customer need is satisfied, further improvements in the underlying technology have diminishing economic value.

But that is not what happened in telecom.

The original use case for a 2G network was being able to make a call without being tethered to a landline. 3G, 4G, LTE and 5G did not merely make the existing use case incrementally better. They made entirely new use cases possible, which in turn created new customer demand.

The smartphone did not become valuable because people suddenly developed a burning desire to make slightly better telephone calls. Faster and more ubiquitous networks made possible video, real-time navigation, mobile payments, streaming, ride-hailing (remember our long debate about Uber?), video calling and an extraordinary number of other things that would have been impractical or uneconomic on earlier networks.

In other words, “good enough” was not (and is not) a fixed point. It is always a function of what the technology makes possible.

Your argument asks whether 4K is good enough for the legal work that clients want done today. Perhaps it is. I am perfectly prepared to concede that for many existing legal tasks, an incremental improvement from one very capable model to an even more capable model may not cause the client to pay twice as much.

But that is not quite the same question as what happens when intelligence becomes sufficiently cheap, fast, reliable and ubiquitous to support things that were previously uneconomic.

The customer cannot necessarily tell you what the next use case is before the technology makes it possible.

And I suspect that is also where we differ on the “bigger pie” argument. If AI merely makes existing legal work cheaper, then your arithmetic is fairly compelling: fewer lawyer-hours per unit of output means fewer lawyers are required.

But if AI also changes the set of legal services that can economically be provided, the arithmetic becomes less straightforward.

A company might not pay a lawyer to review every contract every week. But it might pay for continuous monitoring of its entire contractual estate if the marginal cost of doing so becomes negligible. It might monitor regulatory developments, counterparties, litigation risk, employment relationships, IP portfolios or compliance obligations continuously rather than periodically. Individuals and smaller businesses might obtain forms of legal assistance that were previously too expensive to purchase at all.

Those are not simply “more lawyer-hours.” They are potentially different products, delivered through different business models, on different economics.

Which brings me back to the point about telecom.

I think there are actually two extrapolations in your argument that I am resisting.

The first is from technological convergence to business-model convergence. The fact that the underlying intelligence becomes commoditised does not tell us who will own the customer, the workflow, the regulated service or the economic relationship built on top of it. Airalo is a useful reminder of that.

The second is from today’s customer demand to tomorrow’s customer demand. The fact that clients regard a particular level of AI performance as “good enough” for today’s legal work tells us less about what they will want once the technology makes entirely new forms of legal service economically feasible.

So I am not arguing that lawyers are somehow immune from technological progress, or that every new increment in model capability will magically create another lawyer-shaped job.

Quite the opposite.

I suspect AI will destroy a great deal of traditional legal work. I just do not think that tells us, by itself, what the legal market will look like when the technology has finished changing what is possible.

And yes - that is what Oleg Loukhton and I invest in at Grayhair Venture Partners. We are proud to back founders that are reimagining professional services, starting with legal.