Every startup will at some point be asked to define their vertical - the target market - with most VCs looking for a tightly defined answer. The standard thinking is that a narrow vertical provides a defensible beachhead (anchor, wedge) from which to challenge incumbents. This was a reliable strategy during what SamLessin calls the ““factory model” years of VC. Without wanting to ruffle feathers, our sense is that AI forces a rethink of the strategic logic behind verticalization. Here’s why.

First, AI makes it vastly easier to innovate within established verticals (check out Sam Altman’s most recent OpenAI keynote for scary good examples). This is a bit of a poison pill for startups because, sure, it means you can get to market at speed … but so can incumbents. And those incumbents already control the system of record. There's no competing, except on the terms dictated by an incumbent's ecosystem and/or with an acquisition strategy in mind. As Marissa Moore points out legal services are particularly exposed to this problem because of the relatively monolithic structure of the market (top 0.1% of law firms generate 50% of industry revenues; deep rooted control points; etc.), but the dynamics in other AI-impacted verticals are similar.

Second, by virtue of the pace and scale of social change enabled by AI and other tech, there's probably an increased premium on more radical, research-led forms of innovation. Radical innovation tends to involve thinking outside of established verticals or weaving a path between verticals, focusing on the gaps and/or roadblocks that are likely to define the future of X. Any beachhead here is going to be uncertain, built on shifting sands - but that's a reflection of the world we're in, and it makes more strategic sense to recognize and plan for this uncertainty than pretend otherwise.

There's something else in how AI is likely to require radical change to incumbent fee structures/business models (c.f . Clayton Christensen on the Innovator's Dilemma). In the immediate future we'll see incumbents trying to absorb the capabilities of AI within their established business models ("like a pillow", as the Italian philosopher Antonio Gramsci said of hegemonic power); don't expect the real gains of AI to be passed on to consumers (of legal services) any time soon. Put another way, you don’t pay $650m for Casetext and reduce fees. But this has to change in the longer term, and this is likely to mean existing verticals break down/consolidate around AI-first companies.

This links back to how AI capabilities shift the strategic playbook for startups. To the extent that AI empowers people and businesses to “self-serve” on a huge range of problems or tasks, the incentives for startups skew towards solving deep, firmly entrenched market failures (although it could be argued that any radical tech innovation would trigger similar dynamics). If you view “access to justice” as a metric of success the legal services industry has failed in a massive way: according to UN data 5.1 billion people lack adequate access to justice, many of them based in the global South. On more prosaic grounds, it's been estimated that SMEs under-purchase legal services by as much as 50%. That is, faced with a legal issue, half the time the SME chooses to muddle through on their own rather than engage a lawyer.

Where we've come down is (1) that AI - and particularly LLMs - have specific problems that can be addressed through better knowledge of legal rules, especially when it comes to governance and value alignment. Solving these problems for AI requires building a better bridge to legal data, and orienting our strategy around a future in which the shape of the legal market is defined by the presence of AI systems. For example, we’ll want legal rules to fit into AI/ML workflows, seamlessly integrate IoT and satellite data streams, and so on. These frontier needs will increasingly bite as AI becomes embedded in all aspects of governance and decision-making - as law already is.

What's more, (2) the challenges of integrating law and AI functionally overlap with the drivers behind restricted access to justice. Cost, speed and availability of personalized legal knowledge, opaque judicial reasoning, inadequate systemic intelligence etc. This makes it reasonably efficient to pursue both goals - better AI, better access to law - in lockstep.

Bringing this back to the question of verticals, we’ve taken the position that, for Lautonomy, the gain from focusing on any single, narrow, endangered vertical is outweighed by the gains from a broader, more holistic and future-oriented approach.

To be clear, this doesn’t mean we aren’t thinking about where we fit within legal services, or about how to position Lautonomy as a “sticky” platform..*

It’s just that our answer is more complicated than we’d like.

* To whit: partner with leading organizations whose work is heavily impacted by knowledge of rules → experiment and prove value add during an extended beta → develop a range of applied AI use cases able to solve deep, cross-cutting problems.