Analysis
Last week in New York, investor Jake Saper of Emergence Capital organized a small summit at Crosby, an AI-native law firm, to make the case for what he's calling AINS -- AI-Native Services -- as the next major venture category, according to Newcomer. The pitch is specific: instead of selling software that helps accountants, lawyers or insurance brokers do their jobs faster, AINS startups hire those professionals directly, pair them with AI-focused engineers, and sell the finished output -- the audited books, the closed policy, the reviewed contract -- rather than a tool. I think this is a real and underappreciated shift, and I also think the market is going to over-apply it to categories where it doesn't belong before anyone sorts out which is which.
The mechanism that makes this possible is new. Before 2023-era model capability, an AI system genuinely couldn't do enough of a lawyer's or accountant's actual work to make a services-not-software business model viable -- you'd have been selling a chatbot wrapped around a human team doing the real work anyway. That's changed enough in the last two years that a small team of domain professionals plus AI engineers can plausibly do the volume of work a much larger traditional services firm would need. The pricing follows from that: AINS companies charge for outcomes because they can deliver the outcome in a fraction of the hours a traditional firm would bill, and a pure hourly or per-seat model would leave money on the table relative to what the client would actually pay for the result.
“Accounting and insurance underwriting fit that description well.”
I like this thesis best in categories with three specific properties: work that's currently billed by the hour to a human professional, work where the output is verifiable and standardized (a completed tax filing, a signed insurance policy, a passed audit), and work where the client cares far more about the outcome than about who or what produced it. Accounting and insurance underwriting fit that description well. I'm more skeptical about categories like legal work involving genuine judgment calls or client relationship management, where the professional's name and reputation are part of what's actually being sold, and an AI-augmented team selling under a new brand has to build that trust from zero.
Room for disagreement: the SaaS comparison itself might be the wrong frame entirely. SaaS won because software has near-zero marginal cost to serve an additional customer once built, which is what let it scale to enormous margins. AINS companies still employ real people doing real work on each engagement, which means their marginal cost per customer, while lower than a traditional services firm's, is nowhere close to software's economics. If that's right, AINS may end up looking more like a much more efficient, much higher-margin professional services firm than like a SaaS company at all -- a good business, potentially a very good one, but not the same multiple or growth curve investors are used to underwriting for "the next SaaS."
What I'd actually diligence before writing a check into this category: unit economics per engagement including the human labor cost, not just the AI inference cost, and whether the company's pricing power survives a competitor undercutting on price once the initial technology advantage narrows -- because if the model capability gap that made AINS possible keeps closing across labs, the moat for any individual AINS startup may be thinner than the category-level thesis suggests.
