The second frontier: AI native services companies

For twenty years the SaaS playbook has worked, and it still does. Build a tool, charge a subscription, let the customer do the work. It has minted real businesses and it continues to power excellent companies, several of which we are proud to back. None of that is going away. But alongside it, a new category is opening up, and it deserves attention on its own terms.

Y Combinator gave it a name: the AI native services company. The thesis is simple. Many of the largest companies of the coming decade will be services businesses such as insurance, tax, audit, legal, claims and back office operations, rebuilt from the ground up with AI doing most of the work. The reason is arithmetic. Global spend on services is several times larger than spend on software. A SaaS startup competes for a slice of the IT budget. An AI native services company goes after the labor budget, a market two to three times bigger.

The competitive map is where it gets interesting, and the contrast is one of degree, not of good versus bad. A SaaS company competes against other software companies: sharp, fast, well capitalized, shipping constantly. That is a demanding game, and the best players win it every day. An AI native services company competes instead against the firms that do the work today, namely agencies, BPOs, brokerages and regional operators. These run on people, carry thin margins, grow slowly, and have little software DNA. They cannot match your unit economics once AI carries 60 to 70 percent of the operational load. You are not selling them a better tool. You are absorbing the line item itself.

That asymmetry is the opportunity. A large share of these services is already outsourced, which means the buyer has already decided not to do the work internally. Replacing an outsourced vendor is a much shorter sale than changing behavior. The go to market writes itself: same outcome, fraction of the cost, no implementation project.

There is also a second act that pure software rarely gets. When your AI delivers the service at software like margins, the traditional players become attractive, accretive acquisition targets. You buy the book of business, migrate the work onto your stack, and expand the margin. The rollup becomes a wedge rather than a fallback.

The catch is that the operation is the product. YC is clear that the winning founders combine domain fluency, model fluency and operational rigor. You are measured on throughput, cycle time and how little your output varies. Trust is earned by results.

The takeaway for builders and backers is not to abandon software. It is to recognize a second frontier sitting in industries that never had software at their core, run on manual labor, and have margins waiting to be rewritten.