Earlier this month, Y Combinator released an 11-minute Startup School video called “How to Build an AI-Native Services Company.” The speaker, YC visiting partner Charlie Warren, opens with a prediction: some of the biggest companies of the next decade won’t be software businesses at all. They’ll be services companies, insurance carriers and law firms “rebuilt from scratch with AI doing most of the work.”
If that sounds familiar, it should. In March, Sequoia made the same case: services are the larger spend, and funded AI-native firms are going after that revenue. Sequoia named the prize. YC is now handing the instructions to anyone who wants them.
The video deserves a close read from anyone running a professional services firm, not because it’s hostile but because it’s specific. Warren describes exactly which work the startups are being sent after, and, without meaning to, exactly which work they’re being told to leave alone.
YC warns founders off work that turns on human judgment, because it can’t be scaled. That’s exactly the work your firm hands over unbilled.
Which of your services fit the target profile
Warren gives founders four traits to look for in a market.
Low trust. Not “untrustworthy.” He means work where “the work is already outsourced and the customer cares about the final product, not how they got there.” The founder is “displacing a vendor, not asking the customer to do something fundamentally different.” In his words: “You’re showing up where the budget already lives and doing the work.” The vendor in that sentence is the incumbent firm.
Low judgment at the task level. “If you can break the work into pieces and every piece needs a human exercising actual judgment, you can’t really scale.” The target is work that decomposes into automatable steps, with human judgment needed in only a few places.
A high intelligence threshold. The work has to be hard enough that it takes models plus humans to deliver an outcome the customer accepts. Easy work attracts easy competition.
Regulation as a moat. YC treats regulation as an advantage: higher expectations and legal accountability keep out casual entrants.
The markets he names: tax, audit, insurance, mortgages, parts of healthcare, parts of logistics. The first two are the bread and butter of most accounting firms.
The four traits work as a diagnostic. Your client describes some of your engagements this way: we send it out, it comes back, we review the result. The process is invisible to them, and they don’t want to see it. That work fits the profile.
The pricing lesson they’re getting on day one
The middle of the video is a commercial education, and it’s the same one most firms have spent a decade resisting. “You have to sell outcomes, not seats or tokens.” Price per unit, per return, per claim, per loan, or price on the outcome itself. Never cost-plus, which “caps your upside permanently.” Never undercut the incumbent, because cheap pricing makes the work look cheap. His summary: price on value.
He cites Panacea, a YC company selling FDA regulatory services, which prices on the completed study rather than hourly, “which is the norm in the industry.” The norm is the opening.
The economics behind it: Warren tells founders that traditional services firms top out around 30% margins, and that what he calls “AI operating leverage” can push an AI-native services company toward software margins, in services markets he describes as far larger than software.
Founders with no clients and no track record are being taught outcome pricing as a starting assumption, while it’s still a contested partner-meeting topic inside the firms they’re aiming at.
The ground founders are told to avoid
Trait two cuts the other way.
Warren warns founders off work where every step needs real human judgment, because it can’t scale. He repeats the caution later as a test of intellectual honesty: are you using humans because the work genuinely needs judgment, or because you’re papering over product gaps?
That warning is a map of the defensible ground. Diagnosis, structuring, the call on what the client should actually do, the conversation where a client trusts one specific person with a consequential decision: the playbook tells founders that work is a trap for them. It doesn’t scale, so they shouldn’t build there.
So the threat doesn’t land on your firm as a whole. It lands on specific revenue lines inside it: the ones a client sends out and checks when they come back. The judgment-heavy lines sit outside the kill zone, but only if they’re priced and sold as what they are. Judgment given away inside an hourly bundle, subsidized by the commodity work the startups are now funded to take, is defensible ground earning commodity rates.
The hardest line in the video
The closing section argues founders shouldn’t buy an existing services firm and modernize it. Warren’s reasoning: “You just can’t acquire product-market fit.” A legacy business carries legacy expectations on metrics, hiring, and performance, and bolting AI on top doesn’t change any of them.
He’s talking to founders about acquisitions. Heard from inside the firm, it’s a colder claim: the partners already there face the same metrics, the same hiring model, and the same performance expectations that would defeat an acquirer.
Take that seriously, because half of it is right. Adding AI on top of the current business changes nothing. Faster delivery of the same offerings, billed the same way, is invisible in the P&L.
The half that’s wrong is the assumption that AI is the only variable. The thing a legacy firm can change, and a founder acquiring one usually can’t, is what it sells: new offerings, scoped and priced on outcomes, built on judgment the firm already owns. Warren’s entire pricing section assumes someone gets to make that decision from scratch. Incumbents get to make it too. Most haven’t.
The race runs in both directions
Warren says the best founders in this category need three things: domain fluency, model fluency, and operational rigor. On the first he’s blunt about the stakes: you’re selling to skeptical buyers in regulated markets, so you have to “bleed credibility.” Direct experience is best, he says, but “learned is actually okay.”
That last phrase is a large concession. Warren is telling founders they can acquire, over years, the standing your partners spent careers earning. He’s right that they can. What he doesn’t say is that the trade runs the other way. Model fluency and a commercial design built around it are also learnable, and no one has to spend a career earning them.
Set Warren’s curriculum against the Three Shifts that separate professional services firms gaining margin from the ones losing it. Two of the three are about the work itself: pricing on outcomes instead of hours, and packaging offerings instead of building each one bespoke. His founders get taught both. The third shift is about the relationship, moving from reactive service a client calls for to proactive counsel a client leans on. The playbook never mentions it, because his founders don’t start with the client relationship that makes it possible.
YC’s playbook covers two of the Three Shifts.
Selling outcomes instead of hours
In the playbookCodifying delivery instead of rebuilding it for every client
In the playbookStaying engaged and surfacing problems before the client calls
Not in the playbook
The video’s only real blind spot is assuming the incumbents won’t run. The Big Four are already running. The open question is the mid-market.
Run your firm through their filter
List your service lines. Mark the ones a client would describe as outsourced and judged on output, where the process is invisible and the deliverable gets checked on arrival. Those fit YC’s profile. Expect margin pressure on them, and decide now whether to reprice them, productize them, or shrink your dependence on them.
Then mark the lines where every engagement turns on judgment a client trusts you specifically to exercise. YC tells its founders to stay away from that work. It’s yours to keep, on one condition: that it’s named, scoped, and priced as the offering it is, instead of riding along unbilled with the work that’s now in play.
A founder with no clients and no track record has already spent eleven minutes with that video and written down your service lines. The list they made is the list you can make. Theirs is finished.