Warren / Y Combinator Startup School — How to Build an AI-Native Services Company (2026-06-03)
Some of the biggest companies of the next decade won’t be software businesses. They’ll be services companies like insurance carriers, law firms, and tax practices rebuilt from scratch with AI doing most of the work.
In this episode of Startup School, YC Visiting Partner Charlie Warren walks through the playbook for building AI native services companies, covering how to pick a market with the right traits, why variance kills these businesses faster than anything else, and the P&L math that’ll transform your business model.
(Channel description, Y Combinator Startup School, 3 June 2026.)
A ~11:23 short-form Startup School talk delivered by Charlie Warren — YC Visiting Partner — published on the Y Combinator YouTube channel on 3 June 2026. The wiki’s first YC source explicitly framing AI-native services companies as a distinct go-to-market category, alongside the existing nine YC sources which all sit on the AI-software product side of the line (Caldwell / Akhtar / Momentic / Luminai / GStack / Hu / Replit / Garg / Koomen). With Warren, the YC source-count rises to ten, and the cluster now spans two AI-native-company vertical curricula — Hu on AI-software products from the ground up, Warren on AI-services companies from the ground up. The pieces are explicitly complementary; both reject AI-as-feature in favour of AI-as-foundation, but they apply that thesis to different vendor categories.
The piece is the vendor-side architectural mirror of [[2026-05-05-nishar-nohria-end-of-one-size-fits-all|Nishar & Nohria’s Buy Outcomes model]] (HBR, May 2026), which framed the same phenomenon from the enterprise-buyer firm-boundary angle. Where Nishar-Nohria say “firms should consider buying outcomes for non-differentiating territory”, Warren says “here is how to build the company that sells those outcomes.” The two sources together form the wiki’s first complete buyer-side architecture + vendor-side playbook pair on outcome-as-a-service.
TL;DR
Eleven substantive contributions.
1. The services-rebuilt-from-scratch-with-AI-doing-most-of-the-work core thesis. “Some of the biggest companies of the next decade won’t be software businesses at all. They’ll be services companies like insurance carriers and law firms rebuilt from scratch with AI doing most of the work. These are what we call AI native service companies. And the markets are trillions of dollars in size. Tax, audit, insurance, law, parts of healthcare, and so forth. This opportunity didn’t exist even a couple years ago. But advances in the models have unlocked this new type of business.” Warren explicitly contrasts AI-native services companies with the prior YC default category of AI-software companies — “these companies also look and feel different than most startups today.”
2. The outcome-as-product vs co-pilot-as-product distinction. “Companies provide the outcome to the customer versus build a co-pilot that the customer uses internally.” The wiki’s clearest single-sentence YC-altitude framing of the two-vertical AI-native-company split: outcome-vendors (this video’s topic) vs co-pilot-vendors (the prior YC default). Maps directly onto Nishar-Nohria’s Buy Outcomes (outcomes vendor-side) vs the Build/Compose/Collaborate models (co-pilot or component vendor-side).
3. Four market-selection traits unique to AI-native services. Warren names four traits that make a market a known good fit:
| Trait | What it means | Why it matters |
|---|---|---|
| Low trust | The work is already outsourced; the customer cares about the final product, not how they got there | ”You’re displacing a vendor, not asking the customer to do something fundamentally different. That’s a huge deal because you’re not changing behavior. You’re showing up where the budget already lives and doing the work.” |
| Low judgment at the task level | Break the work into pieces; most steps automatable; judgment focused in a few places where humans stay in the loop | If every piece needs a human exercising actual judgment, you can’t scale |
| High intelligence threshold | The overall work has to be hard | Hard enough that models + humans are needed to deliver an outcome the customer accepts (“this sounds contradictory, but actually it isn’t”) |
| Regulation could actually be good | Regulated industries have higher expectations and legal accountability | ”That raises the bar and the moat for founders” — regulation is a founder ally, not an obstacle |
Warren cites Panacea (YC company providing FDA regulatory services for biotech and medtech) as the regulation-as-moat worked example — “They hire experienced FDA consultants, pair them with an AI platform to deliver faster, higher quality FDA approvals.”
4. Markets YC Likes Right Now: tax / audit / insurance / mortgages / parts of healthcare / parts of logistics. Warren names the YC-internal known-good-fit list explicitly. He immediately adds the counter-prescription: “there are plenty more markets nobody has touched yet. Don’t hold yourself to the obvious ones or what people talk about on X.”
5. The Sam Altman Test. “Will the models disrupt these businesses? It depends on what I call the Sam Altman test. You should ask yourself — as the models get better, does your service get stronger or does the model itself commoditize you? You want to be in the first camp.” The vendor-side test for model-disruption-resistance. Convergent with the agent-harness discipline’s model-rented-harness-owned framing applied to services-company strategy. Warren’s adjacent advice: “Anything involving equipment and on-site labor. The software margin math doesn’t apply when you own and operate physical things. It’s very hard to create real leverage, though these can be really good businesses. Let’s leave this area to the robotics founders.” This is the wiki’s first explicit YC-altitude segmentation of AI-services-vendors from robotics-vendors — they share thematic adjacency but the unit economics are structurally different. Plus the honesty check: “Are you using humans because the work genuinely needs judgment, or are you compensating for product gaps? Be honest here so you’re not papering over product shortcomings with actual humans.”
6. The three founder attributes: (a) Domain fluency — “Direct experience is best, but learned is actually okay. You’re selling to skeptical buyers and often regulated spaces. You have to bleed credibility. How you acquire it matters somewhat less.” (b) Model fluency — “You need to know what frontier models can do today and design the product to ride the curve as they get better. There is no substitute for great tech here. People underestimate this.” (c) Operational rigor — “Variance, throughput, cycle times, SOPs. This is not an exciting set of words for most founders, but you are fundamentally running an operation. You have to learn that skill set and you have to enjoy it or at least you have to respect it. The product is an operation.” The wiki’s clearest single-source articulation of operational-rigor-as-a-founder-attribute — convergent with [[2026-05-28-moon-mckinsey-rewiring-software-delivery-for-the-agentic-era|McKinsey’s humans supervise and improve the system that produces artifacts]] but applied at the founder-team-selection layer.
7. Two worked YC examples:
- Panacea — FDA regulatory services for biotech and medtech. The pairing model: experienced FDA consultants + AI platform → faster, higher-quality FDA approvals. Cited as the regulation-as-moat worked example and again later as the outcome-based pricing worked example (“prices on the completed consultant study versus hourly, which is the norm in the industry”).
- The General Legal Team — AI-native law firm. Founders’ background: actual law firm experience at Cooley and Fenwick + years of technical leadership at Casetext. “Most importantly, they think deeply about throughput and how they staff their firm. They’ve integrated shift work into how they serve clients to reduce cycle times and attract the best lawyers on the team. This is a win-win for scale.” The shift-work-for-cycle-time-and-talent-attraction operational pattern is the wiki’s first named instance of staffing-model-innovation as an AI-services-company competitive advantage.
8. Building the product — humans-as-the-interface, the-product-as-the-operation. “With AI services, the setup is the opposite of most software. The human is the interface of the customer, not the product. The product helps the human scale their work nonlinearly. That changes pretty much everything around building the actual product.” Four implications Warren names:
- Operations mindset. “Find the bottlenecks and build for the bottlenecks. Throughput and cycle time are now product metrics. Track them like you would daily active users.”
- Variance is the existential problem. “By variance, I mean non-uniform outputs from your actual service. Customers will fire you for variance faster than they will fire you for being a bit slower or a bit more expensive than the incumbents. They need to trust the output. Inconsistency destroys trust, which causes churn.” The wiki’s clearest single-source statement of variance-not-cost-not-speed-as-the-customer-firing-trigger in AI-services. This is the operational analogue of Osmani’s ratchet-don’t-brainstorm AGENTS.md discipline applied at the service-delivery layer.
- Humans-in-the-loop should scale nonlinearly. “If revenue scales just in line with the number of humans you add, you’ll have major problems. The humans in the loop also need to enjoy the software. They are your users.” Convergent with automation-vs-augmentation’s augmentation-pole-requires-role-redesign-not-headcount-reduction finding at the vendor-product-design layer.
- OK-to-not-scale-at-first. “It’s okay to do things that don’t scale at the very beginning, but eventually you really do need to scale. Automating the process is the product.”
9. Sales and customer success — the early demand trap + pilot-is-the-product. “The biggest challenge facing founders here is what I’ll call the early demand trap. It’s easy to sign up a lot of pilot customers when you’re just starting out and have nothing. But it can quickly overwhelm your ability to serve them and you won’t be able to build the product to scale. You’ll be stuck using humans. It is a literal trap.” Warren’s advice: “cap your first pilot customers to a small handful. Resist the temptation to sign too many too quickly.” Plus the sell-outcomes-not-seats-or-tokens prescription: “The pilot is the product. For the first handful of customers, don’t try and standardize too early. Use those pilots to learn. Find the spots where AI gives you unique leverage versus spots where you’re just automating something obvious. Build the product accordingly and do it fast.”
10. Pricing playbook. Warren’s frame: “Pricing is harder than traditional software because you’re not competing with other software providers. You’re competing directly with the cost of labor internal or outsourced.” Two GOOD options + two AVOID options:
| Pattern | Recommendation | Why |
|---|---|---|
| Per-unit pricing | GO (per return / per claim / per loan) | “The cleanest. The easiest to explain.” |
| Outcome-based pricing | GO (Panacea: priced on completed consultant study) | “Aligns incentives beautifully, but can be harder to forecast in your business.” |
| Cost-plus pricing | AVOID | ”Caps your upside permanently. Don’t do it.” |
| Straight-line undercutting | AVOID | ”Makes your work seem cheap and potentially low quality. Price on value.” |
11. P&L math + the don’t-buy-your-way-in rule. Warren’s COGS taxonomy: model costs + hosting costs + humans in the loop — “All three of them need a number, a trend line, and someone who owns them. Be deeply suspicious of zero margin or negative margin pilots. They’re fine to learn from, but it’s really dangerous to get hooked on those.” The AI operating leverage thesis: “the more the product is built, the lower the COGS, the better the gross margin.” The opportunity-size frame: “Traditional services firms top out around 30% margins. Pure software and agent companies have more margin, but often smaller TAMs. The bet on these services companies is that AI operating leverage gets you closer to software margin, say 50%+, on a market that’s two to three times bigger than software. You don’t need to be there right away. But the trajectory has to be believable.”
The closing don’t-buy-your-way-in rule: “There’s a temptation we’ve seen, especially among founders with some operating background, to try and buy an existing services business, add some AI on top, short circuit the revenue. This is generally a trap. There’s one decent reason to do it — you need a regulatory moat fast, insurance licensing for example, but otherwise this almost never works. You just can’t acquire a product market fit. Legacy service businesses are legacy. They have different expectations on metrics, hiring, and performance. Adding AI on top of that doesn’t immediately change any of those realities. Building is almost always better than buying.”
The closing recap line — “focus on the process as the product and the product as the process” — is the YC-altitude crystallisation of the entire playbook in nine words.
Linked entities and concepts
Already-promoted entities referenced: Y Combinator (publisher; source-count bump +1 — now 10 YC sources in the wiki), Sam Altman (referenced via the Sam Altman Test framing — body mention only, no new edge required).
Concept pages this source informs (Process step 6 targets): enterprise-ai-adoption (the Buy-Outcomes-vendor-side-playbook enriches the firm-boundary lens; the services-as-AI-target market list), automation-vs-augmentation (the humans-as-the-interface, humans-scale-nonlinearly pattern + operations-mindset-throughput-as-product-metric prescription), agent-harness (the Sam Altman Test + model fluency / ride the curve framing namechecks the harness layer’s model-rented-harness-owned discipline at YC altitude), micro-productivity-trap (the variance-is-the-existential-problem + automating-the-process-is-the-product anti-patterns to shallow agent deployment).
Dangling first-mentions (deferred per the second-source promotion rule):
- Charlie Warren (the speaker; YC Visiting Partner; first appearance — entire content of this source is his playbook thinking, deferred per the same precedent set with Benedict Evans and Carla Peron)
- Panacea (YC company, FDA regulatory services; named twice as a worked example — regulation-as-moat and outcome-based pricing)
- The General Legal Team (YC AI-native law firm; named as a worked example with the Cooley + Fenwick + Casetext founder-team composition)
- Cooley (US law firm; Warren cites it as a founder-credibility anchor for The General Legal Team)
- Fenwick (US law firm; same context as Cooley)
- Casetext (legal-tech firm acquired by Thomson Reuters in 2023; named as the technical-leadership anchor of The General Legal Team’s founders)
W&W cells touched (9 cells — among the broader cells-coverage for an 11-minute YC video):
digital-sensing/digital-scouting— Warren names the YC-internal known-good-fit market list (tax / audit / insurance / mortgages / parts of healthcare / parts of logistics) as a published scouting output.digital-sensing/digital-scenario-planning— the four-market-traits framework + the Sam Altman Test + the robotics-segmented-out + the don’t-buy-your-way-in rule are all scenario-planning artefacts.digital-seizing/balancing-digital-portfolios— the build-vs-buy decision-rule + the cost-plus-vs-value-pricing decision-rule.digital-seizing/strategic-agility— the early demand trap + cap-your-first-pilots-to-a-small-handful discipline are pacing-of-strategic-response prescriptions.digital-seizing/rapid-prototyping— “the pilot is the product” is the rapid-prototyping cell stated as a slogan.digital-transforming/redesigning-internal-structures— humans-as-the-interface, the-product-is-an-operation, throughput-as-product-metric, humans-scale-nonlinearly + the General Legal Team shift-work worked example.digital-transforming/improving-digital-maturity— the three founder attributes (domain + model + operational fluency).strategic-renewal/business-model— outcome vs co-pilot + 30%-services-margins → 50%+-AI-leveraged + TAMs-2-to-3×-bigger-than-software is the business-model-renewal claim.contextual/external-triggers— “this opportunity didn’t exist even a couple years ago. Advances in the models have unlocked this new type of business.”