May 6, 2026 ·

YC Says Agencies Can Be Billion-Dollar Companies — But They’re Picking the Wrong Founders

YC wants technical founders to build AI-native agencies. We argue incumbents with clients, data, and domain expertise are better positioned to win.

YC’s Spring 2026 RFS asked technical founders to build AI-native agencies. We think the people best positioned to actually do it are the ones already running agencies. A founder playbook for the agency community, with a steelman of why the skeptics aren’t wrong about the old model.

AI-Native Agencies Are Already Reaching Billion-Dollar Valuations

Harvey is an “AI law firm” in a functional sense — it sells legal work product, not legal software. In March 2026 it raised $200M at an $11B valuation, reporting $190M in ARR by January.

Crosby calls itself an AI-first law firm and bills clients for executed contract work, not seats. After a June 2025 launch backed by Sequoia and a $20M Series A in October, it closed a $60M Series B in April 2026 at roughly $400M, led by Lux and Index, on the back of 400% revenue growth.

EvenUp serves personal-injury firms specifically, has raised $385M and is valued above $2B, with 20% of the top-100 US PI firms on its platform.

These three didn’t get to those numbers selling tools to law firms. They got there being law firms — or close enough that the distinction stops mattering when you read the cap table. They’re the existence proof for YC’s Spring 2026 Request for Startups: an entire category called “AI-Powered Agencies,” asking technical founders to build the same shape of company in marketing, design, and dev.

We think YC is right about the opportunity and wrong about who’ll capture it.

YC’s AI Agency Thesis: Service Revenue at Software Margins

The pitch from YC and General Partner Aaron Epstein, in one sentence: services companies trade at 1–2x revenue, software at 8–12x, and AI is finally good enough to do the work — so build a company that earns service revenue at software margins, and the multiple re-rates with it.

The math behind it is real. US professional services is a hundreds-of-billions-dollar market (the Bureau of Labor Statistics puts the sector at roughly 10% of GDP). The total knowledge-workforce spend in the US is closer to $5T against ~$230B in B2B SaaS. The arbitrage between “selling the tool” and “doing the work with the tool” is one of the largest mispricings AI has opened up. We agree the prize is billion-dollar-company-shaped.

Where we part company with YC is on who they’re asking. The RFS is written for technical founders building from scratch. The implicit assumption — never quite stated — is that incumbent agencies are too encumbered, too partner-heavy, too billable-hour-shaped to flip the model. That’s the assumption we want to argue with, because the agencies we’ve worked with for the last few years tell a different story.

Why Most Agencies Won’t Make the AI-Native Transition

Before we make the bull case for incumbents, here’s the steelman of why the skeptics think this can’t work.

Linear unit economics is sticky. Saving 30% of execution time per project doesn’t make the agency 30% more scalable. Account management, sales, senior judgment, QA — most of the gating constraint sits with people, and AI doesn’t replace those people yet. The “ten-person agency at hundred-person revenue” line is mostly aspirational right now.

Service multiples are structurally lower. Premium agencies — three-plus years of double-digit growth, low client concentration, sticky retainers — cap at 8–12x EBITDA. To hit a $1B valuation you need ~$80–100M EBITDA, which means ~$300–500M revenue at agency margins. That’s a top-fifty-globally outcome, not a venture base case.

Hourly pricing actively punishes efficiency. If AI helps your team finish a project in half the time and you bill on hours, you’ve just halved your invoice. Moving an agency off time-and-materials onto outcome pricing is a multi-year cultural change, not a quarterly memo.

Talent incentives fight automation. Senior strategists are comped on billable utilization. Asking them to feed their workflows into an agent is asking them to deflate their own bonus. Retrofitting comp is harder than retrofitting tools.

Client concentration tanks valuations. Agencies with a top client at 20–30% of revenue eat 10–20% valuation discounts; above 30%, deals collapse. The agencies that have the cleanest AI-leverage stories often grew that way because of one anchor client.

If you’re an agency owner reading this and feeling defensive, don’t — these are real constraints, and any honest case for going AI-native has to clear them. What’s changed is that for the first time, all five have a path through.

Why Existing Agencies Have the Advantage Over AI-Native Startups

Here’s the part YC’s RFS underweights: the YC-funded “AI law firm” or “AI design studio” starts with zero clients. A 50-person agency with 200 logos and a decade of relationships starts with revenue, retention, and reference customers. In services, distribution has beaten tech in every cycle we’ve seen — from Accenture outrunning Infosys’s early automation bets to WPP acquiring its way past digitally-native shops. We’ve watched too many beautifully-engineered AI products fail at the agency-grade sales motion to think the next wave will be different.

The other thing incumbents own that startups don’t: the proprietary data is already on your laptops. A media agency that’s run 4,000 retail campaigns has the eval set you’d need to fine-tune a creative model. A WooCommerce agency that’s done 800 store migrations knows the fifty edge cases that ChatGPT will fabricate confidently. New AI-native shops have to acquire that knowledge — incumbents have to organize it. Those are very different problems.

And the margin story works on the existing book. If you do $10M/year at 20% net with fifty people, and AI lets the same book be served by twenty-five, you’re now at 50–60% net on the same revenue base. We’ve seen agencies inside our customer base move from “AI helps a bit” to “AI is the operational substrate” inside a year. The valuation conversation that produces is fundamentally different from the one your CFO is having today.

The “agencies don’t scale” claim is mostly true at the boutique level. At scale, the model can work — though not without headwinds. Globant did $2.45B in revenue in 2025 on what is, structurally, an agency-services business (though its 1.6% YoY growth and a Q4 decline show that even scaled services companies aren’t immune to macro drag). Accenture Song became the world’s largest agency company in 2025. What’s missing is the layer between $5M agency and $2B services giant — and that’s exactly where AI-native economics open up new ground.

The AI-Native Agency Playbook: Four Moves for Agency Founders

If you’re running an agency and you read the YC RFS as an invitation rather than an indictment, here’s the playbook we’d run. None of these are clever; what’s hard is doing them in the right order.

Move 1: Pick a vertical and stop being a generalist. The proprietary-data flywheel only spins inside a domain. A generalist agency has nothing to fine-tune on, no eval set worth running, no playbook that compounds across clients. The strongest AI-native services stories we’ve studied — Harvey in big-law, EvenUp in PI, Crosby in startup contracts, Sierra in customer experience — are vertical first. If your agency does “anything for anyone in tech,” you have a positioning problem before you have an AI problem. The good news is that you probably already have a vertical you’re best at; the move is to stop apologizing for it and start owning it on the website.

Move 2: Productize at least one offering and put a fixed price on it. You don’t have to flip the whole book. Take one thing you do — a quarterly SEO audit, a WooCommerce migration, a brand identity, a contract redline pass — wrap it in a defined SOW, price it as an outcome, and fulfill it with as much agentic automation as you can stand. This is the proof point your sales team needs and the proof point your investors will want. It’s also the only way to find out which parts of your delivery are actually AI-leverageable and which parts you were romanticizing.

Move 3: Rent the platform; don’t build it. Three years ago, going AI-native meant hiring an ML team. Today, AI-native platforms exist for most verticals an agency would specialize in — that’s the whole reason YC is funding the new wave. We build one of them, for e-commerce. There are equivalents emerging in legal ops, design, content, customer support, and sales motion. The agencies that win this cycle won’t be the ones that built their own RAG pipeline; they’ll be the ones that picked the right platform stack and put their domain judgment on top of it. Building the platform yourself is a 12-month distraction that costs you the window.

Move 4: Restructure the cap table before you scale. Partnership equity is a $1B-cap. The agencies that exit big in this cycle will look like venture-backed companies on their cap table — small founder team, employee equity pool, growth capital — even if they look like services companies on the P&L. Twenty partners with profit shares is a great way to run a $20M agency forever; it’s a terrible structure for the version of you that wants to be the next Globant. Do this work early — it’s painful and it gets exponentially worse the longer you wait.

The agencies we’ve seen execute even two of these moves well are already running ahead of the AI-native startups in their vertical. The agencies executing all four are starting to look like the kind of company YC’s RFS was supposedly written for — except they got there with revenue from day one.

Why AI-Native Platforms Are the Missing Layer

The reason we’re publishing this from Urumi is that the third move — rent the platform — is the one where most agencies stall, and it’s the one we have a strong opinion on.

We built Urumi because the e-commerce agencies and merchant operators we worked with kept hitting the same wall: they wanted to deliver “AI-native” outcomes, but the platform layer underneath them — hosting, infrastructure, performance, observability, the agentic plumbing — wasn’t built for it. Every agency that tried to build it themselves ended up with a side project that ate the bandwidth they needed for client work. So we built that layer, the way ex-WooCommerce core developers and ex-Google/Meta engineers would build it, and we open it up to agencies as the foundation they put their own brand and judgment on.

That’s the relationship the AI-native platform companies and the agencies need to have for any of this to work. We do the platform; you do the vertical, the relationships, the judgment, the SOW. Neither of us alone is the billion-dollar company. Together, the math works.

Apply to the Urumi Agency Partner Program

We’re opening a small agency partner program for the agencies that want to run this playbook on top of Urumi. This isn’t a referral fee scheme — it’s revenue share on outcomes, joint architecture sessions, early access to the agentic primitives we ship, and a real commitment from us to make your delivery economics work. The only way our thesis pays off is if the agencies on top of us reach the kind of scale YC’s RFS is pointing at.

If you’re running an agency in the WooCommerce or broader e-commerce space and you’ve been reading the YC RFS thinking “they’re describing what we already do, just from a worse starting position” — that’s the right read. We’re accepting five agencies this quarter.

Apply for the agency partner program →

YC is right that this is a billion-dollar company shape. They’re just asking a narrower set of people than they need to.

FAQ

What is an AI-native agency?
An AI-native agency delivers professional services — legal, marketing, design, development — using AI as the core operational layer rather than just a productivity add-on. Instead of billing for hours, AI-native agencies productize outcomes and fulfill them with agentic automation, earning service revenue at margins closer to software companies.

Why is YC funding AI-powered agencies in 2026?
YC’s Spring 2026 Request for Startups includes a dedicated “AI-Powered Agencies” category because the economics have shifted: AI can now execute knowledge work at a quality level that clients will pay for. Companies like Harvey ($11B valuation) and EvenUp ($2B+) have proven the model in legal services. YC sees the same opportunity across marketing, design, and dev.

Can existing agencies compete with AI-native startups?
We believe existing agencies are better positioned than greenfield startups. Incumbents already have clients, revenue, domain expertise, and proprietary data from years of project delivery. AI-native startups have to acquire all of that from scratch. The key is that incumbents must restructure pricing, verticalize, and adopt AI-native platforms — not build from zero.

What does an AI-native agency playbook look like?
Four moves, in order: (1) pick a vertical and stop being a generalist, (2) productize at least one offering with fixed outcome-based pricing, (3) rent an AI-native platform instead of building your own, and (4) restructure the cap table from partnership equity to a venture-compatible structure before you try to scale.

How do agencies transition from hourly billing to outcome-based pricing?
Start with one offering — a quarterly audit, a migration, a brand package — and wrap it in a defined scope of work with a fixed price. Fulfill it with as much AI automation as possible. This gives your sales team a proof point and reveals which parts of delivery are truly AI-leverageable. Scaling outcome pricing across the full book is a multi-year cultural shift, but it starts with one product.

What is the Urumi agency partner program?
Urumi’s agency partner program is for WooCommerce and e-commerce agencies that want to run the AI-native playbook on top of our platform. It includes revenue share on outcomes, joint architecture sessions, and early access to agentic primitives. We’re accepting five agencies this quarter — apply here.

Built by ex-Automattic WooCommerce core engineers

We built WooCommerce core at Automattic — the parts that matter in production: performance, payments, reliability. Earlier we were founding-era engineers at HackerRank (Y Combinator), where the team scaled the company from $2M to $30M ARR. That's why we can operate your store end-to-end, not just host it.

Naman Malhotra - LinkedIn
Vedanshu Jain - LinkedIn

Run your store on Urumi.

Production-ready today. Built for high-traffic WooCommerce stores where downtime moves revenue.

See the WooCommerce platform · Talk to founders

Agent live · 99.99% uptime · shipping today.

Product

Company

Careers

Case Studies

Gruum

🦷 Can you believe it? This page was built by a dentist using urumi.ai