Key Takeaways

  • Vercel scaled to a $600M ARR run rate, with 50% of new business initiated autonomously by AI agents, jumping from 3% earlier in the year.
  • Jason Lemkin builds software 8 to 10 hours a day alongside 21 autonomous agents that make their own infrastructure choices.
  • Software selection has shifted from human relationship funnels to automated codebase inspections that prioritize clean APIs and uptime over brand perception.
  • Dev Ittycheria points out that failing to feed structured documentation to agent indexes is the exact modern equivalent of being invisible on page one of Google.

Software Buyers No Longer Have Faces

For two decades, enterprise software playbooks relied on the same predictable motion: hire SDRs, run paid search ads, optimize landing pages for human eyes, and wine-and-dine VP-level decision-makers. Vercel's recent numbers show that model is cracking. Harry Stebbings pointed out the math directly: “Vercel at 600 million in ARR with agents being 50% of the new business up from 3% at the start of the year.”

That jump did not happen because marketing teams got better at writing copy. It happened because software engineers stopped picking their own hosting, database, and deployment tools. Instead, autonomous coding agents run in the background, identify technical dependencies, select the vendor, and initiate the account setup without waiting for a human signature.

Lemkin runs this exact workflow daily. “When agents pick you, and I'm sure Dev's thinking a lot about this, right? On the database side, when agents pick you, it's a force of nature right now,” Lemkin explained. “I'm literally building 8 to 10 hours a day. And man, my agents have opinions and they're really hard to argue with.”

Why Agents Do Not Care About Your Brand

Human buyers can be seduced by slick keynote decks, brand trust, and vendor dinners. Agents do not care about any of that. An agent reads code, tests endpoints, and checks latency. It does not look at your pricing page testimonials or your case studies.

As Lemkin noted: “An agent knows your whole stack. It knows your application. It knows what you've built on top of the system. So it will make a very specific recommendation what to use.”

If your API documentation is locked behind gated PDFs, your SDK has unhandled errors, or your setup takes twelve manual steps in a dashboard, an agent will skip your product entirely. It cannot recommend what it cannot execute programmatically in two seconds. Ittycheria, who runs MongoDB, framed this as an existential problem for go-to-market teams:

“For the last 25 years, every marketing person, every company has optimized how you show up on page one of Google, right? You educate human buyers. Now, you have to make sure that agents can find you, understand what you sell, have current, accurate information about you across the sources they read. And if you're missing from an AI generated answer, it's like being invisible on Google.”

Ittycheria added that the burden falls squarely on product teams: “It's actually incumbent upon the startup to figure out how they basically educate agents on what they do, the documentation, the APIs and all the infrastructure and the wrapper so that they can then show that they are truly a first class experience for an agent to make a decision on.”

What to Do With This

Run an agent audit on your core product this week. Open Cursor or Claude Code, point it at a blank project, and prompt the agent to install, authenticate, and deploy your service using only your public documentation. If the agent stumbles on vague error messages, requires a manual human confirmation click, or defaults to your competitor instead, rewrite your developer docs into machine-readable markdown files tomorrow.