TraviaTechPie Review

The Story

Here’s a sentence that would have sounded like science fiction eighteen months ago: a company incorporated itself, opened a bank account, set up email, bought phone numbers, spun up servers, and started running — and a human barely touched any of it.

On the coordination problem itself, see how async agents actually cooperate.

That’s roughly the pitch behind Naïve, a Palo Alto startup that just closed a $28.5M Series A. The round was led by Nexus Venture Partners, with Y Combinator, Zetta, and Liquid 2 joining, plus a stack of operator angels — Gokul Rajaram, Apollo.io co-founder and chairman (former CEO) Tim Zheng, and former HubSpot COO JD Sherman among them. That brings the company’s total raised to roughly $32M. The co-founders are CEO Sean Dorje and CTO Dennis Zax.

So what does Naïve actually build? The cleanest way to put it: it’s plumbing for AI agents that want to operate a business. All the annoying, real-world scaffolding a company needs — payments, an email account, a phone number, cloud compute, storage, and the legal act of incorporating — gets packaged behind a single API. An agent calls that API, and the pieces show up. Naïve wires in the usual suspects too, hooking into services like Stripe for payments and QuickBooks for accounting.

The developer experience is the part that makes this land. You don’t sit in a dashboard clicking through fifteen SaaS signups. Instead, you hand a prompt to a coding tool you already use — Cursor, Claude Code, Codex — and that tool talks to Naïve’s API to provision everything the business needs to exist. In other words, the agent that writes your app can also stand up the company that runs it.

The traction numbers are the reason investors leaned in. Naïve says it’s signed up more than 30,000 developer customers within months of launch, and that annual run-rate revenue has grown roughly 10x — to the “low double-digit millions” — over the past six months. That’s fast. Whether it’s durable is a separate question, but the demand signal is real.

The use cases Dorje describes are where it gets genuinely strange, in a good way. People are using Naïve to run AI automation agencies, “faceless” content channels on TikTok and YouTube, and — the example that keeps getting quoted — a rental car agency. The company also ships templates for common patterns: SEO shops, SaaS apps, recruiting, accounting, customer support. There’s even a mobile emulator so agents can tap through smartphone apps on a virtual device when there’s no clean API to call.

Now, the honest asterisks. Naïve doesn’t magic away the parts of business that legally can’t be automated. KYC and KYB — the identity and business-verification checks banks and payment processors require — still land on a human. So do the actual payments. An agent can assemble the LLC and request the accounts, but a person has to sign for who they are and put real money down. That’s not a limitation Naïve can engineer around; it’s the law doing its job.

Where the new money goes tells you what Naïve thinks the hard problem is. The plan isn’t more integrations — it’s making the agents underneath run better and cost less. That means a “model router” that sends each task to the cheapest capable AI model, a memory layer so an agent remembers a business’s context across sessions, virtualized sandboxes where agents can act safely, and a governance system for budgets and human approvals. Read that list again. It’s not a list about starting companies. It’s a list about controlling autonomous software that spends money.

The Takeaway

We’ve been circling this exact spot on the blog for a while, and Naïve is what it looks like when the pieces snap together.

Go back to “Google’s Push into AI Agents” and the “agentic era” framing everyone adopted last year. The pitch back then was agents that could do things — book, buy, browse, execute. What was missing was somewhere for them to do those things safely. An agent with a credit card and no guardrails is a liability, not a product. Naïve is essentially selling the guardrails and the workspace as a bundle: here’s a sandbox, here’s a budget, here’s a memory, now go operate.

And that connects straight to the piece we ran on AI agent identity — the “who is this agent?” problem. The moment an agent is incorporating LLCs, opening Stripe accounts, and moving money, “who authorized this and who’s accountable” stops being philosophy and becomes an operational must-have. Notice that Naïve’s roadmap leads with governance, budgets, and human approvals. That’s not a coincidence. The market is discovering, in real time, that the bottleneck for agentic business isn’t capability. It’s control.

Here’s my read on why the KYC wall matters more than it looks. That human-in-the-loop step for identity and payments isn’t a temporary rough edge Naïve will polish away next quarter. It’s the seam where the fully-autonomous-company dream meets a legal and financial system built entirely around accountable humans. The interesting companies in this space won’t be the ones promising to remove the human. They’ll be the ones that make the human’s job — approve, veto, set a budget, define the rules — as small and as sharp as possible. Naïve leaning into governance rather than pretending the human can vanish is, honestly, the mature call.

There’s also a quieter signal in the “model router” line. Put it next to what we wrote about Liquid AI’s phone-sized “Nanos” models and Thinking Machines shrinking Inkling. The whole industry is converging on the same insight: you don’t run every task on the biggest, most expensive model. You route. A business-operations agent might do a hundred tiny steps a day, and paying frontier-model prices for each one would torch the margins. So Naïve building a router isn’t a side feature — it’s the difference between “autonomous business” being a demo and being a business.

One caution worth keeping. The 30,000 developers and the 10x revenue are impressive, but “developers signed up” and “companies profitably run by agents” are not the same number. A lot of that early demand is builders experimenting — spinning up a faceless YouTube channel to see if it works. The real test is retention: how many of those agent-run businesses are still standing, and still paying, a year from now. That number doesn’t exist yet. When it does, it’ll tell us far more than the funding round did.

Still, the direction is hard to argue with. For two years the conversation was about agents that could think. Naïve is a bet that the next fight is over agents that can operate — and that whoever owns the boring infrastructure underneath gets to tax the entire agentic economy. That’s a big bet. It’s also, suddenly, not a crazy one.

This article is for informational purposes only and is not investment advice.

Photo: Igor Shalyminov / Unsplash