Verso：AI原生公司如何实现全自主运营

Hello everyone. I'm Lydia, the co-founder and CEO of Verso, which I'm building together with my co-founder Camille. If there are any founders in the room, you know that when you're building a company, you don't get much sleep. At Verso, though, there's someone who sleeps even less than I do—and that's Verso itself. We've built an AI-native company that has been almost fully autonomous since day one. Today, I'd like to explain how we made that happen, especially with the help of Codex, and, more importantly, what that changes for us in our day-to-day work.

Simply put, at Verso we're context providers. We give our clients the right consumer context so they can make the right business decisions—decisions that can ultimately have a major financial impact. We provide near real-time access to what their consumers and users think, how they make decisions, and why they make them. That enables both our clients and their AI agents to create products, messaging, and visuals that truly resonate with their consumers.

In practice, we've built an agentic solution that orchestrates the entire consumer research process from end to end—from designing a study, to recruiting respondents, to an AI interviewer that moderates conversations with participants, followed by a complete analysis pipeline that delivers actionable insights to our clients so they can make better decisions. The impact for our customers has been tremendous. Thanks to Verso, they get consumer insights ten times faster and at half the cost of the traditional approach.

That gives them a real competitive advantage in their markets. Our product is powered by OpenAI from end to end. For the interview stage, we use the Voice models, of course. We also use the Realtime Translation model that Romain showed earlier—we started using it as soon as it was released. For the analysis stage, we've built a sophisticated multi-agent pipeline based on the latest frontier models. It's an incredibly powerful product, and we're very proud of it.

But that's not really what I wanted to talk about today. What I really want to share is how we've built a fully agentic, autonomous organization—and how Verso is able to operate with an incredibly lean team given our revenue and growth. Everything in our company revolves around what we call the Company Brain—the Verso Brain. As you know, for any system to be effective and autonomous, it needs to do three things: Listen, Think, and Act.

Listen: our Company Brain listens to everything happening around it by collecting internal data—from emails, calendars, Slack, and Notion—as well as external data. All of this is orchestrated through Codex automations. Think: our Company Brain has two core components—intelligence and memory. The intelligence layer is our Knowledge Agent. It indexes incoming information and retrieves the most relevant context for each use case.

It's powered by Cortex. It also has memory—a flexible database managed by the Knowledge Agent that stores all of Verso's information and organizational context. And finally comes Act, which is probably the most important part. Our Brain can take action in the real world through tools, MCPs, and CLIs. It executes workflows autonomously. To make this more concrete, let me show you how the Verso Brain manages one of our most critical workflows: delivering a consumer research study to a client.

To give you an idea, with the volume of business we handle today, a traditional company would probably need between ten and fifteen Customer Success Managers. At Verso, we have none. Traditionally, delivering a consumer research study takes about one week for scoping, three weeks for fieldwork, and another three weeks for analysis. It's a slow, people-intensive, and expensive process. We've completely reinvented it. As I mentioned, the Verso Brain has access to all of our clients' knowledge, all of Verso's methodology, and it can directly interact with the Verso application.

That's because we designed our product from day one to be fully accessible and controllable by AI agents. As a result, the Verso Brain can autonomously perform almost everything a Customer Success Manager would normally do. It only pings us when it needs human input or flags something that deserves our attention. Let me give you a concrete example. This is our Slack workspace, where we collaborate. Last Friday at 5:00 p.m., we wrapped up a client meeting.

Maxime, who works with us, simply tagged the Verso Brain in the client's Slack channel and asked it to retrieve the information from the meeting we had just finished, along with our email exchanges. An hour and a half later, at 6:23 p.m., after a few iterations, the fieldwork was already underway. Over the weekend, around fifty consumers in the United States interacted with our AI interviewer and completed the study.

By Sunday at 2:00 p.m., the fieldwork was complete. We arrived at the office on Monday morning. The Verso Brain had flagged a few interviews for us to review because they appeared fraudulent. We validated them, launched the analysis, and just two hours later the study had been delivered to the client. We've gone from a world where clients waited six weeks for answers to one where they receive consumer insights in just seventy-two hours.

To show you what that looks like in practice, these are the kinds of studies we deliver to our clients at Verso, complete with all the insights. Because the interviews are conducted by video, clients can actually see the consumers, hear what they said, and make better decisions based on those insights. What made all of this possible comes down to two things. First, the Company Brain I described earlier—it's well structured and equipped with the right tools.

Building something like this is far from easy. Second, designing a product that, from day one, was built to be used by and interact naturally with AI agents. To finish, here's one more quick example from engineering. Today, software engineers at most tech companies spend about twenty percent of their time fixing bugs. At Verso, every bug is automatically detected, triaged by AI agents, and ninety percent of bugs are fixed automatically.

As a result, our engineers spend almost all of their time building the product, improving the technology, and creating value for our customers. I could share many more examples, but the key takeaway is this: we launched Verso not that long ago, we now serve around thirty clients, and we've delivered more than fifty studies across many different languages. And we've done all of that with a team of just four people—something that simply wouldn't have been possible in the old world.

I wish I had more time to tell you how we've also automated eighty percent of our recruiting process, or how Verso sends me a daily WhatsApp digest every morning that I read over breakfast. But I'll save that for another time. Thank you!