Nous Research isn’t being coy about where the money is going. The company has confirmed a $1.5 billion valuation and said the new capital will fund its push into the enterprise sector with a product called “Hermes for Businesses” — customized AI agents that handle multi-step workflows while keeping company data private and secure. That’s the pitch, stated plainly, and it tells you almost everything about what’s happening in AI agents right now.
My reaction? The interesting part isn’t the valuation. It’s that word “private.”
What actually got funded here
Let’s get the numbers out of the way first, because they’re the part everyone screenshots. Nous Research closed a $90 million Series B led by Robot Ventures, with participation from Nvidia, Union Square Ventures, Menlo Ventures, Samsung, and 1789 Capital, where Donald Trump Jr. is a partner. That brings total funding for the three-year-old startup to $158 million at a $1.5 billion valuation.
An investor list like that is worth reading as a signal rather than a guest list. Nvidia invests in companies that will buy and showcase its chips. Samsung invests in things it might one day ship inside products. Union Square and Menlo are classic software bets. When a single round pulls in a chipmaker, a consumer electronics giant, and traditional venture firms at once, it usually means the company sits at a point where several different industries expect to collide.
Translating “multi-step workflows” into English
If you’re not technical, “AI agents that handle multi-step workflows” can sound like noise. So here’s the plain version.
A chatbot answers a question. An agent completes a task. The difference is the number of steps it takes on its own before coming back to you.
- Chatbot: “Here’s a draft email to that client about the late invoice.”
- Agent: Checks which invoices are overdue, pulls the right contact for each, drafts the emails, sends them, logs the follow-up, and flags the three accounts that need a human phone call.
Each of those steps is small. Chaining them together without a person clicking “approve” between each one is the hard part — and it’s the part businesses are now paying real money to get right. “Customized” matters too. A generic agent doesn’t know that your company calls customers “members,” that approvals route through a specific person, or that anything over a certain dollar amount needs a second signature. A customized agent is taught your particular way of doing things.
Why the privacy angle is the real story
Nous Research is leading with data privacy and security, and that’s a deliberate choice. Talk to anyone in a regulated industry — healthcare, finance, legal, insurance — and you’ll hear the same objection. They don’t doubt AI can do the work. They doubt they’re allowed to send their data somewhere to get it done.
Patient records, client files, salary data, unreleased product plans. For a lot of organizations, the blocker on AI agents was never capability. It was the legal and compliance review that happens before anyone is permitted to try. A company that positions privacy as a feature rather than a footnote is telling you exactly which conversation it wants to walk into.
The company is also known for building Hermes as an open-source agent, which is part of why this enterprise move is notable. Open-source and enterprise sales aren’t natural partners in most people’s heads. Open-source means anyone can inspect and run the thing. Enterprise means a contract, a support line, and someone accountable when it breaks. Nous Research is trying to be both, and that’s a genuinely tricky balance to hold.
What I’d watch next
Reported figures suggest the company reached roughly $36 million in annualized revenue by mid-September, with around $100 million expected by year-end. If that trajectory holds, it says something useful about the broader market: businesses aren’t just experimenting with agents anymore, they’re putting them on a budget line.
For non-technical readers trying to decide whether any of this touches your job, I’d skip the valuation talk and ask three questions instead.
- Which repetitive, multi-step process in your week has clear rules and a clear finish line? That’s agent-shaped work.
- Where does your data live, and who is allowed to see it? That determines which tools you can even consider.
- What happens when the agent gets it wrong? A good deployment answers this before launch, not after.
Money raised is a bet on the future. Work completed is evidence. Nous Research now has a solid amount of the first and a product designed to generate the second. The next few quarters will show whether businesses hand over the keys to workflows they currently guard carefully — and whether “private by design” turns out to be the feature that finally gets agents past the compliance desk.
🕒 Published: