Neo raises $100M to govern AI agents inside your SaaS

Ask most operators how many AI agents are running in their business and you'll get a number based on what they deployed on purpose. That number is wrong. The AI agents worth worrying about are the ones your vendors shipped into products you already pay for — your browser, your CRM, your help desk, your IDE — each one arriving with permissions nobody scoped. A Boston startup just raised $100 million on that exact gap.

What actually happened

On July 20, Neo came out of stealth with $100 million across seed and Series A financing from Andreessen Horowitz and Bessemer Venture Partners, with Craft Ventures and Merlin Ventures joining. The founding team is not first-timers: CEO Nick Warner was president and COO at SentinelOne through its IPO, CPO Shlomi Salem led detection engineering there, and CTO Eran Shirazi co-founded EasySend.

The product is a control layer, not another chatbot firewall. Per SecurityWeek, Neo keeps a live catalog of the AI agents, models, extensions, and MCP servers active in an environment, flags over-broad permissions and bad configurations, attributes each action back to a specific user or app, and enforces policy at the level of individual tool calls and API requests — including pausing an operation for review. Warner's framing of the problem is the honest part: agentic capability is being embedded into browsers, developer tools, SaaS platforms, and ordinary business apps, with permission to act on its own. Neo's announcement cites Gartner projecting that agentic features go from 5% of enterprise applications in 2025 to 40% by the end of 2026.

Why it matters for your business

You will not opt into most of this. Your vendors will ship it, enabled, in a Tuesday release note. A 12-person company running Google Workspace, a CRM, a support inbox, and two dev tools now has agentic software touching customer records without a single procurement decision behind it. That's the shadow AI problem, and it's structurally different from the agents you built and scoped yourself.

You don't need a $100M platform to get ahead of it. Do the boring version. Write down every SaaS product you pay for and check what AI features shipped in the last two quarters and what data they read. Turn off the ones with no owner. For the ones you keep, check the OAuth scopes and the connected-app list in each admin console — that's where an "assistant" quietly gets write access to your entire Drive. Then decide, on purpose, which systems an agent may write to and which are read-only forever. The governance question isn't whether AI agents are in your business. They are. It's whether you can name them.

Sources: Neo via GlobeNewswire, SecurityWeek.