Neo: securing AI agents at the endpoint
Bessemer Venture Partners leads Neo Security's $100M Series A to build AI security for the endpoint, giving enterprises real visibility and control over their agents.
In less than two years, coding agents and workplace automation tools on the endpoint have gone from being an edge case to a regular process inside enterprises. While the existing security tools on the endpoint can identify what processes are running, they have no way of seeing what these agents are doing, and have limited visibility into extensions and plugins, which are increasingly the primary interface through which agents operate. As a result, most endpoint agent activity remains unmonitored, let alone stopped when something goes wrong.
Gartner projects 40% of enterprise applications will embed AI agents by the end of this year, and CyberArk found that 68% of enterprises have no identity controls in place for the agents they've already deployed. This problem will only continue to scale, and we've already seen the conversation around it change in real time from within Bessemer’s own CISO advisory group. As recently as mid-2025, agents were hardly a main topic of conversation. By early 2026, agent security quickly became the single most pressing concern raised by cybersecurity leaders.
Neo is addressing this by building an endpoint security platform designed for AI agents from the ground up rather than adapting from a previous architecture. We’re leading Neo Security’s $100M Series A to partner closely with CEO Nick Warner and the team because we believe agent security, and the endpoint specifically, is one of the most important new categories in cybersecurity today.
Built for agents, not retrofitted for them
The rapid adoption of agentic software across the enterprise has created a major security challenge for CISOs. SecOps teams are rushing to gain insight into what's running in their organizations and how they can control it while minimizing friction to the business. Neo was built to address this challenge, and the Neo platform answers a CISO’s three main questions:
- What agentic software are employees using, and should it be allowed?
- What is the security posture of the software discovered, and is it configured correctly?
- How can I enforce guardrails and implement effective controls?
Neo's approach to setting policy stood out to us the most. Rather than asking a security team to define every rule upfront, the system observes traffic first and proposes a policy based on what it sees. Neo's built-in LLM also enables security teams to use plain language to develop, extend, and apply policies, eliminating the need for manual curation of every enforcement rule. This dramatically accelerates time to value and alleviates one of the biggest headaches for teams used to writing extensive lists of deterministic rules, especially when creating policy for agentic behavior is still novel.
Most of the activity in agent security so far has focused on agentless, API based visibility. Neo made a different (and harder) bet, building an actual sensor that lives on the endpoint, because it’s the only vantage point you can truly intercept and govern what an agent does in real time. It's a hard product to build and an even harder one to deploy, but it’s the only path to true enforcement rather than a report of what already happened.
The team behind the conviction
Neo's founding team pairs elite go-to-market experience with technical depth, which is a rarer combination in cybersecurity. CEO Nick Warner spent years as CRO and later President and COO of SentinelOne, where he helped take the company from $5 million to over $500 million in ARR through its IPO. CPO Shlomi Salem is an alum of Israel's Unit 8200 who spent 10 years at SentinelOne, rising from engineer to Vice President of Research. CTO Eran Shirazi previously co-founded and served as CTO of EasySend, which raised $72 million, and led an R&D team at Unit 8200 earlier in his career.
The ripple effect
The stakes go beyond any one company. As agents take on more of the work inside the enterprise, from writing code to managing workflows to interacting with customers, the question of whether an action was taken by a human, an agent, or a script becomes a foundational one for every business built on modern software. Getting this layer right will allow organizations to adopt agentic AI at the pace they want without opening up entirely new categories of risk. Our bet is on Neo to build the platform that makes that possible, and we're thrilled to be backing Nick, Shlomi, Eran, and the rest of the Neo team as they do.
