Venture capital shifts focus to autonomous agent system fleets
Investors are prioritizing companies that build systems to solve classes of problems rather than single-agent solutions. This approach allows for autonomous execution of product plans and rapid integration of third-party tools.
Andy Treidman of Theory Ventures argues that the primary shift in AI development is moving from using agents to solve individual problems to building systems that solve whole classes of problems. He cites Artemis Security, a New York-based AI-native security data platform, as a key example. Artemis replaces the Security Information and Management platform and builds agents on top to handle 50 to 100 tool integrations, such as CrowdStrike and Splunk. Instead of asking agents to build each integration individually, the company builds a system to develop and test integrations, allowing agents to execute on new requirements.
Treidman notes that companies with agent-native workflows complete what looks like an annual product plan in three months. He highlights Reese Sullivan’s Executor project, which uses agents to submit performance-improving pull requests every six hours. He also points to Jen Spark, where agents automatically process customer requests into development queues, often merging pull requests without human review. Treidman emphasizes that these systems require verifiable objectives or quantitative performance indicators to function effectively.
The venture capitalist argues that building systems to encapsulate product judgment allows agents to execute autonomously and reliably. This creates a compounding advantage for companies that adopt this approach, as the systems become more effective over time by managing the complexity of multi-tool environments.