Judgment Labs Raises $32M to Debug AI Agents

Judgment Labs, a San Francisco startup founded by three friends in their early twenties, announced $32 million across seed and Series A rounds on May 12. Lightspeed Venture Partners led both rounds within less than half a year, with Nova Global, SV Angel, Valor Equity Partners and Dynamic also joining.

The company is focused on a problem that old benchmarks miss: agents do not simply turn one prompt into one answer. They plan, search, call tools, revise prompts and leave a long trajectory behind. Judgment stores and visualizes those production traces, then looks for repeated failure patterns and the exact step where an agent went off track.

That makes the product less like another model and more like a feedback loop for enterprise AI. If a team can see that an agent selects the wrong tool in a certain class of tasks, it can patch the workflow, replay real interactions and measure whether the fix actually lifted performance.

Sources: BusinessWire, Judgment Labs Closes $32M in Seed and Series A Funding; Pulse 2.0; Tech Startups; CocoLoop