Today, Judgment Labs, the infrastructure company helping AI-native teams turn production data into continuously improving agents, announced $32 million in combined seed and Series A funding. Lightspeed Venture Partners led both rounds, doubling down on the company less than six months after its initial investment, while Nova Global, SV Angel, Valor Equity Partners, and Dynamic also participated. Judgment’s platform is already in production at a growing list of agent-native companies where it powers the monitoring and improvement cycles behind agents every day. The company is putting the capital behind a single mission: give every team building agents the tools to make their products better with every interaction.
“Judgment is solving the hardest problem in the agent stack — how do you measure and improve something that thinks, plans, uses tools, and remembers?” said James Alcorn, Partner at Lightspeed Venture Partners. “The Judgment team has been productizing agentic evaluations long before the word ‘evals’ became popular. They have a clear technical vision, a product that agent-native startups are already standardizing on, and a market opportunity that grows every time another company puts an agent into production. We led the seed because the bet was obvious, and we led the Series A because the results have been extraordinary.”
For the past few years, building with LLMs has mostly meant building chatbots. A user sends a question, and the model sends back an answer. That paradigm is shifting. A new generation of ‘deep agents’, such as Anthropic’s Claude Code, OpenAI’s Codex, and Cognition’s Devin, don’t answer questions so much as do end-to-end tasks. They reason through open-ended problems, write and run code, browse the web, ask follow-up questions when they’re unsure, and run for minutes or hours on a single task. This trend of LLMs moving from question-answering machines to agents that autonomously execute complex white-collar work is also redefining legal, finance, and customer support.
That shift forces a rethink of how AI quality gets measured. The evaluation methods the industry inherited from the chatbot era were built for a single input and a single output — the answer has issues or it doesn’t. Deep agents often don’t have a single output, but instead produce a trajectory: a long chain of decisions, search queries, partial results, and self-corrections, any one of which can be the place where things went wrong. When a deep agent fails, the final answer often contains subtle errors, whereas the glaring faults are buried somewhere in that trajectory — the agent used the wrong search keywords, skipped a step, guessed instead of asking a clarifying question, or kept going when it should have stopped. Measuring agent quality now means looking at the entire path, recognizing the failure patterns that recur across thousands of real interactions, and fixing them at the root. Judgment Labs was built for this. It gives teams a clear view of the full trajectory an agent takes, surfaces the patterns hiding inside it, and turns every real interaction into a concrete fix teams can ship back into the product.
“We set out to build Judgment because the teams building deep agents didn’t have tools that understood what their agents were actually doing,” said Alex Shan, co-founder and CEO of Judgment Labs. “Input-output evals miss so much of where agents go wrong. Lightspeed has been the right partner from day one: they backed us when we were a handful of researchers with a thesis, and they’re doubling down now that the thesis is playing out in production.”
Judgment’s three founders — CEO Alex Shan (22), Chief Scientist Andrew Li (23), and CTO Joseph Camyre (23) — have been best friends since childhood. Andrew taught Alex natural language processing (NLP) when they were children. Alex was the first customer of Joseph’s middle school Python course. Years later, each took that early head start into the field’s most demanding rooms: Alex became an AI researcher at Stanford’s NLP group within the Stanford AI Lab, working under Professor Chris Manning; Andrew was an early research hire at TogetherAI, one of the fastest-growing training and inference startups in the industry; and Joseph built large-scale infrastructure as a systems engineer at Datadog. The company came together when the first wave of deep agents hit production, and the same failure modes the team had been studying for years started showing up.
The new funding will go primarily toward hiring AI researchers and engineers in San Francisco, and secondarily toward expanding the forward deployed engineering team that serves their burgeoning customer base.
“Our agents are in front of customers every day, and the quality bar keeps going up,” said Aqil Naeem, Chief Executive Officer at E3 Group. “We tried other tools, but none of them could automatically point toward where things failed. Judgment is in a different league; we can see exactly where our agents make mistakes, fix them, and measure the lift. It’s the difference between guessing and knowing, and it’s showing up directly in our customer experiences.”
About Judgment Labs Judgment Labs is the platform for improving agents from production data. The company’s infrastructure helps teams evaluate long reasoning traces, tool use, and memory, then turn production data into continuously better agents. Judgment Labs is headquartered in San Francisco and backed by Lightspeed Venture Partners, Nova Global, Valor, and Dynamic. Learn more at judgmentlabs.ai.
About Lightspeed Lightspeed is a global, multi-stage, venture capital firm managing over $40B in assets. Since its founding in 2000, Lightspeed has been the first investor and an early backer of some of the most innovative companies in the world including Abridge, Anthropic, Anduril, Castelion, Databricks, Glean, Mistral, Navan, Neko Health, Netskope, Thinking Machines, Reflection AI, Rubrik, Snap, Skild AI, Vinted, Wiz, and more. Learn more about the firm, team, and why we’re bullish on the potential of AI to transform the world at lsvp.com
View source version on businesswire.com: https://www.businesswire.com/news/home/20260512621556/en/
Media gallery
