How Braintrust is Becoming the Observability Layer for Production AI

Braintrust, the AI observability platform helping teams measure, evaluate, and improve AI in production, led by Ankur Goyal and the team, has raised $80M in Series B funding led by ICONIQ, with participation from prior investors Andreessen Horowitz, Greylock, Elad Gil, Basecase Capital, and others.

Like many founders, building Braintrust came from a personal pain point. While at Impira and Figma, internal tools had to be created just to run evals. That repeated experience showed the problem was widespread, so conversations with peers shaped a user-first approach that still guides the product today.

Since the founding of Braintrust, AI has shifted from experimentation to production. Agents are now embedded across engineering workflows, powered by models from frontier labs and open-source challengers. Non-technical teammates are vibe-coding apps in their spare time, while enterprises and startups alike ship AI features that simply have to work, reliably, and at scale.

Despite AI’s growing criticality, teams often have less confidence than ever in what might fail next, and when something breaks, it’s harder to explain why. Traces now span long-running, multi-step agents with tool calls and intermediate reasoning, generating hundreds of megabytes of data per interaction.

Traditional observability tools can’t keep up with that scale. At Braintrust, this strain was felt firsthand, ultimately leading to the development of a custom database purpose-built to handle the complexity of AI observability.

As customers scale AI in production in 2026, observability must be treated as core infrastructure. AI behaves like a constantly evolving operating system, often beyond direct human inspection, and teams need clarity, accountability, and confidence with every update.

AI powers mission-critical products used by major companies, which is why organizations like Notion, Replit, Cloudflare, Ramp, and Dropbox rely on Braintrust.

Customer feedback drives the product, and constant conversations ensure it genuinely improves workflows and outcomes.

Matt Jacobson, General Partner at ICONIQ Capital, reinforces customer focus. Conversations consistently center on what customers are building and how best to support them, a philosophy shared by companies that scale successfully alongside their users.

“ICONIQ has observed that a defining trait among generational companies is deep, authentic customer obsession. Ankur and the Braintrust team embody this mindset, building the product from the start to serve evolving customer needs. The strong validation from leading AI teams using Braintrust reflects the depth of that commitment.”

— Matt Jacobson, ICONIQ Capital

“With this new funding, the company is expanding engineering and GTM teams, opening new offices, and preparing new product launches to be unveiled at the Trace user conference next week, while keeping customer focus at the core.”