Braintrust raises Series A from A16Z to assess and improve AI tool performance

AI evaluation startup Braintrust secures significant funding: Braintrust, a company specializing in AI product evaluation and monitoring, has raised $36 million in a Series A round led by Andreessen Horowitz (a16z), valuing the startup at approximately $150 million.

Key details of the funding:

• The round was led by a16z partner Martin Casado
• Cloud leaders Databricks and Datadog also participated in the funding
• The investment values Braintrust at about $150 million, according to sources

Braintrust’s core offering:

• The startup provides software to evaluate and monitor AI product performance
• It helps companies identify and fix issues when AI tools produce nonsensical answers
• Braintrust’s software development kit (SDK) integrates with a company’s existing IT infrastructure

Impact on AI product accuracy:

• Companies using Braintrust typically see their AI products’ self-reported accuracy improve from below 40% to over 80% within weeks
• The tool allows businesses to gauge performance improvements from tweaks such as using customized prompts or switching between different AI models

Notable clients and adoption:

• Braintrust is used by several unicorn companies, including Airtable, Brex, Instacart, and Stripe
• The startup has dozens of customers, with the number doubling in the past three months
• Clients typically pay tens of thousands of dollars, sometimes exceeding $100,000 for Braintrust’s services

Founder’s background:

• Ankur Goyal, Braintrust’s founder, previously built AI infrastructure at Impira and Figma
• Goyal’s experience with in-house evaluations at these companies led to the creation of Braintrust
• He launched the startup in August 2023 after identifying a common need for AI evaluation tools across enterprises

Investor and industry support:

• Prominent AI research labs are power users of Braintrust
• The startup counts CEOs of Datadog, Hugging Face, Airtable, and Instacart among its investors
• OpenAI cofounder Greg Brockman and Quora CEO Adam D’Angelo are personal backers

Competitive landscape:

• Other companies like Galileo and Humanloop are also entering the AI evaluation space
• OpenAI recently announced its own basic evaluations tools
• Investors see Braintrust as providing neutral building blocks for companies looking to work with multiple AI model providers

Future outlook and potential risks:

• Braintrust is well-positioned if AI continues to be impactful and disruptive
• The biggest risk, according to Goyal, is if AI tools fail to live up to their promise
• Investors view Braintrust as part of a new systems layer needed to make AI programming work effectively

Broader implications: As AI adoption accelerates, tools like Braintrust may become increasingly crucial for companies looking to implement and optimize AI solutions effectively, potentially shaping the landscape of AI development and integration across industries.