Recall.ai lands $38M to unlock spoken data at scale – News

Recall.ai, a startup that provides backend infrastructure for artificial intelligence that understands human conversations, has closed a $38 million Series B funding round that brings its total funding to $51 million and values the company at $250 million.

The firm’s core product is a unified application programming interface that abstracts the complexity of building and scaling meeting integrations across Zoom, Google Meet, Microsoft Teams, Webex, GoToMeeting and Slack Huddles.

Recall.ai, which is the business name of Hyperdoc California Inc., plans to use the new capital to expand its agentic AI integrations, storage and playback infrastructure, and to broaden its support for new types of conversation data beyond meetings.

While large language models have made human conversation more accessible to AI, the infrastructure needed to ingest and process that data reliably at scale is a large and complex task.

“It’s a highly complex technical domain because you’re doing large-scale audio and video processing, which is very computationally intensive, so you need to write very high-performance and efficient code,” said David Gu, the company’s 25-year-old cofounder and chief executive. “Otherwise, you’re burning a lot of money on compute.”

Recall.ai is betting that more companies will want to add conversation-native features—like automated follow-ups, sales enrichment or clinical documentation—without building the plumbing themselves.

“There are five times more words spoken at work in a single year than all the data on the internet,” Gu said, citing internal estimates. The company says more than 400 billion hours of work conversation happen globally each year, almost none of which is captured.

Labor-saver

The API eliminates the need to manually spin up bots and handle each meeting platform individually, allowing for deployment in days instead of months. Recall.ai said development teams consistently report two to three times faster time to market using its API compared to building meeting recording capabilities in-house. Costs are lower because of Recall.ai’s economies of scale.

“We process over three terabytes per second of raw video at our peak load every day, and we launch over eight million EC2 instances monthly,” Gu said. “We have very optimized services to get the maximum out of the servers we run.”

The company also offers a desktop recording software development kit and is building support for dialers and offline conversations. Recall.ai said it has over 1,500 customers – including HubSpot Inc., Datadog Inc. and Instacart Inc. – and processes millions of meetings monthly.

The company’s customers are primarily software firms that want to add audio/video capture and transcription to their products. “We’re a part of their infrastructure stack in the same way they might use AWS or OpenAI,” he said.

The company is now pushing into phone and in-person conversations, which Gu says account for about 70% of workplace communication. A new Desktop Recording SDK allows applications to capture spoken audio locally. Integration with voice over IP and cell phone platforms is on the roadmap.

Recall.ai isn’t profitable but is on a $10 million-plus annual run rate and is more interested in growth than profitability at the moment. “We’ve grown revenue more than three times in the last 12 months,” Gu said. “If we didn’t hire more people, we would be cash flow positive very soon, but we are hiring more.”

Gu, whose previous startup was a live transcription service, sees the infrastructure challenge as both a burden and an opportunity. “It’s a highly complex technical domain,” he said. “But we’re solving the hard problem so others don’t have to.”

The funding was led by Bessemer Venture Partners LP. Other investors include Salesforce Ventures LLC, HubSpot Ventures, RTP Global UK Ltd., Audion Management Co. LLC d/b/a Ridge Ventures and Y Combinator ES20 LLC, along with angel investors.

Photo: Unsplash