Recall.ai Raises $38M Series B to Power the AI Stack for Conversation Data

Recall.ai, the platform providing comprehensive infrastructure for AI products built on human conversations, announced today the successful close of its $38 million Series B at a $250 million valuation. The round was led by Bessemer Venture Partners, with additional backing from HubSpot Ventures, Salesforce Ventures, Ridge Ventures, RTP Ventures, Y Combinator, and angel investors including Paul Graham, Solomon Hykes, Michael Siebel, and Eoghan McCabe.

Building the AI Foundations for Real Conversations

Recall.ai delivers a powerful, unified API that enables developers to access recordings, transcripts, and rich metadata without building their own integrations from scratch. This infrastructure powers more than 1,000 conversation intelligence products and processes billions of minutes of conversation each year.

Its suite of products includes the Meeting Bot API, a Desktop Recording SDK, and a Mobile Recording SDK, supporting capture of conversations across video meetings, desktop apps, phones, and even in-person interactions. By integrating directly with major platforms like Zoom, Google Meet, Microsoft Teams, and Slack, Recall.ai reduces what would normally be a months-long engineering effort into a matter of days.

The company has built its backend to operate at massive scale. Recall.ai processes over three terabytes of raw video every second and launches more than 8 million EC2 instances monthly to keep pace with demand. Developers gain access to transcripts and metadata within 10 seconds of a meeting ending, regardless of its length, enabling AI-driven applications to act almost immediately.

Unlocking Conversation as an AI Dataset

Much of the world’s business knowledge and human interaction doesn’t live in documents or spreadsheets—it’s spoken in conversations. For AI systems to become truly useful, they need access to this context.

“Conversation data is the world’s largest untapped dataset,” said David Gu, co-founder and CEO of Recall.ai. “To fill out a CRM, AI needs to know what the customer actually said. To write a follow-up email, it needs to know what was discussed. To generate a clinical note, it needs to know exactly what the patient said.”

Advances in large language models over the past two years have made it possible to take unstructured speech and turn it into actionable insights at scale. By providing a standardized way to capture this data, Recall.ai allows developers to focus on building features such as meeting summaries, coaching assistants, or compliance tools—without managing the infrastructure headaches that typically come with real-time voice and video processing.

Accelerating Enterprise Adoption

Companies such as HubSpot, ClickUp, and Apollo.io already use Recall.ai to launch new conversation-aware features significantly faster than before. Development teams report achieving a 2–3x improvement in time-to-market compared with attempting to build similar functionality in-house.

“Recall.ai allows us to build AI-powered meeting recording features without needing to worry about infrastructure or platform-specific edge cases,” said Jared Williams, EVP Head of Engineering at HubSpot. “It has helped us move faster than we could have with an in-house build.”

The adoption extends beyond sales and productivity software. Recruiting platforms, legal technology companies, and healthcare AI providers have also turned to Recall.ai as a backbone for their conversation-driven products. For healthcare developers in particular, the ability to capture patient speech accurately and securely provides a foundation for building clinical documentation tools and AI medical scribes.

The Road Ahead: Implications for AI

The $38 million raise underscores a bigger transformation taking place in artificial intelligence: the move from static, text-based inputs to dynamic, speech-driven context. Until now, most AI systems were designed to parse documents, emails, or database entries. But the majority of decisions, negotiations, and exchanges of knowledge happen in conversation.

Reliable infrastructure for capturing and structuring speech could reshape multiple industries. In healthcare, real-time transcription and summarization could reduce physician burnout by handling documentation automatically. In customer service, systems that truly understand tone and nuance could anticipate needs rather than react to tickets. In the workplace, platforms could build a memory of decisions and discussions that improves collaboration across distributed teams.

At the same time, the expansion of conversation data raises new questions about privacy, security, and consent. Organizations adopting these tools will need to balance the promise of AI-driven productivity with safeguards that ensure data is handled responsibly. The debate over how to govern this new layer of AI infrastructure is likely to become as important as the technology itself.

While Recall.ai is one of the companies pushing this space forward, the broader story is about AI’s evolution into systems that can interpret nuance, tone, and intent—not just words on a page. As this shift accelerates, the ability of intelligent systems to truly “listen” may become the defining factor in how seamlessly AI integrates into daily work and life.