Recall.ai: Unlocking the World’s Conversations

Recall.ai provides a unified API for accessing conversation data from platforms like Zoom, Google Meet, and Microsoft Teams, serving as a fundamental infrastructure layer for over a thousand AI companies. The company operates at a massive scale—processing three terabytes of raw video per second at peak and launching eight million EC2 instances monthly—all while maintaining a lean team of fewer than 30 people and approaching $20 million in ARR.

The Pivot: From a Product to an API

The journey to becoming an infrastructure provider was born out of necessity. The initial product, launched in 2020, was a call recorder built to solve the team’s own challenges with user research. For nearly three years, they ran this product and discovered that the most critical and demanding aspect was the underlying infrastructure.

• Infrastructure as the Core Problem: 70-80% of all engineering effort was dedicated to ensuring the reliability, stability, and scalability of the recording infrastructure. The stakes were incredibly high; any downtime meant customers would lose critical, irreplaceable recordings. This experience gave the team a deep, visceral understanding of the problem’s difficulty.
• The Rise of LLMs: As Large Language Models evolved at a frightening pace, a new wave of companies emerged, all building products on top of unstructured conversation data. These companies were creating products that would have seemed like “magic” just 18 months prior.
• The Strategic Shift: The team realized that the robust, battle-hardened infrastructure they had built for themselves was precisely what every one of these new AI companies needed. In February 2022, they pivoted, productizing their internal infrastructure and selling it as the Recall.ai API. Their first customers were their old competitors, as they could immediately demonstrate the value of offloading the infrastructure burden to move faster.

Building a High-Agency, Lean Organization

Recall.ai’s remarkable revenue-per-employee is a direct result of a deliberate strategy focused on hiring a small number of exceptional individuals.

• The “Full Stack” Employee: The company culture avoids rigid roles, seeking high-agency individuals who can operate across domains. Engineers are expected to not only write low-level systems code but also talk to customers, understand product needs, and even help close enterprise deals.
• Avoiding Information Loss: This approach is designed to combat the “90% loss of information” that occurs during handoffs between different departments (e.g., from product to engineering). By having a single person own a problem from customer interaction to implementation, the company gains massive efficiency and quality.
• Hiring Former Founders: Over half of the engineering team consists of previous startup founders. The company looks for people who are motivated to push themselves, learn new skills, and see their role as an opportunity to achieve something great.

The Future is Conversation Data

The core thesis behind Recall.ai is that conversations are the most critical, untapped data set for AI.

• A Vast Data Source: David notes, “There’s five times more words spoken at work every year than all the words on the entire internet.” He argues that 99% of the context within a company exists in person-to-person conversations, not in documentation.
• Fueling Effective AI: For an AI to be truly effective, it needs access to this conversational context, just as a new employee needs to talk to colleagues to get up to speed.
• Roadmap: Recall.ai is focused on expanding the surface area of data capture with new products like a Desktop Recording SDK (for bot-less recording) and future support for phone calls and mobile. The goal is to handle the entire infrastructure stack—capture, storage, querying, and pre-processing—so that AI companies can focus on their unique value proposition.

Core Lessons for Founders

David’s journey from a 19-year-old hackathon winner to the co-founder of a major infrastructure company was defined by persistence.

• Perseverance is Key: The early fundraising efforts were a grind. The first attempt for the call recorder product took four months to raise just $160k. The seed round for Recall.ai took eight months, over 1,000 handwritten emails to other founders, and 120 investor meetings to close $2.5 million.
• Demystifying Sales: The first $2 million in revenue was pure outbound, driven by co-founder Amanda. David’s view is that sales isn’t “black magic” but a learnable skill focused on two things: 1) clearly explaining the value of your product and 2) helping the customer navigate their internal process to adopt it.
• Patience for Product-Market Fit: He advises founders to “not give up,” noting that it takes time to build the necessary skills to run a successful company. He also cautions that it takes at least a year—not a few months—to truly know if a new product is going to be successful.

• Tags:
• Startups
• LLM
• Infrastructure
• Feature
• API
• Fundraising
• AI Data
• Team Building