Unified API Recall pulls conversational data from Zoom and Google Meet to power AI bots

Join us in returning to NYC on June 5th to collaborate with executive leaders in exploring comprehensive methods for auditing AI models regarding bias, performance, and ethical compliance across diverse organizations. Find out how you can attend here.

Enterprises bullish on AI are racing to combine top large language models (LLMs) in the market with their internal business data. The approach works, but most teams often miss out on utilizing the biggest data goldmine: conversations. Today, Recall, a startup helping enterprises mobilize their data asset, announced it has raised $10 million in funding.

The investment comes from Ridge Ventures with participation from Industry Ventures, Y Combinator, IrregEx, Bungalow Capital, Hack VC and other existing backers. Recall’s co-founder David Gu told VentureBeat the company plans to use this capital to build its product – a unified API for different platforms – and add more members to the team, gearing up for the next phase of growth. Over the last twelve months, the company has grown tenfold.

“Conversations are the world’s largest dataset. Large language models continue to unlock conversations in exciting ways, and the demand for developers to capture this data has never been higher. Every SaaS company in the world should be using conversations as a data source. Recall’s unified API makes it as easy as possible,” he said in a statement.

Mobilizing conversational data for AI

In the age of remote work, teams have millions of hours of meetings across video conferencing platforms like Google Meet and Zoom. These interactions happen daily and have a lot of intelligence about day-to-day business ops, covering aspects like sales, marketing and technology. However, with each platform being used, this information – spanning audio, video and chat – gets locked in a silo. To access this conversational data at scale, teams are forced to build in-house infrastructure and integrations – which takes a lot of time.

VB Event

The AI Impact Tour: The AI Audit

Join us as we return to NYC on June 5th to engage with top executive leaders, delving into strategies for auditing AI models to ensure fairness, optimal performance, and ethical compliance across diverse organizations. Secure your attendance for this exclusive invite-only event.

Request an invite

“It takes a year or more of engineering time to build the infrastructure and integrations in-house in the most basic form. Once it’s built, companies face another challenge – maintaining the infrastructure they built. Hosting the infrastructure requires hundreds to thousands of servers to handle the processing and a team of engineers to monitor, scale, and maintain everything,” Gu told VentureBeat.

Gu and fellow co-founder Amanda Zhu were initially building a product to transcribe video meetings from different platforms in real time. However, they soon saw the infrastructure challenge and realized that accessing data from video conferencing platforms was a bigger problem than transcription (which was just one application). So, in 2022, they redirected their engineering efforts from the initial product to launch Recall.

At the core, Recall provides a unified API, backed by associated infrastructure, to integrate with all video conferencing platforms an organization has in use and pull conversational data from them. It covers a range of video tools in the market, including Google Meet and Zoom, and takes away the entire hassle of building dedicated infrastructure just for accessing meetings’ data.

“Users simply make a single API call to Recall to pull data from meetings, such as meeting recordings (both audio and video), transcripts, metadata and more. From there, users can take the raw data they get from Recall and pipe it into an LLM like GPT-4 to summarize and do further analysis on,” Gu told VentureBeat. They can plug the API to build applications – like meeting copilots and bots – that ingest and analyze data from video conferences.

Gu noted that the API takes only a few days to ingest data and start powering downstream applications. Plus, it comes with strict data policies, storing ingested audio and video recordings for a maximum of seven days. After that, the information is automatically deleted.

“We also provide an API endpoint for our users to delete the data immediately at any time, if they want to minimize the duration we store the data for,” the co-founder added.

300 companies on board

Since its launch, Recall has roped in more than 300 companies from different industries, ingesting millions of hours of meeting data and powering downstream AI applications. This includes use cases like live note-taking during telehealth sessions, analyzing virtual depositions and powering conversational intelligence by providing sellers with live guidance during sales calls.

“For a CTO, this technology means 5+ engineers freed up to focus on higher impact work. For a VP of Product, this means shipping a new feature in 2 months instead of a year, which is critical in fast-moving fields like AI technology,” Gu noted. He did not share the exact revenue numbers of the company but confirmed it has grown from zero to several million dollars in two years with a straightforward usage-based model (charging per hour of audio and video processed through the API). 

As the next step, the co-founder said the company plans to build its product, adding new tools and integrations for more conversational data sources and delivering more value to customers. It also plans to grow the team across key domains.

Source link

About The Author

Scroll to Top