AI-powered meeting plugin Read AI emerges from stealth with $10M

Read AI, a startup developing an AI-powered platform for meeting metrics, today emerged from stealth with $10 million in seed funding from Madrona Venture Group with participation from PSL Ventures and individual investors. With the funding, cofounder and CEO David Shim says that Read will be able to make its service available for free on Zoom and via Zoom’s app marketplace, as well as through a Google Calendar integration.

Video calls have become a larger part of daily life compared with before the pandemic. For example, in the past 12 months, Zoom has seen a 3,000% increase in meeting minutes, and the top four video conferencing platforms have crossed over 500 million daily active participants combined.

But virtual meetings aren’t perfect. For example, when broadcasting to a large group of people, presenters can usually only see thumbnails of a few participants off to the side of their deck, requiring them to spend time scrolling to see participants’ reactions. Beyond this, when it comes to recurring meetings, priorities can shift over time, making the same hour-long sync no longer a good use of the workday for some attendees.

“We’ve all been in that meeting, where we know it is about to go downhill, yet we hesitate on taking any action as we might have the wrong read. The current feedback mechanism is an emoji or a message in chat that more often distracts than improves the conversation,” Shim, previously CEO at Foursquare, told VentureBeat via email. “Read is ‘feedback 2.0,’ where in real time, all meeting attendees have a shared dashboard that encourages collaboration by validating intuition with data.”

AI-powered meetings

Read was founded in 2020 by Shim, who teamed up with longtime colleagues Rob Williams and Elliott Waldron to build the prototype. The three previously launched location analytics company Placed, which was acquired by Snapchat in 2017 and spun out into Foursquare in 2019.

Read leverages AI, computer vision, and natural language processing models to power its meeting analytics backend. The models, which Shim says were trained on an “internationally and demographically diverse set of training examples,” deliver real-time sentiment, engagement and participation metrics encouraging collaboration among attendees, and aggregate metrics across meetings.

A host of startups plug into meeting platforms to analyze the content with AI. For example, Fireflies recognizes questions, assigned tasks, major topics, and general sentiment in meetings. Kronologic is developing what it describes as “calendar monetization” tools, also powered by AI. As for Read, Williams says it can be “a second set of eyes and ears,” seeing an entire audience in a meeting and providing aggregate reactions in real time.

“Read provides that longitudinal awareness and can help … [identify] where engagement and sentiment drop-off, allowing [companies] to adjust the content or attendee list improving both operational efficiency as well as morale,” Williams said via email. “In an enterprise sales setting, Read augments an account executive’s ability to read the room. [The product] provides real-time feedback on if the pitch is resonating, if the audience is interested in the new features, and if the pricing is well received. This type of feedback is not only beneficial to the executive, but to the prospect as it allows the seller to make adjustments to deliver a more productive meeting in real time.”

While ostensibly benign, Read’s tracking capabilities might not sit well with all employees, many of whom have had to contend with increased workplace surveillance in recent years. A 2018 Gartner survey found that 17% of organizations are monitoring work computer usage, while 16% are tracking Microsoft Outlook and calendar appointments. Moreover, even though 62% of executives say that their companies are using new technologies to collect data on employees, fewer than a third say that they feel confident that they’re using the data responsibly, according to a recent survey by Accenture.

There’s also the potential for AI bias to skew the meeting metrics. While Shim insists that Read’s models were trained on a sufficiently diverse dataset, even tech giants like Google and Facebook have historically deployed flawed models into production. For example, Zoom’s virtual backgrounds and Twitter’s automatic photo-cropping tool have been shown to disfavor people with darker-colored skin. And some experts argue that accurately assessing a person’s sentiment by analyzing their face is premised on shaky evidence.

But Shim insists that Read’s platform is “built on the foundation of privacy and transparency.” In meetings, Read notifies attendees that it’s listening in on the conversation and provides access to meeting metrics for all attendees. Read also offers an opt-out function to participants in the meeting and doesn’t provide meeting playback in the form of recorded or transcribed conversations.

“The future of meeting analytics isn’t recordings and transcripts. Rather, it’s real time, understanding how the meeting is going and what are the moments that make or break an interaction,” Shim added.

Read anticipates doubling the size of its team to over 30 people in the next six months.