Meeting summaries are a hot new workplace task for generative AI

As companies grapple with how to put generative AI tech to work in the office, taking notes on meetings is emerging as a clear use case.

Salesforce said last month it would roll out a suite of tools including a summarization feature for calls and messages in Slack; Zoom recently teamed with OpenAI and Anthropic to offer similar tools for video meetings; and Microsoft now includes AI-generated recaps in its subscription tier for Teams.

Now, Cisco has joined the fray with new tools that use AI to help catch workers up on chats, meetings, and any other mentions on Webex and Security Cloud.

The features, set to debut by the end of the year, are based on an undisclosed large language foundational model and fine-tuned using proprietary data, according to Jeetu Patel, Cisco’s EVP and GM of security and collaboration. It includes a conversational interface where users can ask AI questions like, “Were there any urgent action items for me?” or “Who am I meeting with?” as demonstrated in a video provided to Tech Brew.

The tech will also be deployed in Webex Contact Center to provide its customer service agents with summaries of conversations with customers.

Patel said Cisco decided to focus on meeting summaries because it was an area where the technology could have a big impact on the way people work in the near term.

“What we wanted to do was make sure that the use case was something that would have high value rather than just be interesting for a demo but not really high-value,” Patel told Tech Brew. “[Summaries] seemed like a massive kind of 10x productivity boost for people if they missed a meeting.”

While it may seem like a more utilitarian function than some flashier demonstrations of AI, Patel said meeting summaries are, in some ways, more high-stakes than consumer applications. The team worked to train the AI so they could feel confident that it would not hallucinate—the technical term for when the language model makes things up.

“It’s one thing when you have a consumer application, and the model hallucinates for, like, 40% of the time. That is not cool in a meeting summary,” Patel said. “So you have to make sure that you’re really starting to fine-tune and train the model—the specialized piece—in a pretty thoughtful way.”

The rollout will build on some of the other ways Cisco already integrates (non-generative) AI into Webex, including instant transcription and translation, noise cancellation and voice separation, and real-time sound equalization, Patel said.

Beyond meeting summaries, Patel sees generative AI as a major shift in the technology landscape that will ultimately transform the way people interact with computers in the workplace and beyond.

“This is going to be just as big as the internet, and just as big as the mobile revolution, and just as big as the client server revolution—all of those major kinds of shifts that you saw on the market,” Patel said. “We are in the fourth major revolution of how humans interact with machines.”