Charting Cisco's AI Roadmap
Collaboration Group CTO Jonathan Rosenberg talks future plans for AI.
Cisco last week announced Spark Assistant, an artificial intelligence-powered digital assistant focused on improving the efficiency of meetings, as you might have read on No Jitter (UC analyst Zeus Kerravala gave an overview of the assistant, while industry analyst/consultant Brent Kelly dove deeper into where and how AI comes into play).
As intriguing as Spark Assistant is, it's only the beginning of a decade-long AI journey Cisco has embarked on, as I learned in a briefing with Jonathan Rosenberg, VP & CTO of Cisco's Collaboration Technology Group. "This is a different kind of product... a technology that learns along the way," he said.
While voice assistant traffic was negligible just three years ago, today it makes up 20% of traffic, estimated Rosenberg based on public statements from Apple and Google. "Cortana, Siri, Google, and Alexa get 50 billion voice queries per month," he calculated using publically available statistics. This technology is now mainstream." The significant consumer adoption of voice assistants drove Cisco to start questioning what this type of technology should be able to do for the enterprise workplace, he added.
The answer was kind of simple, Rosenberg said. "You would want an assistant for you and your team that could do everything for you, that has access to the wealth of knowledge of your entire company... its people, its places, its data... and be able to help you through your day."
Five Levels of AI
Spark Assistant is at level one of a five-level AI strategy, Rosenberg said. Level one is relatively simple, putting aside all the backend work required. It's about command and control, being able to direct Spark Assistant to start a meeting or place a call to Joshua, he said. These are the kind of capabilities we all heard about in last week's Spark Assistant announcement.
Level two will get into natural language understanding and use context, Rosenberg said. When Spark Assistant gets to this level, it will be able to do things like remind a user to send a note to a particular business team. The assistant will be able to create action items, keep track of conversations, and know what files or slides to share, he said.
At level three, Spark Assistant will have reached the point of "semantic understanding," Rosenberg said. In this stage, Cisco expects Spark Assistant to be able to synthesize meeting content and deliver actionable summaries. Upon hearing a meeting participant say something like "Josh, follow up with Michelle about this," Spark Assistant would create an action item.
In level four, the assistant would go a step further and be more proactive -- asking Josh if he would like to share a file with Michelle, and then doing so based on a "Yes" response. At this level, the assistant would be connecting dots and starting to take actions that would offload some of the more menial action items to automation.
Finally, at level five, which Rosenberg emphasized is a decade out, we'll see the assistant create strategic recommendations and provide insights into things you should do. For example, an AI-powered assistant at this level might be able to tell a company that it needs to increase its number of sales people in Europe in order to meet quotas.
Having achieved level one development, Cisco plans to roll out the initial version of Spark Assistant to select customers in the beginning of 2018, Rosenberg said. Based on feedback from this customer group, Cisco will be able to solve a lot of preliminary technical hurdles, Rosenberg added. Spark Assistant will learn from the voice commands it receives, and Cisco will use that insight to refine the capabilities it envision for levels two through five.
On a final note, a conversation with Rosenberg isn't complete without discussion of security and privacy. "This is one of several things that makes this product harder [to develop] for the enterprise than for consumers," he noted. "We have a whole set of things around privacy controls and user opt-in/opt-out. We have to have our statements set about how long we are keeping data and provide clarity and granularity around how we are using that data."
Hear directly from Rosenberg on Cisco's vision and product direction in a keynote address at Enterprise Connect Orlando 2018, coming March 12 to 15. He'll take the stage on Tuesday, March 13, at 10 a.m. Register now using the code NOJITTER to save an additional $200 off the Advance Rate or get a free Expo Plus pass.
- Hey Spark, How Is Cisco Partner Summit?
- Where's the AI in Cisco Spark Assistant
- Salesforce Expands AI Toolkit