Microsoft Copilot – Teams Edition

by Heidi Elmore

Can We Trust AI with the Specifics?

Meetings. We want less of them, and yet we want more out of them. Microsoft Teams announced their own meeting optimizers at last week’s AI event, with more undoubtedly coming next week at EC. But the value of many of these features makes some important assumptions about what specifics you are ready to give (or give up) to succeed on this productivity mission.

We’ve been inundated with meeting features from all providers since the start of 2020. Many of those capabilities are quite helpful in some specific use cases. For example – Front Row. But ask the average end user outside of the enterprise communications industry, and all most individuals will remember is “custom backgrounds.”

The latest meeting features from Microsoft are about deriving intelligence from the meeting experience. But what if you or your meeting participants don’t trust the AI?

I’ve been a massive fan of the potential for meeting insights since I started using meeting recordings as a Microsoft employee in 2018. (To be fair, those were stored in Stream vs. SharePoint and OneDrive today.) Even the early days of just being able to search transcripts by keywords were incredibly useful.  

Now, with Meeting Recap, the promise of insights increases exponentially. Like, Zoom, and others, there are now options with Copilot after the meeting. You can receive an AI-generated summary of talking points and action items. Unlike other providers, we are told we can even ask questions of Copilot that will query the meeting transcript. 

It’s a very exciting prospect, for sure. But it makes a lot of assumptions about if you can trust the AI, specifically in the inclusivity of the exercise. The most obvious concern is for individuals whose language is not the primary spoken language of the meeting. If accents can’t be transcribed appropriately, will their voices be heard? These individuals already face bias in real-time conversations. Will AI create an inherent bias in our post-meeting assets?

It also creates a potential bias for the most vocal person in the virtual room. If conversations live in perpetuity, and future decision making is hinged on the summary of what is said in volume, dissenting opinions may be overwhelmed.

Inclusivity is an essential tenet of many meeting providers. And yet, I haven’t seen anyone decisively solve this challenge in real-time, let alone with post-call transcripts. Don’t get me wrong. I am excited about these prospects. I just feel like meeting providers have to do more first to capture every participant equally before we decide to make intentionally-detached decisions based on transcripts. 

The next feature of Copilot for Teams is real-time feedback in meetings. This feature isn’t without precedent in other parts of enterprise communications. In the contact center world, live sentiment tracking has been an exciting innovation powered by conversational intelligence. 

For example, Amazon Connect launched live analytics for Contact Lens in 2022 to help agents and supervisors monitor real-time customer satisfaction. But this isn’t to influence a conversation with a coworker. This is to streamline repeatable and bottom-line impacting interactions with a customer.

Meetings Copilot for Microsoft Teams promises to summarize the meeting conversation in real-time, but the demo video also highlights how the group “felt” about the discussion so far. Can we trust the AI to properly analyze our feelings in real-time?

I “feel” like many of these features will promote an environment where individuals will either not want to join a meeting on time to reduce recording and analysis or not attend at all. Will these features force a culture of backchannel conversations to avoid oversight and intrusion? 

In our effort to make meetings more useful, maybe vendors have missed the obvious – how do we just have less meetings? That’s where some of the genius exists in some of the other Microsoft 365 Copilot functionality. With that capability, we can search across assets, meetings, and chats for the latest on a customer or subject to avoid needless update meetings. 

In addition, many people create meetings to transfer knowledge to start asset creation. The ability of Microsoft 365 Copilot to crawl an existing tenant, its content and generate a starting point helps augment many existing needs for meetings. 

What’s the difference between the Microsoft Teams Copilot features and the Microsoft 365 Copilot features? Distance. The further insights are from the interactions, the less it feels like big brother and the more it feels like value. We can trust in aggregates but not specifics. 

Microsoft has learned a lot in its intelligence journey. They’ve come a long way from Workplace Analytics. Rather than boring us with video or audio AI features, they are digging into where the Office Graph can provide value. Time will tell whether these features can be delivered in a way that will stay compliant with industry or country-specific requirements. And that’s ok, because trust takes time.

All of these features are yet to be released to the public rings. More announcements are coming from Microsoft. In addition to the Enterprise Connect announcements due on Monday, 3/27, a series of announcements are coming from Microsoft Viva (a cousin of Teams) on 4/20.