Last week’s dust-up over Zoom’s terms of service revealed, if nothing else, how contentious, complex, and contradictory the adoption of AI within communications platforms is likely to be. The more we plan to rely on AI to “improve” meetings and other elements of the employee experience, the likelier it is that users and/or enterprises will object to some aspect of the implementation.
The first generation of pandemic-era AI-driven improvements to videoconferencing was largely uncontroversial, partly because they were fairly limited in their ambitions, and partly because end users were usually the ones who controlled their use. People used background blur to hide their messy work areas and sometimes chose whimsical background images, though nowadays you mostly see one of the generic office space images, some corporate branding, or plain blur. In any case, whether there’s a corporate mandate or just a general practice, end users control it from their interface and seem to have a consensus about what is and isn’t acceptable.
Part of the concern over allowing vendors to train their AI on end user inputs had to do with the permission structure, which was controlled by the enterprise via admin settings. The end user’s option, if they objected, was simply to not attend the meeting, which isn’t usually a realistic decision for someone in a professional organization.
But as we talk about systems—any system—becoming more sophisticated in order to provide meeting equity, we’re talking about handing over more control to the centralized system, potentially at the expense of the end users’ desires. I’ve written before about how just the experience of smarter meeting technology may take some getting used to—but I also touched on the concern that any AI system might inadvertently exacerbate, rather than overcome, the biases that tend to emerge in human interactions.
In the meantime, end users continue to want to exert their own measure of control; for example, some users are coming up with life hacks for the meeting space. More than one enterprise IT leader has told me that their users are avoiding the “bowling alley” effect in old-school single-camera meeting rooms by simply bringing their laptops into the room and using the internal camera so they appear individually in a frame, the same as if they were joining remotely. The technically elegant solution here is multi-camera rooms with sophisticated AI “directors” choosing just the right shot; but some users don’t want to wait for that to get rolled out.
The history of the Internet and modern communications technology is the balance between technically elegant systems and “good enough” solutions that can be implemented quickly and with as little bureaucratic intervention as possible. We’re seeing this dynamic play out in the world of video meetings, with the added challenge that we don’t just have to solve technical problems anymore. We need consensus from enterprise leaders, both in business units and IT, as well as buy-in from end users, about the rules of engagement when it comes to AI-enabled meeting systems.
And that’s to say nothing of the vendors, who as this VentureBeat article reports, are taking steps to protect their own intellectual property in AI systems from any encroachment by users or other outside parties.
In short, AI for collaboration is going to be a battleground between vendors, enterprise leadership, and end users. If you think it’s not, you might be, well, hallucinating.
