It probably goes without saying that two of the most important capabilities of artificial intelligence (AI) is its ability to exactly reproduce a conversation, but much more importantly, its ability to derive some useful insights about the content of the conversation as a whole – for instance determining sentiment. Together these capabilities are sometimes called ‘conversational intelligence’.
The automated admin assistant?
At this point in time though, it’s safe to say that conversational intelligence is a solution looking for a problem. The sorts of uses it’s currently being put to include transcribing a meeting and summarize the key points – which is not exactly shaking the world (especially as humans can still do this much better). In the meantime, there’s a whole bunch of innovative firms like BlueCap.ai and Sonero.ai all working hard to find other applications for it.
Analysis not admin
One area would be better for is setting up workflows. Say a meeting conversation includes the sentence: “The quality department needs to update the documentation on solvent-free paint.” It’s conceivable – is it not – that the AI tool could automatically send that request to the appropriate people?
But important though the automated admin assistant angle is, I believe the most exciting potential of conversational intelligence lies in analysis, not transcription. Using it for analysis would be like having a coach, a superhuman, able to sit in on every meeting.
The value they could bring is almost endless. Here are just some of the possibilities:
- In an organization with a lot of project teams, a conversational intelligence tool could flag any teams that seem to be getting into trouble so that the unit leader can check in to see how to help.
- In an organization where professionals complain that they waste too much time in meetings, it could seek insights on how to cut down on unnecessary meeting time.
- In an organization with an active high potential program, it could provide coaching to the high potential employees based on how they behave in meetings.
Alerts rather than dashboards
As you might be able to deduce, I’m particularly in favor of using these tools for specific purposes, usually with a limited lifespan, rather than seeing them as an “always-on” monitoring tool that gathers data you hope will be useful.
I especially like the idea of using them to issue alerts, rather than providing a dashboard that an analyst needs to make sense of. For example, the head of IT may be intrigued by the idea that they can look at a dashboard for each of their project teams and analyze the meetings from multiple perspectives. However, in practice they won’t have the time to keep such a close eye on each meeting. What they will want, however, is for the conversational intelligence tool to sit in the background and when it detects a problem (eg: “There appears to be a lot of conflict on the ERP integration team”), it will send an alert and the IT leader can decide if they ought to check in with the team leader.
Privacy problems are solvable
People are always concerned about the privacy aspects of any AI tool that collects data. But while this is an appropriate concern, it’s often treated as a kind of boogieman, not a business problem to be solved. In a particular application in a particular company, decisions will be made on what data is collected, how it is anonymized, and who gets to see it, and so on. In other words, privacy concerns are issues that can be addressed; not a reason to shy away from a powerful tool.
How well does conversational intelligence work?
Conversational intelligence, like so much of AI, sits in that strange valley of being able to do some amazing things (including accurately identifying pictures of kittens!), while being oddly stupid at times (mistaking a kitten for an elephant). Their saving grace is that they are used to provide suggestions to managers, not give definitive answers. For instance, a conversational intelligence tool may tell a high potential employee that they seem to be taking up too much airtime in meetings; but it’s up to the high potential employee to decide, based on the context, whether they agree it’s something they should change. Put simply, the only way you’ll really know whether these tools help with a particular use case is to try it out.
What’s next for these tools
I was surprised, but probably shouldn’t have been, that Sonero is already looking at using AI to assess the video component of online meetings to gather additional insights. You can imagine the tool noting when someone in the meeting looks aghast at some suggestion. Again, we’re in a situation where we have a solution looking for a problem, but if you’re old enough you’ll recall lasers started in that same situation, and look where they are today. It’s also worth noting that there is a good chance that we’ll see applications designed for specific niches, such as in the medical or legal sectors, before we see widespread adoption in assessing general business meetings.
Conclusion
I want to share an insight from Graham Westwood, an advisor to Uncanny Labs, the developer of Bluecap.
He says: “Don’t think of these tools as just providing an insight about a meeting. The meeting is a window into what is going on in the organization.”
The potential for these to be important strategic tools is here; now we just need innovative leaders to explore what they can do.