Meeting notes and AI Governance

I’ve seen a raft of articles recently discussing the growing popularity of AI note taker tool in meetings (example) . These articles haven’t been particularly deep – focusing on the comedy of having more AI attendees than people.

The use of these tools to transcribe and summarise meetings is a great opportunity to practice AI governance, both as an individual and as an enterprise. This is a low-stakes, easy example of AI usage that can help build your governance muscle memory.

The norms for this are going to change over the next few years. At the moment there is an extremely wide spectrum of possible responses – from “we don’t need any rules in this area” (a perfectly valid governance outcome” to “we do not allow these tools to be used under any circumstance”.  A more nuanced approach will end up somewhere in the middle, with some guard rails that don’t stop innovation, productivity or experimentation.

The concept of ‘hybrid’ meetings is becoming embedded in corporate culture. We are all gaining skills in how to make these work for all. But at the same time, we are seeing a broadening of what ‘attending’ a meeting can mean – I can attend in person. I can be online. I can be online but openly working on other things; lurking in case my attention is needed. I can watch the recording later-on. And now with an AI tool I can review a summary of the meeting.

Here are some questions that such a governance process might consider:

·         Are all meetings equal – are there some that might be treated differently: One-on-ones? Board meetings?

·         What notification, permission, consent is appropriate when an AI tool is in use. The current rules you use around recording meetings are a good starting point.

·         How are meeting transcripts and summaries handled after the meeting: Are they distributed to all? Are they regarded as personal notes of the user?  Are they retained?

·         What if the meeting has official minutes? What is the relationship to any AI summary?

·         Are their any risks to be considered: Bias? Hallucinations? Accuracy?

If you are doing this as a personal exercise, then make sure you write down the answers, even just for your own reference. This will allow you to monitor over time. As you gain more experience you may find that you have learnt lessons that can be reflected back into the guard rails.

Next
Next

When is ‘not-a-bot’ (ie human) a differentiator?