All Hail Our New Meeting Notetaker: Zoom AI Companion

Never ask for a volunteer to take notes again!

Posted on November 18, 2024 in Blog Posts

“Before we get this meeting started, who would like to volunteer to take the minutes?”

[Complete silence follows]

“It is so easy to take the minutes. Just capture the major points and action steps.”

[Uneasy shuffling ensues, along with complete silence]

“Well, as the chair of this committee I guess I’ll share some minutes after the meeting ends.”

[No minutes are ever produced]

Does this sound familiar to you? It should. A call for a volunteer to take the meeting minutes is as welcome as a call for a volunteer to clean the staff room refrigerator.

I recall one virtual meeting where nearly every participant wholeheartedly expressed in the chat how impossible it was to find anyone to volunteer to take notes for their library’s committee meetings. This was followed by no one volunteering to take notes for this meeting. A threat from the meeting leader to set up an advance schedule of notetakers for each meeting was poorly received.

There are two things on which all librarians can agree. First, no one wants to take meeting minutes. Second, we all think someone else should take the meeting minutes. You have literally heard multiple times in your career someone in a meeting say, “I can’t take the minutes because I am really bad at taking minutes.”

Is it possible that among all of the technologies we have encountered in our careers we might finally have the technological solution to the conundrum of who is going to take the notes? I think so.

The next time there’s a call for a note-taking volunteer, we will no longer need to settle it by drawing straws or taking pieces of folded paper out of a jar where all are blank except the one that says, “You have to take the notes, loser.”

Now we can all point to our computers and say together, “Zoom AI Companion, please perform this dirty deed for us.”

I’ve been using the AI Companion to create meeting summaries since this feature debuted in Zoom. I have deployed it for one-on-one meetings with colleagues. More frequently, it is used for committee meetings. While Zoom’s AI Companion is hardly your only option if you want to use AI to record meeting notes, its integration within Zoom gives it a significant advantage over competitors.

Here I want to share some reactions, from both my own perspective and those who have participated in the meetings where I have used AI Companion, and offer some tips for when you can expect the AI Companion to perform well or less well than hoped. Overall, my experience has been mostly good, and I think yours will be as well if you have realistic expectations about the capabilities of AI note-taking assistants.

---

🌟 Subscribe to the LibTech Insights newsletter for weekly roundups and bonus content, including: 

• Ready for It? Ensuring Web Content Accessibility in Libraries
• Getting the Best Results from the Smithsonian Digital Collection
• An Open Knowledge Partnership: Wiki Movimento Brasil and Libraries

---

Getting Started

First, confirm that your institution’s Zoom administrators have enabled the AI Companion in your version of Zoom, as the default institutional setting is disabled. Second, make sure your version of Zoom is a licensed version, as the AI Companion is unavailable in the free version. One way to quickly ascertain your access to the AI Assistant is to navigate to the general settings of your Zoom software. You should see this:

The one setup item to consider carefully is whether you want to allow the AI Companion tool to activate automatically when the meeting starts. If you plan to use the Companion tool regularly, it may be preferable to set the toggle to “Turn on meeting summary automatically when meetings start.” This eliminates the need to remember to turn it on, and attendees will get a notification that the AI Companion summary tool is activated. If your use will be more limited or selective, I recommend leaving the above setting off and turning it on manually as needed.

Second, start a new Zoom meeting session to confirm this operation succeeded. You should now see the icon for the Zoom AI Companion in your toolbar, as illustrated below:

Before using AI Companion in an actual meeting, I recommend pairing up with an interested colleague to conduct a trial run. This will allow you to familiarize yourself with the AI Companion’s operation from initialization in the meeting to receipt of the meeting summary in your email. To Zoom’s credit, the AI Companion is simple and unfettered by multiple options that lead to confusion. As is the case with other operations, such as polling or screen sharing, this can be both a plus and a hindrance at times, depending on your needs and intentions. It takes little time to gain enough familiarity with the AI Companion to put it into practice at your next meeting.

AI Companion Features

While it’s easy to get started with AI Companion’s most basic operations, primarily creating a meeting summary, there are some additional features and options that could be of interest. One of the simple ones is in-meeting questions. This feature allows participants to ask questions that can be answered if the meeting summary has coverage. Questions can range from “catch me up on what I missed” if arriving to the meeting late, “what decision was made,” or just about anything else that a participant might want to know about the proceedings.

AI Companion can interface with other Zoom features, such as Whiteboard and team chat. One of the more interesting features, not yet activated at my institution, is “smart recording.” This opens options for a host of features for better organizing the meeting summary output along with advanced editing features. One of the features organizations will want to consider is the ability to capture complete audio transcripts of meetings when AI Companion is used to create meeting summaries.

My understanding of how the AI Companion works is that Zoom records an audio transcript of the entire meeting. It then generates the meeting summary from that audio transcript. This is a temporary file that is deleted once the summary is generated. However, organizations have the option to enable access to the full transcripts, in which case they would not be deleted but would be available on the cloud. This adds an additional option for getting a full transcript without needing to record a Zoom meeting.

Tips for Beginners

As is the case with most new tech tools, you can anticipate some unexpected results with your initial experiments with AI Companion mode. The most important thing to keep in mind, now that the AI Companion is integrated into all the Zoom meetings you host, is to decide if you want AI Companion to create a meeting summary. That means giving some thought to the nature of the meeting and whether or not the generation of notes is appropriate or necessary.

If you choose to enable AI Companion for a meeting, the new AI etiquette is to let attendees know in advance that you plan to use AI Companion to assist with note-taking. This notification could be shared via a pre-meeting email reminder or an announcement at the start of the meeting. Reactions could vary from outright relief from those who feared having to do that job themselves to some degree of resistance to having AI involved in any way whatsoever in the meeting. We may have colleagues with AI trust issues.

I have yet to experience a situation when a meeting attendee was downright opposed to my use of the AI Companion. That said, I’ve tried to be judicious about when to use it, which is primarily for committees where there is a traditional, assigned notetaker. If you do choose to enable AI Companion for automatic summarizing, it will generate a notice to your attendees informing them that Companion is in use. As is the case with Zoom video recordings, participants receive a consent notice.

Managing the Summary Output

Where the AI Companion may produce the most uncertainty for users is the quality of the actual output. Put simply, sometimes the summary is remarkably accurate and sometimes it just stinks. A meeting summary, which is typically sent to the meeting host’s email within five to ten minutes after the meeting ends, consists of several components. First is a high-level summary of the entire meeting called “quick recap.” It condenses the summary into a single paragraph. Then comes “next steps,” a bulleted list of actions to be taken after the meeting. I’ve always found it tricky, as a note-taker, to get the next steps just right, so this is my favorite feature of the entire summary output. Then comes the actual summary of the meeting where AI Companion attempts to organize the meeting by individual agenda items with section headings. I appreciate the attempt at organization, but the results are sometimes less than useful.

Here are some issues you’ll see in these summaries:

• Garbling of names of who said what in the meeting or outright attaching a statement to the wrong individual.
  
  
• Section headings that are inserted oddly in the summary or fail to accurately reflect what actually happened.
  
  
• Repetition of meeting conversation; for example, you will see things like “Person One will work on a newspaper strategy to improve the dissemination of information to members. Person Two then discussed the idea of rethinking how the newsletter could better disseminate the organizational update.” While that might not be awful to have in AI-generated notes, it’s hardly how you or I would write it.
  
  
• The quick recap section can vary wildly from spot-on to utterly useless.
  
  
• Mashing together statements that were made by two different participants.

While these are nuisances that may leave one frustrated, even an imperfect summary is a far better meeting experience than having to take copious notes on one’s computer. It’s far more rewarding to be an engaged participant than an awkward appendage stuck in note-taking mode. You can likely recall meetings where you were both the notetaker and a presenter. It’s almost impossible to both speak and take notes about what you are saying. Sure, you can fill that in later, but it’s a real time-saver to grab what you yourself said and integrate it into the final meeting minutes.

The other way in which I and others have found the meeting summaries that AI Companion generates somewhat frustrating is the way it is delivered in an email message. The summary is formatted as a table. This makes it rather inconvenient to move the text into a word processor for editing. The process involves deconstructing the table by removing rows and spacing to get to the text. It’s a tedious but not difficult task. I’d like to see Zoom improve this by providing the summary, post-meeting, in a format that is easy to import into an editing tool.

Once you anticipate the types of errors the AI Companion creates it becomes easier to quickly correct them, for example, removing the provided headings and replacing them with more accurate headings. Based on my experience, the time involved in reviewing and editing the summary—and the summary should ALWAYS be reviewed and likely edited before it is shared—should be approximately 15 to 20 minutes on average to complete the task. No, far from perfect. Does it have the potential to save human notetakers time? Absolutely.

Not All Meetings Are Equal (for Companion)

More commonly we find ourselves in meetings where the participants are not occupying the same physical space. When Zoom is in use, everyone usually logs in from disparate remote locations. That is typically the case with a professional society committee where members are from diverse geographic regions. It might be a meeting of coworkers who are all working from home. In these cases, AI Companion works well. It manages to (mostly) accurately identify who’s speaking and summarize what that person said.

That makes sense. Each participant is identified when logged into the group Zoom meeting. If everyone is speaking one at a time, Zoom can easily differentiate you or me from other speakers and record me when my microphone is activated. But what happens if the meeting gets heated and two or more attendees are speaking simultaneously? There’s a much greater chance that AI Companion will confuse which attendee is speaking and what they are saying.

Another tricky scenario for AI Companion is the hybrid meeting. The few times I’ve used AI Companion in this situation is that it will work reasonably well at summarizing what the remote attendees are saying. That is not always the case with those attending the meeting on site. The summary will attribute what anyone in the on-site room said to the meeting host. Again, that makes sense. How else can AI Companion identify who is at the on-site meeting? One possibility is to ask everyone on site to always preface their remarks with something like “This is Steven speaking….”  I’d say that AI Companion for hybrid meetings is a work in progress. It can work, but anticipate doing more cleanup of the summary.

Where’s the Content?

When I speak to colleagues about using the AI Companion, they see the potential benefit, but the issue that rises to the top of the heap is the collection of meeting content. Where exactly is what’s spoken in our meetings being stored and what’s being done with it? As stated above, the audio transcript used to create the meeting summary is a temporary file that exists for approximately 24 hours and is then discarded. You or your colleagues can learn more about how AI Companion handles your data. You’ll see the usual types of disclaimers from Zoom concerning sharing of your data with third parties. It also isn’t used to train Zoom’s own AI or third-party AI models.

There’s no question that some meetings are better kept private, such as one-on-one meetings between supervisors and their reports or those involving personnel matters. Since these types of meetings rarely need start-to-finish recording, it’s unlikely the AI Companion would be considered at all.

What we humans know is that smart, competent note-taking requires nuance and good judgment. In any meeting, what’s being said can be even more important to keep out of the minutes than what goes into it. Knowing the difference between what stays in the room and out of the notes will always require a human touch. As such, it is even more important to submit AI Companion notes to a review when there is potentially sensitive information discussed by participants so that it can be identified and edited or entirely removed.

As this Chronicle of Higher Education article indicates, academic institutions are developing policies that address the use of AI note-taking tools. While most of these policies focus on use by students and faculty in classroom settings, it can also apply to administrative use for committee and other meetings. It is advisable to determine if your higher education institution, or local government in the case of public libraries, has established policies that may impact your adoption and use of AI Companion and similar AI note-taking software.

Conclusion

Here’s the bottom line. Until the day that everyone in a group meeting jumps at the chance to take meeting notes or official committee minutes, what else have you got to compete with the Zoom AI Companion? Sure, it has its flaws. When it comes to capturing meeting notes, we can live with some imperfections. What we do need in our notes is accuracy.

In certain circumstances, it is critical to know exactly what was said and who said it because we all know there is going to be a time when there is some sort of misunderstanding or miscommunication after a meeting when you thought I said this and I thought you said that. What saves us from a nasty situation is going back to the notes.

For now, we are only at the stage of what I’d call machine-assisted note-taking. Tools like Zoom AI Companion can ease the burden of taking notes in meetings but are hardly ready to eliminate the role of humans in the process of generating notes. For now, it will remain important, when using AI Companion, to review the summary carefully to compare what it provided with what actually happened in the meeting. Sometimes it will be surprisingly accurate. Other times you’ll wonder why you bothered using it. There’s only one way to find out what results you will get with Zoom AI Companion.

---

🔥 Sign up for LibTech Insights (LTI) new post notifications and updates.

🧠 Join us for a free webinar on AI literacy in libraries.

✍️ Interested in contributing to LTI? Send an email to Daniel P. at Choice with your topic idea.

Related Posts