AI Meeting Assistants
AI Meeting Assistants will join a meeting, transcribe the conversation and provide a summary. They integrate with remote meeting tools such as Teams and Zoom. Ideally the user should be present and taking an active part in the meeting although they could just leave the note-taker on and ignore the meeting progress. Some AI assistants will even join and record a meeting if the participant hosting them is not present.
There have always been issues with overlong meetings or information that could have been more easily distributed by email. Nevertheless meetings do fulfil a real need by allowing information to be exchanged and decisions to be made.
The technical concerns over AI Meeting Assistants revolve around what they record and what happens to the data that they process. Many of the tools including ‘Otter AI’ and ‘Tl;dv’ offer comprehensive free plans. Although these would not be suitable for a larger business user they could attract meeting attendees to use them without the permission of a corporate network. The application bots should advertise their presence and ask for consent; a company policy needs to be in place to set clear rules on their use.
If the application is only providing a complete transcript of a meeting then it needs to be accurate. Such a transcript could contain a large degree of irrelevant content that will need to be waded through and key content identified. This is only of limited use to the busy worker. Having a clear and concise summary is far more useful and should be the output of a good note-taker. The degree of reliance on the AI to do just that is a measure of how useful the tool is. It should be able to take out irrelevant asides or comments that seemed appropriate at the time but were not thought through and which the speaker would rather not seen recorded. Speakers might even be intimidated to not contribute in case their ideas are noted and clearly seen to be unwise. Others might contribute with more content with an aim to making the summary rate their contribution more highly; making the overall meeting longer but to no useful purpose.
The more effort a human has to put into analysing the output the less useful is the tool. On the other hand although the notes need to be concise then no core point should be omitted. Continued use of a tool should allow the AI to learn what to look for and improve the final output.
A privacy worry stems from where the harvested data goes. If an AI note-taker joins from outside an organisation it will record corporate information that ought not to have been disclosed. This needs to be addressed by rules on which individuals, locations and AI note-taking systems will be allowed to join corporate meetings. The AI engine will be storing its findings and using that to further improve the decisions that it will make in future. This would include data that does not appear on the notes that it outputs and could include sensitive information. Although unlikely this information could pop up in a subsequent output accessed by that engine.
