A riddle for you…

The typical corporate employee spends 50%+ of their professional existence in meetings. Historically, that’s a lot of time spent taking notes, transcribing, driving follow-ups. Meetings often represent a goldmine of information shared but hard to capture, organize, and drive actions. 

Recently, I worked with a non-profit that designated a ‘secretary’ for each in-person board meeting, tasked with taking detailed notes. That person was talented and could be contributing if they weren’t spending all their time jotting down every note spoken. The organization needed to change the way they ran their meetings to integrate technology. Now, they use an owl to capture the meeting audio, a transcription tool to convert the audio to text (Otter.ai), and an LLM to analyze the output and summarize. They start out each meeting by introducing each attendee so that the AI learns their voice and make sure to speak clearly into the mic. Meanwhile, the secretary has the Otter app open and jots down important context that might not be captured through the transcript - which goes into the summary. This simple change freed up a human that can now think, participate in the discussion and focus on annotating the transcript with important context and insights. 

This might sound like an extreme example, but I am confident that almost every meeting in Corporate America could benefit from further infusing AI into the planning, execution, and follow-up processes. I’m sure you are already seeing some benefit to using AI in meetings - you might have a Zoom AI companion or use a tool like Gong, but I can almost guarantee you that you are leaving some efficiency opportunities on the table by not adapting your meeting practices to fit new world.

In order to get the most out of AI, you need to change the way you approach meetings and ingrain it into the three parts of a meeting: Prep, Execution, Follow-up

Meeting Prep:

I like to come to meetings over-prepared. In terms of what I know, what I share with my teammates and what I share with my customer/partner/client. Coming prepared and sending out a detailed agenda prior can be taxing if you do it all manually, but can be a superpower if you automate it.

For me, research for an important meeting starts in SearchGPT or Perplexity with a prompt like this:

Please help me prepare for an upcoming meeting by following these guidelines:

Meeting Details:

Who I'm Meeting With (People, company, role): <Insert Name of Person/People, company, role>
Meeting purpose: <Insert quick blurb on topic of meeting>
Success Criteria: <Describe What Success Looks Like for This Meeting>

Research Objectives:

Research the Person(s):

Gather recent news, background information, and notable achievements about the individual(s). Share their current role, prior roles, alma matter. 
Identify any recent projects, publications, or statements they've made.

Research the Company:

Provide an overview of the company's recent news, performance, and significant events. Think about it specifically in the context of what I already know and success criteria, try to pull in relevant info
Highlight industry trends or challenges relevant to the company.

The second piece of meeting prep is synthesizing all notes from prior conversations (which gets easier as you start to take better notes). I throw all previous relevant notes into ChatGPT along with the output from SearchGPT/Perplexity above to create a prep doc, agenda and list of questions I want to ask.

Meeting Execution

We need to change how we run meetings in order to maximize the impact of our AI companions. What this looks like is somewhat dependent on the type of meeting (in person vs remote) and technology available (recording, transcription, note taking).

So first you need to understand what tools are available to you and fit within company’s policies. Then optimize around them.

Audio: For example, if you are doing audio recording of an in-person meeting, make sure folks introduce themselves in order to associate the voice identification with a name in the transcription.

Transcription: For any sort of transcription, I like to have certain trigger words I then use in my follow up prompt like “action item” or “next step” that I try to use consistently in order to improve the accuracy and comprehensiveness of the automated notes. I am a big fan of Granola for transcription capture:

Manual note taking: There are many meetings with audio and transcription are not an option for many reasons. In cases where I need to manually take notes, I use a system to make it easier for the AI to summarize the key points I am capturing, I include different symbols and assign each a meaning in my follow-up prompt (see below).

• ““= verbatim quote

• *= important point

• ^= action

• ?= follow up questions or further research

Meeting Follow-Up

Any good meeting will succeed or die based on the follow-up. And here, again, AI can help you keep the momentum going without taking too much of your time. The first thing I do after a meeting is generate a refined summary based on the transcript or notes available.

I use a standard base-prompt and then insert meeting notes at the bottom. You will see that I also identify and label the various symbols above. I have generally found that ChatGPT’s o1 model does the best job for this - although any state-of-the-art model should work well.

Please summarize the raw meeting notes included after the "//" below according to these guidelines:

Executive Summary: Start with a brief overview of the meeting's main objectives and outcomes.

Important Points: List all important points marked with '*'. Organize them clearly under this section.

Key Sections: Split the notes into sections based on the topics discussed. Elaborate on each key point within its respective section.

Action Items: Include all action items marked with '^', specifying who is responsible for each action. List them under 'Action Items'.

Follow-Up Questions/Research: List any follow-up questions or items needing further research marked with '?', under 'Follow-Up Items'.

Verbatim Quotes: Incorporate any verbatim quotes noted by double quotes ("") in relevant places throughout the summary in italics. Include all quotes, as verbatims are important!

Style Guidelines:

Conciseness: The summary should be as concise as possible while maintaining the meaning of the notes.
Language: Avoid using flowery or superfluous words.
Tone: Write in a neutral, colloquial tone.
Recommendations: Suggest any additional meetings or actions if necessary.

//
<meeting notes>

Lastly, if you really want to get fancy, you can use Google’s NotebookLM to create a custom podcast that summarizes your meeting. This can be useful in disseminating the knowledge to a wide audience.

We are all living and dying in meetings. It’s not about having more meetings; it’s about having better ones. AI can make that difference, starting now.