In this section I review one AI-powered application and demonstrate how it can be used to create new value. This week is a bit different, because the application is one I built myself.
What does a month of weekly meetings add up to?
In my ongoing work with HID Global, I am part of their AI core team, which meets weekly to sync on our AI initiatives. Each meeting leaves behind notes and action items, and we wanted to see what the month as a whole added up to. To do that, I built a custom agent in Microsoft 365 Copilot. It reads our weekly meeting transcripts and drafts an Amazon-style monthly business review (MBR). After a couple of rounds of iteration and refinement, it now works well.
What does the review include?
Many business updates at Amazon are written as narrative documents instead of slide presentations. An MBR is one of those narratives. It keeps information flowing up and down the organization, and around it, in a format leaders can aggregate, and it helps them dive deeper where something needs their attention. My agent uses the same structure: a short executive summary and three main parts, with a list of open questions for me at the end.
- Health of the business - how the main AI initiatives and goals are progressing, adoption, the month's key decisions, and any business metrics showing progress toward the goals.
- Opportunities, risks, trends and blockers - for each one, it helps spell out the evidence and why it matters. It also lists any actions and asks from leadership.
- Highlights and lowlights - one or two anecdotes of each, like something a stakeholder said or an incident, and what each one teaches the team.
Amazon's original MBR structure also includes a section of metrics for the KPIs the writing team has identified as important. For now we left that out of our MBR at HID Global, since the KPIs are still being worked out.
How do you keep the agent from making things up?
On top of the structure, a few rules in the agent's instructions help keep the review accurate. Facts have to stay separate from interpretation, and every important claim needs a meeting date or a quote behind it. It must never invent progress or outcomes, and the same goes for risks and quotes. If the evidence for an item is missing or unclear, the agent asks me how to handle it before writing the final version. It also compares the month with earlier reviews, and if there is no earlier review yet, the review says so.
How can you build your own MBR agent?
Microsoft 365 Copilot lets you build an agent like this in Agent Builder, with no code. Turn on transcription in your Teams meetings, and it can read the transcripts. A ChatGPT or Claude project holding your meeting notes and the same instructions can do something similar if you don't use Microsoft 365 Copilot. A prompt like this one is a good place to start:
Here are the notes and transcripts of our weekly team meetings for [month]. Write a monthly business review with a short executive summary and three sections: 1. Health of the business - progress on our initiatives and goals, with any business metrics mentioned. 2. Opportunities, risks, trends and blockers - for each one, the evidence, the business implication, our response and what we need from leadership. 3. Highlights and lowlights - one or two specific anecdotes each, such as a quote or an incident, and what each one teaches us. Tie every claim to a meeting date or a quote, and never invent progress or quotes. If more information is needed, interview me before you write the final version.
Go deeper
For a look at how an agent like this gets built, watch Ned Friend create one that writes weekly status reports from meetings and email.
Your action step
Pick one recurring meeting in your organization, collect a month of its notes or transcripts, and run the prompt above on them. Before you share the draft, check each claim against the meeting date or quote it points to.
Which of your organization's recurring meetings could become a monthly business review?
If you want to find where AI agents can create value in your own leadership routines, that is the kind of question I take on in AI strategy advisory engagements and in AI agents executive sessions.