Notes

My AI roasted me

4.5% airtime. In my own meetings.

By Samer Azar, Fractional CFO · 2026-03-14 · 5 min read

Hi Reader,

Fifteen years ago, a mentor handed me a copy of Patrick Lencioni's The Five Dysfunctions of a Team.

I was early in my career. I thought finance was about numbers. He told me finance was about people. The numbers were just the surface. The real work was understanding how the people in the room made decisions, what made them trust you, what made them go quiet. Board members, investors, other executives. You're always positioning a financial argument to land with someone. And whether it lands depends less on the numbers and more on how that person processes information.

I've been trying to apply that ever since. Tracking it in my head. Adjusting on the fly. A vague sense that one client responds differently to bad news than another. No system. No data. Just mental notes that fade by the next call.

Last month, I had about fifteen Fireflies transcripts from a client's board meetings sitting around. On a whim, I fed them all to the AI and asked one question: what patterns do you see in how I'm showing up?

The first thing it came back with: my average airtime was 4.5%.

I thought that was a typo. I'm the CFO. These are financial board meetings. I checked. It wasn't a typo.

Then it got more specific.

The hedging. "I don't want to sound overly pessimistic, but..." before every hard finding. Asking "is that a good idea?" after my own recommendations.

It told me I was building up to conclusions instead of leading with them. Long methodology explanations, then the verdict at the end. "Lead with the verdict. The forecast is wrong. Here's why."

It told me I was asking a colleague privately what the CEO thought of my analysis, instead of asking the CEO directly. "Challenge him with data. That's how you earn his respect."

It told me I was offering three options when I had a clear recommendation. "I recommend A. Here's why." Not "we could consider A, or alternatively..."

These weren't generic coaching tips. The AI pulled them from specific moments in specific transcripts. It quoted me back to myself.

I ran the same analysis for a different client, different meeting type. My airtime was 40%. Zero hedging. Driving the agenda. Assigning tasks. Same person. Same month.

The variable wasn't the client. It was who else was in the room, and what role they implicitly assigned me.

I shared some of these profiles with a couple of CEOs I work with. The reaction surprised me. They were fascinated. Nobody had ever analyzed how they behave in professional settings with that level of objectivity and accuracy.

One of them told me: "I've been trying to understand myself for 20 years. I didn't think an AI would get me from a couple of calls."

That experiment isn't a one-off anymore. It runs after every client meeting. Here's the sequence:

1. Fireflies transcript drops. Within 10 minutes of ending the call. Raw audio to text, speaker-labeled, timestamped.

2. Three analyses run automatically. Meeting quality score (was the agenda covered, were decisions made, did action items land). Behavioral comparison (how did each participant show up versus their baseline profile). Communication coaching (what did I do well, what should I change next time).

3. The report emails to me. One document. Scores, flags, specific quotes, coaching notes. I don't review a single transcript. The system reads them all for me.

4. The client brief updates itself. Action items from this call get written into the brief for the next one. Commitments with deadlines. Nothing falls through the cracks between meetings.

5. ClickUp tasks get created. Every team commitment becomes a tracked task with a deadline attached. Not in my notes. In the project management tool where it belongs.

6. Profiles get flagged for refresh. If someone's behavior shifted in this meeting (airtime dropped, engagement changed, new stress signals), the system flags their avatar for an update.

7. The meeting ends. Six things happen without me touching anything.

Last week was what happens before the call. The brief that arrives in your inbox so you walk in prepared.

This is what happens after. Together they form a loop. Before: prepared. After: accountable. Every call feeds the next one.

Lencioni wrote about trust and accountability in teams fifteen years ago. What I've built is the system that actually does it. The book was the principle. The system is the memory.

How fast would your career move if someone with infinite patience, experience, and objectivity was watching every meeting and coaching you after? Reply and tell me what you'd want them to watch for.

Next week: how your inbox and calendar become live inputs. Emails classified automatically. Meetings transcribed and analyzed without you reviewing a single transcript.

P.S. I'm doing a live session in a few weeks. The full system you've been reading about, on screen, with anonymized client data. Architecture overview, then 5-7 workflows walked through end to end. How the briefs get built. How the meetings get analyzed. How the follow-ups happen without you.

Register for the live walkthrough