My AI rated my performance in an all-hands 40/100

Some situations are very easy for us humans to decipher but tough as a nut for your AI to crack. This often leads to us humans thinking AI is not up to the task, when in reality, we are at fault for being messy. The first time I ran into this was when seeking meeting feedback but I since learned it applies to all work.

A year ago, I wanted to build a meeting feedback coach for myself. The idea was simple: No one ever gives me actionable feedback on how well (or badly) I am doing in meetings. But meetings had become the majority of my day. To grow, I needed feedback.

AI was just getting good, so I decided to try to make it work.

One of the first pieces of feedback I received was:

Regarding a 2-minute all-hands update. But AI thought that, because I am a manager, I should own the meeting.

I was disillusioned. But I was also dedicated to making it work. Turns out it was much more complicated than anticipated. Let me walk you through some examples:

Your Role Matters

You are in a meeting with your boss. Unloading your problems, seeking help and talking most of the time might be expected here.

How about when you are the manager? Imagine you are unloading all your problems to your direct report. You talk most of the time and you ask them what to do about it. Pretty shitty performance, right?

For us humans this is clear. Your AI does not know what relationship you have with the person in the room. What your role is. Without that context, it can’t evaluate your performance. It’s just impossible.

The Room Matters More

Now, imagine these two meetings for a project manager leading a big initiative:

• One hour company all-hands. Two minute slot to share progress and update all employees of what is happening.
• Steering committee meeting for your project.

Your role is different depending on the room you are in. That is critical context your AI needs to evaluate your performance.

For us humans this is clear as day. For AI it is not. It doesn't necessarily know about the purpose of the meeting. The relationship between the people attending or the role you should have assumed in that room.

Try it out yourself. Provide Claude/ChatGPT a transcript of a meeting and ask it for feedback. After that do the same but this time explain what people expect from you in that meeting, explain company politics, the relationship you have with people.

Compare the results. Do you see the difference?

Mapping The Messiness For AI

When people talk about the context AI needs, we often discuss documentation, facts, thoughts. We rarely discuss the human part in it. The relationships we share. The conflict between teams that had been going on for months. The corporate politics that might be at play.

But our reality is mostly driven by this human messiness. Not by facts.

AI tends to default to the perfect fact-driven world without emotions.

Mapping out our messy environment for AI to understand our complex and chaotic reality is crucial if you want to use it as your personal assistant. That is not just true for meeting feedback. It's true for nearly everything you do with AI.

You don't have to try it yourself. I have packed all of my learnings about the meeting feedback coach into a skill for you.

That is what mine sends me now. It knows the room, so it knows that 40% is too much in that chair.

Download Free Skill

Best

Myles

PS: There is another cohort of Mastering AI: Become an AI-empowered operator starting next week. If you want to learn to build skills like this yourself, join us.