What a Year of Building AI Agents Taught Me

What I've been noticing lately

A year ago, I didn't want to spend months watching AI evolve from the sidelines while making avoidable mistakes along the way. So I took a class and learned to build an agent myself.

We put 25 years of how I think into a box and called her Nancy.

What surprised me was not what the agent could do. It was what the process revealed about me.

To teach Nancy to write a proposal or prepare discovery questions the way I would, I first had to explain how I actually make decisions. And that was harder than I expected.

Much of what I rely on was never documented. It lived in experience, instinct, and years of recognizing patterns without consciously thinking about them.

Why it keeps showing up

The hard part was never the technology.

The hard part was explaining the judgment behind the work.

For years, I assumed our consultants' greatest value was their knowledge of NetSuite. What I came to appreciate is that their real value lies in understanding what is actually broken in a business versus what appears to be broken.

That distinction matters.

Building Nancy, and later a growing workforce of agents, forced us to make that thinking visible. We had to describe how we approach problems, what questions we ask, and how we arrive at conclusions.

In the process, we captured knowledge that previously existed only in people's heads.

New team members can now start from our best thinking instead of starting from scratch.

Not because the technology is extraordinary, but because it pushed us to become more explicit about how we work.

What it makes me wonder

I wonder how much expertise inside organizations remains trapped this way.

Critical knowledge. Daily judgment. Years of experience.

Essential to the business, yet largely undocumented.

AI may be the catalyst, but people are still the ones who see patterns, exercise judgment, and make decisions.

What I didn't expect was that building these agents would sharpen my own thinking too.

Explaining my reasoning to a machine forced me to understand it better myself.

AI is not a product you buy once. It is a capability you develop over time.

And sometimes, the process of building that capability changes the builder as much as the thing being built.