On a Tuesday morning, all our team members gathered for the daily standup.

Then my turn came.

I said something like: “Hey, the AI tool is available now. You can go request your license.” Then I moved on to my work update.

Nobody asked a question or anything. The next person started talking about their tasks, and the standup kept going.

I walked out thinking I had done my part.

Weeks later, I found out some people on the team still hadn’t requested a license. Not because they were against it, but they had no strong reason to go do it.

And I remember being confused. I made the announcement to them. What else was I supposed to do?

When I was appointed as an AI champion, I built a mental model of the job almost immediately.

It looked like this:

The company rolls out AI tools → I learn about them → I tell my team → I collect feedback → I pass the feedback to our AI team → Repeat …

Information flows from the top, through me, to the team. I am the relay station.

So that’s what I did.

I posted announcements in Slack when new features dropped. I mentioned tools in the standup. I gathered opinions and forwarded them. I was diligent about it. I took the role seriously.

But after a few months, I started noticing something. The people who were going to try AI tools had already tried them. The people who hadn’t been were still in the same spot. My announcements weren’t moving anyone from one group to the other.

I kept doing the relay work because I didn’t know what else the job could be.

Here’s the thing about being picked as an AI champion: you get picked because you are already comfortable with AI.

You prompt naturally. You automate small tasks without thinking about it. Talking to an LLM feels like second nature.

I assumed everyone on the team had that same baseline.

I thought prompting was table stakes for software engineers in 2025. I figured the only barrier was access to the tools, and once I removed that barrier (by announcing availability), people would start using them the way I did.

Not everyone thinks about tasks the same way.

Some engineers look at a repetitive task and immediately think “I can ask an AI to handle this.” Others look at the same task and think “let me write a script” or “let me knock it out manually, it will take 20 minutes.”

Neither approach is wrong. But the second group doesn’t naturally reach for AI, and no amount of Slack announcements can change that instinct.

The gap I missed wasn’t about tools or access. It was about the distance between my own comfort level and the team’s. I treated my fluency as the default. It wasn’t.

A few months in, I started to see how AI champions on other teams operated.

In cross-team sharing sessions. In Slack channels. They were doing something I wasn’t.

One champion walked through a specific task, showed exactly how they prompted the AI, and shared the output. Another posted a use case on Slack, including screenshots of the conversation. At least something concrete. “I asked it to do X. Here’s what came back. Here’s where I had to correct it.”

My reaction wasn’t “I don’t know how to do that.” I knew all of it. I had been doing similar things on my own for months.

That was the moment. It was a quiet realization that I had been sitting on exactly the kind of knowledge my team needed to see, and I had kept it invisible because I assumed it was obvious.

I confused being good at using AI with doing the champion’s job. They are not the same thing.

I started doing something simple. In conversations (meetings, casual chats, wherever it came up), I began sharing specific experiences. Sentences like: “I asked the AI to do this thing yesterday, and it handled the whole task end to end.” Or: “I tried using it for this, and it didn’t work well.”

The shift was slow, but I noticed it. People started responding in kind. Leads began mentioning their own experiments in passing. Someone would say, “I asked the AI about this, and it was useful,” or “I tried it for that, and it wasn’t great.” These weren’t formal presentations. They were the kind of observations that surface when one person opens the door first.

Early on, the conversations were about limitations. The AI hallucinates. It doesn’t read the attached files properly. The usual friction of a new tool.

But over time, as people built more familiarity, the conversations changed. We started talking about what the AI handles well and what it doesn’t. We figured out how to integrate it with other tools through MCP, connecting it to our Atlassian suite and building custom skills. The discussions moved from “can we even use this?” to “how do we use this better?”

I didn’t cause that shift, but I was the first person on the team to say “here’s what I am doing with it” out loud. That’s a smaller contribution than it sounds, but I think it mattered more than any Slack announcement I ever posted.

If I look at my own path honestly, I went through a few stages.

The first stage was relaying information. I passed information. I announced tools. I collected feedback and sent it upstream. I was a channel, not a participant. This felt productive because I was busy. But the team’s relationship with AI didn’t change during this period.

The second stage started when I began using AI visibly. In front of people. Sharing what worked, sharing what failed, in the normal flow of work. Not scheduled demos or formal sessions. Conversations. This mattered because it made AI usage a normal thing to talk about, not a special initiative to adopt.

I am now early in a third stage of this process. I moved to a new team recently, and I am trying to be more deliberate from the start. I have been building more AI-driven automations and sharing them openly, hoping to spark ideas rather than prescribe tools. My dev manager and I have talked about creating a low-friction way for people to log their AI wins and fails, something I can consolidate and share back on a regular cadence. We are still figuring out the right format.

These are stages I went through. I am not saying they are universal. I am saying that the first one felt right and wasn’t, and I burned many months on it.

There’s one problem I keep running into, and I don’t have an answer yet.

The company invested real money in these AI tools. Enterprise licenses for an entire engineering org aren’t cheap. At some point, someone asks: what did we get for that investment?

And I don’t have a clean number. I can point to specific tasks that got faster, specific workflows that improved. I can tell anecdotes. But quantifying the aggregate impact of AI tooling across a team in a way that satisfies a finance review? To be honest, I haven’t cracked that. I am not sure most champions have.

Part of the difficulty is that we never measured the baseline. We didn’t track how long tasks took before AI tools arrived, so we can’t produce a clean before/after comparison. The ROI question is real, and I think it’s the next hard problem for anyone in a champion role.

I am now in a different team and standup.

When it’s my turn, I don’t announce tools anymore. I talk about what I built yesterday with AI, what worked, what broke, and how I fixed it. Sometimes people ask a follow-up question, sometimes they don’t. But the conversation has started differently this time, because I started it differently.

I spent months as a relay station before I realized the job was simpler and harder than I thought. Simpler because it comes down to one thing: let people see you use it. Harder because that means using it in the open, including the parts where you look like you don’t know what you are doing.

I am still figuring out most of this. But I am figuring it out in the room now, not behind a Slack announcement.