Back to Syllabi
Skills for CareersAI StrategyCareer Advice

The Hidden Tax of Being the AI Person on Your Team

6 min read

Get posts like this in your inbox weekly

Subscribe Free

A few weeks ago someone asked me an analytics question and I could not answer it.

That should not happen. Analytics is my bread and butter, the thing I teach and built a career on. They had been digging into agent harnesses and wanted to know how you would actually build one for analytics work, or whether it was even worth the effort. And here is the uncomfortable part: I knew what a harness was. I had watched the sessions, read the threads, explained the concept to other people. What I could not do was answer how to build one, because I never had. I had spent six months consuming this stuff and teaching it, and almost no time actually applying it.

That is the trap, and it is a quiet one. Listening to content feels like keeping up. Teaching it feels like mastery. Neither is the same as putting your hands on the work, and the gap between knowing about something and having done it is invisible right up until someone asks you to do it. If you are the person your team turns to for AI, this is the bill that comes due. Mine came a few weeks ago, and it told me exactly where the last six months had gone.

The thing that happened while I was helping

Back in January, people came back from the holidays gassed up on Claude Code. Everyone wanted in, and there was almost no shared knowledge anywhere. So I did the obvious thing. I started a community, ran trainings, answered questions. A lot of questions. North of a hundred users went through training directly, and the community answered hundreds more.

It worked, mostly. People who had been quietly afraid of what AI meant for their teams got to see a bigger picture than "this one task is done differently now." Hesitant users turned into capable ones. That was worth something.

But here is what I missed. Every hour I spent raising the floor was an hour I did not spend climbing. I was not falling behind because I was lazy. I was falling behind because I was being useful. Teaching felt like contribution, so I never questioned it. Nobody questions the thing that feels productive.

Then an advanced user asked me a question I should have owned, and I found the bill.

Teaching is frontier work disguised as a favor

There is a story going around that AI will flatten organizations. Everyone becomes a builder, the gap between junior and senior collapses, we all rise together. Parts of it are true.

But it skips something. In an org racing to adopt AI, the people who already get it become a resource. They get pulled into teaching, into Slack threads, into "can you just show me real quick." That pull is constant, flattering, and almost impossible to refuse, because every individual ask is small and reasonable.

The cost is not any single ask. It is that the person doing the teaching stops being the person doing the work. Your expertise came from applying things, and teaching quietly replaces applying with explaining. You keep consuming, you keep current on paper, but you stop building, and explaining a thing you have not built is a debt that comes due the moment someone needs you to actually build it.

Here is the number that makes it concrete. We trained well over a hundred people and answered hundreds of questions, and there are still thousands of users not at the baseline we want. Six months of real effort, and the gap is still enormous. That is the math of teaching at scale: the returns are real but slow, and the slowest, most expensive stretch comes last. And the baseline will not hold still. AI is moving faster than anything most of us have worked through, so the bar keeps rising while you teach to it. You are not filling a fixed gap. You are bailing a boat the industry keeps making bigger.

How to tell if you are paying this tax

This is the part for you, because if you are good at your job, you are at risk for exactly the reason you are good at it.

Watch the questions you spend your time on. For the last stretch, a lot of what landed on me was some version of "how do I use AI to clean up my email." Fine questions, but old-world ones. They are about doing the existing job faster, not about what the job is becoming. If most of your AI time goes to helping people speed up the old way of working, you are not on the frontier. You are running a help desk for it.

The advanced users were asking the agent-harness kind of question, and those were the ones I could no longer fully answer. That contrast is the whole signal. The questions you cannot answer matter more than the ones you can.

It is also worth being honest about who you spend the teaching on. Not everyone who is behind is behind for the same reason. Some genuinely cannot keep up right now, and that is the honest case: they are good at their jobs and simply do not have the hours to learn a new toolset every quarter. They need a real block of your time and will use it well. Others are not slow, they are unwilling, and the tell is that even when they reach for AI it is only to do the old job a little faster. Light-touch effort goes a long way with the first group. With the second, you can pour in a lot and watch little come back. You do not owe every hour to whoever asks loudest. You owe it to where it compounds.

None of this means stop helping people. Raising the floor matters. But generosity with your expertise has a price, it is paid in your own edge, and almost nobody puts that on the invoice.

Pay the tax on purpose

So here is the rule I am holding myself to now. Listening is not learning, and teaching is not doing. The only thing that keeps you on the frontier is putting your hands on the work, and that is exactly the thing teaching crowds out, because explaining feels productive enough that you never notice you stopped building.

Protect the block of your week where you actually build something, and protect it like a meeting you cannot miss. Not the block where you consume content about AI, and not the block where you teach it. The block where you do it badly, get stuck, and figure it out, because that is the only version that survives contact with a real question. The day you are no longer the most informed person in the room is the day being the AI person stops being worth anything, and you will not see it coming, because the whole time it felt like you were keeping up.

The teaching was worth it. I just forgot it was not free.

Justin Grosz

Justin Grosz

Product Leader | Adjunct Professor, Northeastern