Automate the Tasks. Keep the Glue.

AI was supposed to give you time back. It didn't. It gave you more work.

That part is real. So is this. AI isn't replacing you. It's highlighting you. The trouble is that most people have forgotten what it's supposed to highlight.

What doesn't go on the résumé?

You didn't forget who you were all at once. You forgot one bullet point at a time.

Say you spent ten years in commercial real estate. Leases, owners, brokers, deals. Then the AI boom, and a shot at project management with a solar developer. So you rewrote the résumé. The kitchen-table conversations became "managed client relationships." The owner who trusted you with a second building became "portfolio growth." The deal that only closed because you picked up the phone on a Sunday didn't make the page at all.

The page didn't have room for who you were. It only had room for what you did.

What you did got you hired. What you left off is why you're good at it. Your review comes down to a tracker. Permits, interconnection deadlines, study submittals, the weekly update. Every one of them has a date, and the date is what gets measured. The kitchen-table conversations are still part of the job. They're just called "RFIs and landowner relations" now, and they live in a single row with a status next to it. Green means on time, red means it isn't, and nobody asks how it got green.

Ten times what, exactly?

Somewhere above you, someone read that AI makes people ten times more productive. Nobody said ten times at what. The targets went up anyway. Two projects became four. So you did what everyone was told to do. You built a system. You handed AI the updates, the study summaries, the forms. You mapped it straight off the tracker, row by row, because as far as anyone could see, the tracker was the job.

The system looked a lot like your résumé. It treated every row the same, because the tracker does. It didn't know which studies you can skim and which ones you read twice. It didn't know that the utility contact goes quiet when they're behind, or that a short email gets an answer and a long one doesn't. Those are the shortcuts that made you good at this, and none of them were ever written down. So the system did everything the long way, and its drafts sounded like anyone's. You spent the time it saved fixing what it got wrong, working around a system that was supposed to work for you.

Then you noticed the one row that never went red. The landowner one. You went looking for the step in the system that kept it green. There wasn't one. There was a note on your phone, a list of names you call the week before every county meeting. Nobody here taught you that. You'd never shown it to anyone.

What if the system started with the note?

Measured by your tasks

If AI makes you feel replaceable, it's worth asking how you got there. It probably wasn't one decision. It happened slowly, in résumé bullets, status updates, and every time someone asked what you do. The tasks were always easier to name than the rest, so the tasks are what you named. After enough years, that's all you could see. Put a tool that's fast at tasks next to that picture, and of course it looks like a replacement. You didn't become your tasks. You just stopped counting everything else.

It's an easy habit to fall into, because the rest is hard to see. When you get good at something, the decisions behind it go quiet. You stop noticing them because you don't have to think about them anymore. A study in Academic Medicine found that experts explaining a task they know well leave out about 70 percent of the decisions it takes. They aren't hiding anything. They just can't see it.¹101Source: Sullivan et al. Academic Medicine, 2014 That's why so much of what makes you good never makes it onto the page.

What would you keep?

You can't see it. You can feel it.

Think about the last time work felt good. Not busy, good. A problem you saw coming before anyone else did. A conversation that turned a no into a yes. Something you made that you were proud of before anyone told you it was right. That's the glue. It's the part of the job that holds everything else together, and it's usually the part you like most. Write it down, even if it feels too obvious to count. The obvious stuff is the 70 percent.

Then look at what sits between those moments. The status update you write the same way every week. The form you fill out by copying from three other forms. The notes from a meeting that never get read. None of it is wrong, but none of it is glue either. It's the distance between the parts of the job you like.

That distance is where AI belongs, around the glue and not on it.

Here's what that looks like for me. I built a Claude Code skill called /span-hunter. It searches across our data and finds the hot paths, the routes people take through our site most often, so I can see where we're asking for more data than we need. Pulling the data is the part I hate. So I hand it off. I tell it to go get the data around the latest hot paths, and it does. Then it stops. What comes next is mine. Finding the story in the numbers, why it matters, what gets better for the people using the site, and how to sell that to the team. The system doesn't touch any of it. It just gets me there faster.

That's the difference between automating tasks and building a system around yourself. A good system knows where to stop. Every task you hand off shortens the trip back to the work that's yours. Less rows. More glue.

Try it this week

1. Write down three moments from the last month when work felt good.
2. Under each one, list the tasks that came right before it.
3. Pick one of those tasks and hand it to AI. See if you get back to the good part faster.