​

A few weeks ago, my company let go of an entire engineering team.

Not because the work stopped. Because the math changed.

(don't worry, we were able to offer them other roles inside the company or gave them the option to take a nice severance package)

The funds got redirected toward product development, and specifically toward technical depth in the places where we actually need more help executing. AI would cover the surface-level work. A smaller, sharper technical team would handle the rest.

I'm the one building the workflows to make that work.

And here's what that experience made clear: the headline everyone keeps writing, "AI is taking jobs", is technically accurate and completely misleading at the same time.

What's actually happening is a budget redirect, not a replacement.

Just this past week...

Sam Altman walked back the jobs apocalypse framing.

The Uber COO recently admitted they can't draw a line between AI spend and actual consumer features.

The Linear CEO said nobody can find the 10x productivity gain in revenue.

What companies are doing is something more specific. They're keeping the technical roles where execution is irreplaceable. They're cutting the middle layer of work that AI can now absorb. And they're asking the operators who remain to pick up the surface-level execution gap with AI as the tool.

That's a different problem than "your job is gone."

The new job description nobody posted.

The people who get ahead in this shift aren't the ones using AI the most. They're the ones who can do three things:

• Understand enough about the technical work to explain it to a customer or a senior leader
• Build or manage the workflows that cover the 80% of work AI can handle
• Know where judgment actually lives, the 20% where you can't outsource the call

I'm not talking about becoming a software engineer. I'm talking about being the person who can translate what the tool did and whether it was right.

That's the role that doesn't get cut. That's the role that gets pulled into more rooms.

The practical move this week

Most people try to automate their whole job at once and end up with nothing running.

Start with one workflow. Here is the exact process I used when building out systems for my own team after we restructured.

Look at your last two weeks of work. Find the task you did more than twice that followed the same pattern every time. Information retrieval, status updates, summarizing a document, formatting a report, drafting a first response. That is your target.

Before you open any tool, write down what the finished output should look like. Three bullet points. A table. A two-paragraph summary. Whatever it is, describe it in plain language. If you cannot describe the output, you are not ready to build the workflow.

What does the workflow need to start? An email? A meeting transcript? A spreadsheet row? Write that down too. You now have an input and an output. The workflow is just the bridge between them.

Before you automate anything, run the task by hand using AI. Paste in your input. Give the model your output description. See if it produces something usable. This step tells you whether the workflow is worth building or whether the task actually requires your judgment to do it well.

If the manual version works, now you can look at tools to run it without you. If the manual version requires too many judgment calls to be consistent, the workflow is not ready. File it. Come back in 90 days.

The operators who move ahead are not the ones who built the most. They are the ones who identified the right 20% first and then built backward.

What is one task in your world right now that you did more than twice this week the exact same way? Hit reply.

If this helped, forward it to someone on your team who is trying to figure out where to start.

​