Blog

The Amazon problem AI couldn't fix

David McCandless
David McCandless
09 November 2026
The Amazon problem AI couldn't fix

Last week I wrote that I’d go back to Amazon if AI took the parts that taxed me. PRFAQs. Writing “crisp.” Forecast bridges.

I promised the other list. The stuff AI can’t touch.

Listen to this post on Substack.

Here’s the big one.

Amazon has two job families: tech and non-tech.

My last title was an L5 Business Intelligence Engineer. That’s the tech family. But I sat inside a business org, not a tech org.

Sounds like a minor detail. It wasn’t.

When my boss suggested it was time for L6 promotion, the way I remember it, I needed six people to write feedback on my work. Four of them had to be tech employees at L6 or above.

Here’s the problem.

In a business org, I barely worked with any tech people. Let alone four L6+ ones who’d seen my work up close.

The math didn’t work. For people like me, promotion was close to impossible.

This is the part that stung.

What I loved most about Amazon was how much people just worried about taking care of the customer. And 95% of the time, if you took care of your customers, everything else worked out on its own. Including your career progression.

Except for people like me.

Meme: angry Nicolas Cage labeled Amazon BIE next to laughing Pedro Pascal labeled Amazon Business Analyst. Both write SQL + Python to delight customers. Only one gets promoted.

Being a tech person embedded in the business meant I always worked on high-impact projects. I was a scarce technical resource, so I was high leverage for the business.

That exact combo is what stalled my growth. At least on paper, where promotions get decided.

It’s a vicious cycle.

A business org hires a few tech people.

They can’t get promoted, because there aren’t enough senior tech people to vouch for them.

So they leave.

So the org never grows its own senior tech people.

So the next juniors can’t get promoted either.

Repeat.

Which brings me to boomerangs.

A boomerang is an employee who leaves a company and later comes back. Often at a higher title and/or higher compensation than when they left.

It’s not rare. ADP Research found returning employees made up about a third of new hires in March 2025. In the information sector, it was nearly two-thirds. (ADP Research)

When the fastest way up is out, people take it. Get the bigger title somewhere else. Come back later at the level you couldn’t get promoted into.

Smart for the employee. Expensive for the company.

Where AI fits, and where it doesn’t.

AI can draft my promo doc. It can make it crisp. (See last week.)

It can’t conjure four senior engineers who’ve watched me work.

This one’s a structure problem. Not a writing problem.

To be fair.

Of all the publicly traded companies I’ve worked for, Amazon had by far the healthiest culture. The most focus on results. The least bureaucracy and the fewest dumb policies.

This was one of the exceptions. A policy that made sense in a tech org and quietly broke everywhere else. Hopefully it’s changed for the better since then.

If you lead a business team with a few tech people in it, ask one question: who’s qualified to vouch for them at promo time?

If the answer is “nobody on this team,” you already know how the story ends.

Take care of the customer and the rest works out.

Unless “the rest” requires four people who don’t exist.

Share this post

Let’s Turn Your Data into Growth

No more guesswork or missed opportunities.