Running an Amazon business is 80% execution and 20% ideas. AI takes the 80%.
I’m standing in my kitchen, holding my phone out to my fourth child.
I have ten kids. At the time, I had nine. Every night we hand out small rewards, and every night she logs them in the allowance app we use. That app makes her do it one kid at a time. Open a kid’s account. Type 50 cents. Save. Back out. Open the next one. Nine times.
I open the app I just built and tap a few buttons.
All nine balances go up by 50 cents at once.
She looks at the screen. Then at me. She smiles.
“That was cool.”
The app I didn’t write
The app on my phone is called KidsBank, and I didn’t write a single line of it.
On February 17, 2026, I opened Claude Code inside Visual Studio Code for the first time. I have a computer engineering background, but I’m not a coder. Every time I tried to learn, the courses put me to sleep.
So I didn’t learn. I described what I wanted. Claude wrote the code, deployed it to the web and put it on my phone. I asked for a feature, and the change showed up on the screen. About ten days later, my family was using it.
The day before, building that app would have meant hiring a full-stack developer for one to two years. Somewhere between $100,000 and $200,000.
I did spend a lot of late nights those first weeks. But none of them went into code. They went into deciding what the app should do, and making sure I’d thought it through right.
That’s when I felt like the world had changed. It also made coding fun again, for someone who never actually codes.
The 80/20 flip
Here’s how I explain it. Work used to be 20% idea and 80% execution. That’s why companies needed so many people. Ideas were cheap. Carrying them out took a crowd.
AI is taking the 80%. People keep the 20%. And once execution stops being the constraint, you can act on five times as many ideas.
That’s my model, not a measurement. But on a call with a friend in October 2026, I said in passing that this was the biggest reorganization of work in a hundred years. He stopped me mid-sentence.
“That’s your next video. Since the Industrial Revolution.”
So I went looking for the numbers.
Why the comparison holds
The Industrial Revolution reorganized work because one input got cheap: power. Machines took over muscle. In England, the share of workers in farming fell from about a third around 1801 to under a fifth by 1871.
Epoch AI tracked what it costs to get the same quality of answer from an AI model. The cost of hitting a fixed score on a benchmark has fallen about 13x a year since 2023. From late 2021 to 2026, it fell roughly 190,000x.
For comparison, electricity got about 5% cheaper a year between 1892 and 1973. Batteries, computers and DNA sequencing all fell far slower than AI. No technology on Epoch’s chart has dropped in price this fast.
When execution gets 13 times cheaper every year, work reorganizes around the part that’s still scarce: the idea, and the judgment about which ideas are worth acting on.
What this means if you sell on Amazon
KidsBank was the first thing. It wasn’t the biggest.
Paying bills and reconciling accounts used to be close to a full-time job. Claude does that work for me, and it got there in four steps:
- I had Claude do the task once while I watched.
- I had Claude write a playbook from what it had just done.
- I put the playbook on a schedule, using a task scheduler I built.
- I gave it a browser, so it could work on websites it couldn’t reach before.
Between April and October 2026, that ladder produced about 30 finance playbooks. Claude still shows me the final review screen before any money moves.
Most of running an Amazon business looks like that list. Pulling the ad report. Adjusting bids. Checking stock against reorder points. Matching settlements to what actually hit the bank. Filing reimbursement claims. Each one is a recurring task that can climb the same four steps.
Then the question changes. It stops being how many people it takes to run the business, and becomes what you would test if running it cost almost nothing. More products. More listing experiments. More campaigns.
What this means for agencies
Agencies have it harder, because most agencies bill for the 80%. Campaign management, weekly client reports, listing updates: execution, priced by the hour or by the account. When the cost of execution keeps falling this fast, that pricing gets hard to defend.
The agencies that come out ahead sell the 20%: the strategy, the judgment about what to test, and the ideas a client wouldn’t have had. AI lets them deliver more of it without adding headcount.
Who actually becomes 10x
The obvious guess is that AI makes the best executors 10 times better. The research doesn’t say that. In one well-known study of customer-support agents, the least experienced gained the most from AI, and the top performers barely moved. On a fixed task, AI levels people out.
My own story fits. I’m not a better coder than I was on February 16, 2026. I still can’t write the code. What changed is that every idea I have can get built.
AI doesn’t make the best executors 10x. It makes the people with the most ideas 10x, because execution is no longer the bottleneck.
Where it shows up first
If this is right, it shows up first in jobs that are mostly execution.
Stanford’s Digital Economy Lab, working from ADP payroll data through June 2026, found employment for 22- to 25-year-olds in the most AI-exposed jobs running about 19% below their peers in less exposed ones. The researchers are careful about it. They don’t claim AI caused it, they see no economy-wide displacement, and the gap comes from fewer hires, not more layoffs.
That’s the pattern the 80/20 flip predicts. Companies aren’t firing the execution layer. They’re not refilling it.
And if it’s a bubble
It might be. The railroads were a bubble too. Britain’s Railway Mania of the 1840s was, by many measures, the greatest technology mania in history, and its collapse one of the greatest financial crashes. Investors lost heavily, and the railroads still reorganized how the country worked.
A bubble in the financing doesn’t cancel the change in the work.
Start with one task
Pick one repetitive, recurring task in your business, like the weekly ad report or the settlement reconciliation, and have AI do it once. Watch it work. Then have it write down what it did.
That’s step one of four. It’s also how mine started.
Back in the kitchen
My fourth child did the 80% every night, one account at a time, nine times over. Then it took a few taps.
She called it cool. I call it the biggest reorganization of work since the Industrial Revolution.
We were looking at the same thing.
