Artificially Intimidating

Artificially Intimidating

Deep Dives by Artificially Intimidating

7 AI Agents Got Bank Accounts: Why Digital Interns Burn Cash Without Scoping

Seven bots, 72 hours, $12,431 invoiced, zero revenue. The fix is the same one that works on a brand new hire.

Nicholas Rhodes's avatar
Nicholas Rhodes
Sep 15, 2026
∙ Paid
Watercolor of a man in an orange cap holding out a blank sticky note beside a row of machines spooling paper into a heap, with an invoice reading $12,431 and an open ledger reading $0.
Seven agents. Seventy-two hours. One very expensive pile of paper.

Think of the last thing somebody handed back to you that was technically exactly what you asked for and also somehow completely useless. The deck with fifteen slides and no recommendation, or the lead list full of people who will never buy.

They did what you said.

You just said it badly .

Now hand that same instruction to something that never gets tired and never asks a follow-up question.

A viral experiment recently gave seven autonomous AI agents funded bank accounts and a single directive: go make money.

They had 72 hours, full internet access, and permission to start businesses and invoice real people.

The final scorecard:

Go Deeper:

Seven AIs Got Bank Accounts. They Invoiced Strangers $12,431. — the daily-brief version of this same experiment, with the other four stories from that morning.

OpenAI’s Agents Found a 25-Year-Old Wiki and Started Passing Notes — what agents do when nobody gave them a boundary. The companion piece to this one.

A brain coach says you’re surrendering, not offloading — the other half of the delegation question. This post is about what you hand the machine. That one is about what you should keep.

User's avatar

Continue reading this post for free, courtesy of Nicholas Rhodes.

Or purchase a paid subscription.
© 2026 Nicholas Rhodes · Privacy ∙ Terms ∙ Collection notice
Start your SubstackGet the app
Substack is the home for great culture