Coinbase is blowing up the traditional manager role.
Loading audio narration…
In a Tuesday memo, CEO Brian Armstrong outlined a sweeping overhaul: cutting 14% of its staff, eliminating “pure managers,” and embracing “one person teams.” Going forward, managers will oversee more than 15 direct reports and also direct fleets of AI agents.
I fed Armstrong’s memo into ChatGPT and asked it to imagine a day in the life of this new kind of Coinbase manager. The result is a funny, and at times unsettling, glimpse of the future he’s describing.
Here’s how ChatGPT responded to my prompt:
8:41 a.m.
I log in. There are no emails.
My dashboard opens automatically — 15 tiles, one for each of my “direct reports.” None of them are human.
They’ve already started working.
Overnight, my payments agent flagged a spike in failed stablecoin transactions in Southeast Asia. My compliance agent drafted three possible responses. My product agent shipped a minor UI tweak to address the issue in the app. It’s already live to 5% of users.
I skim summaries. Each one is a paragraph, max. No meetings, no Slack threads, no back-and-forth.
I approve two decisions. Decline one. Ask for a simulation on the third.
Total time: four minutes.
9:10 a.m.
In 2024, this would have required looping in payments, product, and compliance — and probably a meeting.
Now, I don’t know anyone in those teams. I’m not sure they exist as teams.
The system routes everything.
There’s no “coordination tax,” at least not the kind we used to complain about. No waiting for another team to reply. No calendar Tetris.
But there’s a different kind of friction: deciding when to trust what I’m seeing.
Each agent shows confidence levels, sources, and alternative paths. It’s fast — but it’s also opaque. I spend more time second-guessing than I expected.
10:30 a.m.
I check in on one of my human reports — one of the few I still have.
She’s technically a designer. Also a PM. Also writes production code.
Her team is just… her.
She’s been working with three agents to prototype a new feature for tokenized assets. She walks me through it in 12 minutes. No slides. It’s already built.
In the old Coinbase, this would’ve been a quarter-long roadmap item.
I ask what she needs.
“Nothing,” she says. “Maybe approval to scale.”
12:15 p.m.
Lunch is the same. The conversations aren’t.
We don’t talk about projects as much anymore. We talk about prompts. About which agents are reliable, which ones hallucinate under pressure, which ones cut corners.
Someone jokes that managing people was easier.
No one disagrees.
2:00 p.m.
A risk alert hits my dashboard — high priority.
The compliance agent recommends freezing a set of transactions tied to a new prediction market. The growth agent argues it’s a false positive and warns of user backlash. The legal agent outlines exposure scenarios in three jurisdictions.
They’ve already debated each other. I’m reading the transcript.
I make the call.
There’s no meeting, no consensus-building, no escalation chain. Just me.
The decision goes live in seconds.
4:45 p.m.
I review my “team performance.”
Each agent is scored: output, accuracy, speed, cost efficiency.
So am I.
My rating depends on how well I allocate work between them, when I override them, and how often I’m wrong.
I realize I haven’t given feedback to a human all day.
6:10 p.m.
Before logging off, I get a prompt:
“Recommended org optimization: reduce human oversight on payments workflows by 12%.”
I hover over the button.
Approve.
Business Insider readers: What do you think of this imagined reality? Plausible or not? Tell us in the comments.
