🟡 5 MINUTE SUNDAY · Every Sunday. One tip. Under 5 minutes.
Cutting headcount for AI sounds like a cost cut. Most of the time it's not. It's a bet, and almost nobody checks if they won it.
Everyone I talk to right now is doing the same thing. Fire people, hand the work to AI, call it savings.
I'm a COO person. I love cutting costs. But not like that.
01 — ONE LEADERSHIP TIP
A founder called me last week. Puzzled.
"We have a consultant who advised firing 3 engineers and replacing them with AI. I did it."
Then the result. "Now delivery takes longer because the CTO has to QA and debug everything on the weekends."
Then the part that made me sit up. "The AI bills are astronomical. I'm not even sure it's less than the salaries I paid for the 3 engineers."
Nobody had run that math. Not once.
So I asked him some questions. Not advice yet. Just questions.
Have you actually compared the AI bill to the salaries it replaced? Line by line, not guessed. He hadn't.
Why did you hire those three people in the first place? He paused. No clean answer.
What was their actual output before you cut them? Vague. Feature velocity, roughly. Not a number he'd defend in a board meeting.
Is this genuinely one for one? Did anyone define what the AI does, at what quality, checked by whom? No. The consultant said "AI can handle it" and that was the plan.
Nobody told the CTO he was signing up to be the human safety net for three people's worth of output. He inherited it by accident, one merged pull request at a time, until his weekends belonged to the backlog. I havent talked to the CTO but i can imagine he doesn like his new weekend extra job, how long do you think he will stick around like that?
My guess? Not long…
And then you are left with a very fancy AI agents and a very high bill attached to it , but no smart people who will actually ship your software. Is that what you were trying to do? One person reviewing everything a machine produces eventually buries them and your company with it…
The Shift: Cutting a role isn't the same as cutting a cost. If your total spend didn't drop, you didn't cut anything. You relocated it, straight onto your CTO's calendar.
We were talking about a long-term contract at that point. I told him not to pay me for that. He needed one focused fix, not a retainer. Find the actual bottleneck, fix that one thing, and if it means rehiring for a role you just cut, you rehire.
02 — ONE COMPANY WORTH KNOWING
Meta ran this exact bet at scale, under a program called Project OT.
Code changes jumped 220% year over year once AI coding tools spread. Features that actually reached users rose only 36%. Major incidents climbed 40%. Time spent firefighting rose 70%. Employee sentiment dropped from 74% favorable to 55%.
Same lesson as the founder above, just with more zeros. More AI activity is not more value delivered. Someone always ends up cleaning up the gap. At Meta that was engineers. In a 40 person company it's your CTO on a Saturday.
03 — ONE THING TO ACT ON
Do this before you cut the next role:
- Compare the AI bill to the salaries it replaces. Actually compared, not estimated.
- Ask why you hired that person in the first place. No clean answer means the hire was the mistake, not the headcount.
- Write down their actual output, the number you'd say out loud in a board meeting.
- Confirm it's genuinely one for one, defined, at what quality, checked by whom. Not sure means wait.
Time required: 15 minutes
If you're about to cut headcount for AI, or you already did and the numbers don't add up, that's exactly the ops work I do with founders and CTOs: designing the AI transition properly so the savings are real. Let's talk →
See you next Sunday.
— Alisa
Fractional COO · Executive Advisor · Keynote Speaker
Helping founders and senior leaders scale themselves, not just their business.
This issue also ran on LinkedIn — join the discussion there
