But I’ve been thinking about a different angle that’s way more messed up and likely where we are heading soon.
Just recently, I've read about companies that had monthly AI bills get into millions of dollars. Some smaller companies could really be facing bankruptcy if they don't cut costs. Firing 3 devs just to keep the lights on then stops looking like "greed" and looks like survival, basically avoiding going out of business.
You could make a moral argument "just stop paying for the AI and keep the human workers," but not many will want to work at a tech company today that doesn’t provide top tier AI tooling, so you kind of have to pay for AI...
So here’s the precedent this sets: Layoffs won't be framed as "AI is stealing your job" anymore instead they’re going to be framed as "we literally cannot afford both the tech and the people, so the people have to go."
It gives executives the perfect shield against bad PR. They can just point at a massive OpenAI or Anthropic invoice and say "look, it was either lay people off or we go bankrupt and then lay off everyone." It turns humans into the only thing that can actually be cut. (instead of cutting AI usage)
We're trapped in this weird loop where you need the AI to stay competitive, the AI costs a fortune, so you have to fire the humans who needed the AI in the first place or likely won't even take the job.
A premium developer setup—like Claude Code Max combined with Gemini Pro—costs under $300-$400 per developer a month. Compared to a standard developer's $10k+/mo payroll, that's a negligible rounding error (around 1.5% to 3%). The argument that companies are forced to lay off developers to afford their tooling seems highly exaggerated at this point.
wat. No. There are plenty of devs who hate how AI is changing the work, and would be thrilled to go back to the old ways. I am seeing so many arguments that amount to "We all have to use AI because... we all have to use AI." People start with the assumption that AI will take over and then use that to work backwards and prove that we all must use it for everything.
If anything, we're starting to see the opposite. I'm hearing more and more discussions from my clients that the increasing cost is not sustainable, and the increase in problems is not the result they were hoping for.
There is a strong argument that such clients aren't using agentic processes correctly. But at the same time, when I show them how to improve such processes, the bills go even higher.
We have not yet landed on the tools and processes that will make AI take over all work. Nor have we proven that such a scenario is inevitable.
Take a baseline as 5 devs costs 5 salary units and produces 5 output units.
AI does something like “for 1.25 salary units you get 1.5 output units” so it’s costlier on a per-dev basis, but you get 6 output units for 5 salary units, but only have the budget for 4 devs with AI. You would be a failing leader to argue for 5 output units over 6. That’s essentially the calculus that’s going on today. How much more output is gained from spending on AI compared to hiring more devs. Nobody has figured out how to measure it yet, but it’s clear that spending more on individual devs is the path to increased productivity.
You seem not to understand how companies work.