A non-programmer manager describes spending $2,000–3,000/month on multi-model AI setups (Claude Code, Codex, multiple reasoning models) that deliver 10x–50x productivity gains, arguing that white-collar workers are becoming replaceable and society is at a pivotal moment similar to early COVID.
So I have both 200 dollars subscription to Claude and 200 dollars subscription to OpenAI. I use Claude Code and Codex for coding related stuff/execution, discussions involve Astra or Opus 5.5 on the problem at hand at important intervals, we summon two indepedent Astra and Fable models for guidance (would love to add the third with Google Gemini 4.0). Double checking for mistakes/errors in progress as well as internet searches also come from individual models. And with this, I spend about 2000 dollars to 3000 dollars total on LLMs (including API usage) per month. I am not a programmer but I do have roughly 20 people working under me. Now, here is the thing. I don't know what your setup is like and I don't claim to have an optimal setup here either but this type of setup has me 10x to 50x my productivity compared to 6 months ago and also, compared to this process being down with other people who work for me (all M.S. degrees higher and Ph.D.s), it is so much faster and also more accurate. So while 2000 to 3000 dollars a month seems like a lot, if you compare that to a white collar worker's salary, it is nothing. And I suspect that this setup is better than having 5 people under me, so you can do the math. I feel like the whole society is right now going through the February 2020 beginning of the COVID moment. Only a few people are recognizing that in the past year or even half a year, these models have improved and crossed an important intellectual threshold line such that many people who are working right now across all industries are pretty much replaceable or very close to it. And the thing is, the type of setup I have is nothing. It will take most people 1-2 days to get comfortable with it and this type of setup, as opposed to mere vanilla of telling ChatGPT (e.g. make me this document, create this PPT) will soon be the standard to which the white-collar person has to outperform. I really think that anti-AI and pro-AI people should now about get together in one voice and voice the concerns. There is going to be a lot of problems in job markets in the next few years and no country is ready for this.
The article explores why American white-collar workers react strongly to AI, citing cultural emphasis on uniqueness, positivity bias, and the threat AI poses to their economic status and identity.
Companies across various industries are throttling employee AI usage and switching to cheaper models due to skyrocketing costs, with some spending tripling to over $15 million per month.
A BCG study finds that 74% of non-managerial white-collar workers regularly use AI, saving at least a day per week, but many companies struggle to convert efficiency gains into measurable value.
The author expresses skepticism about AI replacing white-collar jobs, citing a lack of evidence and personal experience in network engineering, while questioning the business case for AI adoption in various industries.
The article compares AI's potential impact on white-collar jobs to industrialization's effect on farming, arguing that just as farm automation drastically reduced agricultural labor demand, AI will similarly shrink computer-based white-collar work.