I work at a large non-tech company and for the last 6 months they’ve been pushing hard for adoption of GenAI, from CoPilot to Claude and Claude code. I finally gave in the pressure and jumped in feet-first at the beginning of this year. I can certainly see the appeal. I’m able to knock out tasks in record time, I can throw it at complicated coding challenges and it will come up with more complete solution than I would. Our team is outputting code and story points at a break-neck speed. My main focus, as an individual contributor, is to try and manage the firehose of code that can come out of the machine: to control the output into reasonable chunks that me and my coworkers can read and reasonably consume and understand.

But now, after this heavy push to adopt the tool, we’re starting to get serious push back on token consumption. I have a very reasonable token limit per month, but I’ve heard rumours they’re pulling those back, and starting people with only small amounts when onboarding. I just completed a half hour highly produced training on limiting token consumption, clearly someone is starting to look at the checks we’re sending to Anthropic.

This all brings me to the following conclusion: the company is deskilling it’s workers for improvements to clock efficiency and spending money to do so. I smell a trap!

First off: deskilling. This should be pretty clear, but in my case I’m substituting writing code for prompting for code (and then reading it, because I really am trying to keep myself sharp in some way). These are different skills, but one I can do with the tools built into a $300 laptop, the other I need a hundred dollar monthly subscription. The training on how to be efficient in token usage is another great example of this, we’re not training on the business we’re actually in, but instead in how to use the tool we’re paying for.

Clock time versus calendar time. This whole post was inspired by the coincidence of reading Pavlov’s blog post, then doing the training all in one day. The tldr: we’re improving individual efficiencies while ignoring the larger scale coordination needed to solve the actual hard problems that large companies are trying to solve. The thing I’ll add as someone in the trenches: the clock time efficiency looks and feel really good at the worker level. Our team is hitting numbers like no-one’s business. (Clock time is a good metaphor, but the scale is slightly off, we’re improving our two-week sprint output, not actual output hours, what no-one has proved yet is if that increase in story-point output will actually translate to a more quickly delivered system that will increase the metrics that actually matter).

GenAi is clearly changing the enterprise and corporate world as well as a lot of other sectors. I hate what it’s doing to visual design, and I loath to think what pop music sounds like a couple years. Some companies are chomping at the bit to lay off workers, and that’s gross and I suspect another trap. However, I think even conservative corporations who are looking at GenAi as a accelerator, not a game-changer are still playing with fire, and not one they can easily put out.