Do managers really like generative AI more?
Here is John Wang asking and offering conjectures about a good question: Why do executives like AI more than individual contributors do? His answer is that executives' domains are less deterministic, so they're more comfortable with non-determinism and more capable of wrangling it.
Wang only briefly mentions the latter point (that executives' training makes them more adept with AI), but it's interesting, both in itself and as an instance of a more general idea. What if some traits that make people good at traditional software engineering are not just less relevant to, but actually anticorrelated with, those needed to be good with AI? I don't buy it, but it's worth thinking through.
Here are some other reasons why managers1 might appear more enthusiastic about AI:
- It's their job to promote organizational priorities. In most places it's not so hard to say "I'm an IC doing my job, but I don't like AI very much." In many places it's impossible to do a managerial job well while also saying "I don't like this corporate priority."2
- Managers feel less professionally threatened by AI. It seems pretty obvious to me that a lot of what managers do is extremely replaceable by AI, but it's even more obvious that rank-and-file software engineers are more immediately in danger of replacement, at least as a matter of official organizational policy.
- Managers mostly do not see themselves as craftspeople. To the extent they see themselves as employing hard-won skills, they do not think AI makes those skills obsolete. They don't feel as much that their ways of life are threatened.3
- Managers are automating away a lot of pointless drudgery with AI. The first successes of AI-enabled automation have been better for managers than ICs, just from a hedonic perspective.4
- Software has long been unique in that ICs have more power, respect, and authority relative to managers than they do in other fields.5 Managers think that AI will make them more relatively powerful.
- Adopting AI is simply easier for managers than for software engineers. We need to overhaul all our ways of working, down to the basic rhythms of life. Managers are largely keeping the same schedules and doing the same things, but with new tools that are straightforward to learn, at least for their most common current managerial uses.
None of these are claims about how managers should use AI or what AI-assisted management will look like. They're just conjectures about why many managers are currently liking, or professing to like, AI more than many ICs are. Comfort with non-determinism is an interesting thesis, but it's pretty far down my list.
Wang says "executives;" I'm talking about "managers." The manager-IC distinction isn't perfectly exclusive and exhaustive, but it's a lot closer than the executive-IC distinction.↩
It's a bit more complicated than this; many tech managers obviously disagree with all sorts of corporate positions. Mostly, though, if something is important enough, their official position has to be in favor of it.↩
Whether or not they're right about this.↩
Again, I'm not sure this is true, but there are a whole lot of reports that, were it my job to write them, I'd be downright thrilled to get AI's help with.↩
This isn't the claim that AIs are more respected or powerful in software than managers are; it's that an IC, on average, gets more respect relative to their manager in software than they do in the average profession.↩