I’m starting to get feel queasy anytime I read about AI at workplace as a software product person.

It’s great that I can quickly research, vet and validate any idea in minutes thanks to Claude. Even prototype it and pull together an actual user testing campaign. In minutes. That’s great.

And so many of my colleagues & counterparts are doing a ton with AI as well to speed them up in various ways, making previously unscalable things scale.

But not all outcomes are created equal. Some folks are shipping faster, and with fewer resources, but not everyone is. There are plenty of reports of folks who end up spending more time checking their little AI friend’s work than they would had they just built it themselves.

And either way, I don’t see scalable software being built fast enough (outside of limited vibe code cases) to handle me spitting out any number of business strategies, product specs, test plans, “quick wins”, and so on with my little AI friend. I know for a fact not all the output is great, but it technically exists.

I worry that this is the expectation now. I don’t think this is good.

Just think - for just like 18 seconds - about how this plays out in practice in any not-small team setting:

• The engineers get slowed back down, needing to check every aspect of the AI’s work, even refactor significant chunks of what the AI spit out, then deal with conflicts arising from different solutions put forth by combinations of AIs and other human engineers. There are already tons of studies out there showing how meager the software development gains are for lots of software engineer
• This results in engineers not always working nearly as quickly as the “AI gurus” want you to believe.
• Therefore, all the PMs, designers, execs, etc. with great ideas that have been rapidly & thoroughly vetted by AI are simply sitting, waiting for their ideas to get done
• Everyone could vibe code their own solutions, testing them with real users, which is all good until you need to actually integrate those solutions - at which point everything gets stuck.

So it’s just as it was before, but with even more vetted ideas (read: noise) needing to fit through the development cycle. Software slows back to a crawl.

And instead of trying to deeply understand the problem, the execs, typically under insane pressure, just start laying people off.

Instead of thoughtfully considering the ways those tools are used by different roles. The outcomes that usage is realistically yielding. The ways their culture incentivizes people to work. The expectations they put on people. The expectations put on themselves. The realistic, durable, unique value the business provides.

Yes, AI will get better. But teams won’t without this.

This is my takeaway: Caring and thinking deeply matter a lot. Possibly more than ever, thanks to AI. And not just for certain functions or use cases. It’s everything. Not just so you can keep your job. But also so your job continues to be useful.