"Do nothing" rather than turbo-adopting AI

I just read Han Lee's The AI Great Leap Forward, an eye-opening comparison of the current promotion of AI to China's 1950s Great Leap Forward. Like a lot of discourse recently, it's managed to capture my thoughts in a way I've struggled to express.

There's a lot of questionable things about AI adoption at the moment, but professionally one of my biggest concerns is the adopt-at-all-costs narrative outlined in the article. We're not baselining, measuring or assessing, just using AI because we must and calling it a day.

Yesterday I went to an event promoting an AI-based proof-of-concept for searching content (I'd tried the proof-of-concept in advance and been underwhelmed). Ostensibly this tool is solving a discoverability problem, but the context is so young that it's not a real problem that people have. We've jumped to the most complicated and expensive solution without trying the years of well-established working solutions already available (and a lot cheaper).

I raised this in the Q&A, how do you know it's worth the cost compared to something much simpler like token-based searching. The answer I got back was a vague summary of the value of AI search from an industry expert, not a practical explanation of A/B testing or iterative development.

All of which is to say, it reminds me that I'm lucky to work in a department of actual professionals who care about their work. I've been in a few planning sessions recently where we've had to decide what to develop next: every one of them has had a "do nothing" option, sometimes we've decided—on balance—that it's the correct choice. We've done the right thing for our users, saving them money by not wasting our time on unnecessary functionality.

Lee's article contains a picture that I think well summarises a lot of people's expectations from AI:

I'm thankful to say this isn't the experience I have every day, where we do genuinely work as a team and recognise each other's skills. I think it's because our approach is genuinely to provide solutions to users, not just to build software.