You can’t define your professional self through tasks you’re handing over to AI

You can’t define your professional self through tasks that you’re consistently handing over to AI.

In the final stretch of prepping for my Advancing Research 2025 talk on “coexisting” with AI, I’ve been thinking about what advice I’d give to researchers, designers, and PMs, and I think it’s this.

In knowledge work, we all have different thresholds for what “good enough” looks like.

For example, on the subject of research synthesis (because it’s been coming up a lot):

Lots of folks think AI synthesis is totally adequate in many cases. Others have tried it and think it does an abysmal job.

One reason for differences like this is that AI is great at producing generic, average output (I heard Ben Jackson call AI a “cliché generator,” and I think that’s exactly how to think of it).

Average output is exactly that–average.

To a layperson, it looks fine. To a specialist, it’s lacking.

Ethan Mollick cites studies that say this is typical for AI–it elevates the work of novices and may actually reduce the work quality of top performers.

Average work may be perfectly adequate to keep bringing home a paycheck, but if you don’t value it yourself, it will not be valued by others in the long term.

This is why I think you should decide where to use AI based on what you value.

If research synthesis is drudgery for you, and you just want to move on to designing wireframes–then maybe it’s okay to use AI to automate it. Sure, it’ll miss some themes; it may be outright wrong sometimes, but novice researchers are wrong sometimes too (and have more guardrails as a result).

But if you love research synthesis; if you love pawing through quotes on a Miro board and discovering that one powerful theme that nobody else recognized; if you love building up trust with your stakeholders over time that your results are going to have a superior level of quality; then maybe you should be using AI for something else.

Of course, there's nuance here that can't be captured in a blog post.

But I think the worst thing you can do is succumb to external pressures to use AI for the things you want to be doing yourself, simply because everyone else is.

I’ll be continuing the discussion on Tuesday at Advancing Research 2025. See you there.