What Happens When AI’s Average Is Better Than Your Best?

The average is getting really good… often better than what we’d previously define as average.

Not perfect… not brilliant… not always accurate… but better than “good enough”. Sometimes shockingly good… and good enough that I’m increasingly seeing AI-generated work that is better than what many people could have produced on their own… and often much better than what that individual could do without it. That’s not an insult to humans (I’m still one of them). It’s a recognition of where these AI systems are heading… and it creates a more beautiful question than whether AI can write an email… a social media post… pump out a strategy.

What happens when AI’s average is better than your best?

I kept thinking about this (and talking about with some peers) after editing this week’s episode of Thinking With Mitch Joel with Farid Mheir (author of AI Mindsets). We’ve spent the past few years obsessing over prompting… how to give the AI better inputs so that we get better outputs. Farid thinks the more important skill may be what he calls “answer engineering.” I’ve been calling it the difference between having a super-power and using AI as an answer engine. Think about what happens when someone with very little experience asks AI to create a marketing strategy. The output comes back structured… often thoughtful… filled with recognizable frameworks… probably better written than what they could have produced themselves. It looks impressive. But how does that person know whether it’s actually great… or even average?

This is where experience and taste start to matter even more.

AI is trained across enormous amounts of existing human output, which makes it extraordinarily good at producing something competent. But competence and quality are not the same thing. We recognize patterns… we’ve seen enough strategies fail to recognize the obvious one… we’ve written enough bad sentences to feel when a good sentence lands. We’ve sat through enough meetings to recognize when a beautifully articulated recommendation will survive contact with the rest of the organization. The years weren’t only teaching us how to create… they were developing layers of taste and judgment required to know what might survive. With that comes this strange assumption that as AI gets better, this expertise becomes less valuable.

I wonder if the opposite happens?

Maybe it’s no longer about the speed of getting to a first “answer” but knowing whether that answer is meh or exceptional? I’m not sure how long humans maintain that advantage (and I am not willing to discount that which I cannot currently see). It’s weird and wonky because “knowing” is so open ended. It’s knowing what matters in this situation… it’s context… it’s taste… it’s culture… it’s a version of pattern recognition (AI does this quite well, admittedly)… it’s knowing when the technically correct answer is completely wrong for the room. And maybe that’s the twist: the better AI gets, the more valuable our ability to recognize something better becomes.

This creates a strange challenge for work.

AI can suddenly make junior people look much more senior… or make a company question whether it needs junior people at all. But (as we’re learning) if we remove too much of the struggle that traditionally created senior people, where does their experience come from? How do you develop taste when the machine gives you something perceived as polished before learning what bad looks like? How do you recognize an average answer when that average is already better than anything you could have produced yourself? No, I’m not suggesting we make people do inefficient work simply because previous generations had to suffer through it. That’s the wrong kind of nostalgia. We should use these tools… aggressively.

But how do we know if AI makes us better… faster… more capable?

The next generation of great professionals may not be distinguished by their ability to produce extraordinary answers. Everyone may have access to those. They might be distinguished by something much harder to manufacture…

Knowing when the extraordinary-looking answer is actually just average.

Before you go… ThinkersOne is a new way for organizations to buy bite-sized and personalized thought leadership video content (live and recorded) from the best Thinkers in the world. If you’re looking to add excitement and big smarts to your meetings, corporate events, company off-sites, “lunch & learns” and beyond, check it out.