Bill Gates wrote an alarming 6000-word blog post about AI, and a lot of people are talking about it.
In it, he warns that AI could be creating one of the most tumultuous periods in human history (one that could take more than a decade to get through).
It’s a bit of a scary read. He covers how quickly AI is advancing and the fact that many AI companies have already gone beyond the safety limits they once promised they would keep.
Both the essay and an interview he did afterwards with MIT Technology Review are worth a read. But the paragraph that stayed with me wasn’t one of the frightening ones. It was Gates’ answer to a different question:
What can’t AI do?
Gates talked about his friend Warren Buffett, one of the world’s most successful investors. From the age of about 13, Buffett became obsessed with figuring out what businesses are worth. More than 80 years later, he holds an enormous amount of knowledge in his head that no one can easily copy.
Buffett has shared plenty of his ideas over the years. But there’s no way to capture everything that goes on in his mind when he makes a decision (the judgement, thought process and instincts he’s built over a lifetime). So there’s no complete record for AI to learn from.
Right now, you have your own version of this.
You know which enquiries to walk away from, and you know it within 90 seconds of the phone call. You know which jobs always take longer than the quote says. You know that when a particular client uses the phrase “just a quick one,” it’s never a quick one. You know who to put in front of a nervous customer and who to keep well away.
And I’m going to guess that none of that is documented anywhere. It’s in your head, another leader’s head, or possibly the heads of many people on your team.
That’s both good news and bad news.
It’s good news because this kind of accumulated judgement is genuinely one of the hardest things for AI to replace. It isn’t a simple task that AI can learn and copy. It’s years of solving problems, making mistakes and learning what to notice until eventually, it becomes instinct.
And it can also be a problem.
Because so much of what makes your business successful isn’t stored in a document or a system. It’s stored in your memory and the memories of others on your team. If your longest-serving person retired next year, much of that knowledge would walk out the door with them.
So here’s the cheerful conclusion I’ve landed on:
The most useful AI-related thing you could do this year probably has nothing to do with AI. It’s writing down what your best people actually know.
If you want to get started, here are two questions I think are worth asking your team:
- What do you know that a new employee would have no way of finding out?
- Who is the only person who knows it?
One honest caveat: Gates says AI is “not there yet” when it comes to capturing our accumulated human judgement. He doesn’t say we’ll never get there.
I’ll take his word on that. But “not yet” is a decent amount of time. And regardless of what AI can do in five or ten years, writing down your best people’s instincts, ideas and ways of making decisions is always worth doing anyway.
Moral of the story: Your most valuable business knowledge may be the stuff that only exists in someone’s head.
