You probably don't need an AI Strategy

Right now the mad hype around AI is making everyone scramble and wonder

Am I ready for AI?
(or at the business/organization level) What is our AI strategy?

Of course the real question is: what is my biggest problem and can exploring AI get me there faster?

I’m a mid-30s dad, my goal for the year has been to exercise regularly because of back pain. But every month my daughter (who has just joined daycare) brings home pathogens that knock me out and kill my health and motivation. Short of cruel isolation, it’s just a risk I have to accept. Reminder apps and tracker apps have existed for decades. AI brings minimal impact to the core problem (my health situation, habits, and motivation).

In the world of operations, there is the concept of the bottleneck, the slowest step slowing the whole process. Let’s think about a law firm, you might think that the bottleneck is case document review but the actual bottleneck might be getting court dates! Reviewing cases 10 times faster with AI then doesn’t increase your throughput.

If your goals or your bottlenecks are not a problem that is amenable to AI then you simply don’t need an AI strategy.

Of course saying this openly is haram. AI theater is the norm. Especially for public software companies who are getting punished in the market if they don’t have a strong AI narrative. There is questionable merit to the stories of their demise; is AI a sustaining or disruptive innovation? Can’t existing companies simply incorporate AI the way Microsoft and Adobe transitioned their desktop products to the cloud? I won’t cover narrative building (formally called bullshitting by real professionals) here, will save that for another blog post.

While it is hyped as a general purpose technology, it is not there yet. Even something like paper and writing could only deliver mass scale utility once we had pens, ruled notebooks, templates, high literacy rates, etc. Until then it was just like parchment - a way of preserving old religious books; a narrow use case. The same thing will happen with AI. In time people will forget about tokens, models, etc. the way they forgot about dial-up, TCP/IP, etc. and just think about Facebook, Google, Twitter, etc.

Technology should rarely be treated as a general purpose thing; new technology is being invented all the time; cheap drones, EVs, CRISPR, etc. Most people never felt the need to have a strategy for those either. I’m betting most don’t need one for AI either.

My core recommendation is that people sincerely evaluate if they even need to do anything. Even if it does apply or has promise, the goal would be to play with it. Playing with tools is how actual R&D happens! People try all sorts of things, whatever clicks spreads. We do it for the joy of learning, avoiding any pressure to forcefully deliver results. But will management allow that?!