AI is making firms more similar

Learning will make them more different

Same, same, same, everywhere

AI is being sold aggressively, the costs are falling like a stone, and so it’s getting everywhere. Stanford University reported that inference cost at roughly GPT-3.5 quality dropped more than 280-fold between November 2022 and October 2024. So it’s not surprising that most businesses have it in one form or another. Having AI isn’t different.

That’s not all that’s getting rather “me too”. For all practical purposes, AI models are universally good at what they do, and similar. Not only have most firms got AI, but it’s almost always producing similar outputs, regardless of which AI model they bought. The choice almost doesn’t matter because each one is much like all the others.

The impact of AI is reducing the differences between firms as well. Numerous studies have shown that generative AI often gives the largest gains to novices and below-average performers, reducing the differences between them and experts. As an example, the Boston Consulting Group conducted an experiment among 758 people, and below-average performers improved by 43%, compared with 17% for above-average performers, on tasks within GPT-4’s capabilities.

Where should ambition focus?

So almost everyone has it, all of it produces much the same stuff, and it’s reducing the advantage of being a high performer. Ambition must look elsewhere for competitive advantage, higher performance, and enhanced value.

As competence loses superiority, the premium moves to better systems: capturing proprietary demand signals, unspoken operating knowledge, unusual customer insights, style, accountable trust, and owning a superior mechanism for systematic learning from outcomes and putting that learning to good use.

Data is where knowledge lives, so it had better be good. It isn’t understanding, though, that comes from careful analysis. Learning comes from observing what went well, what went badly, and working out why.

Learning how to learn

Differentiation begins there, but sustainable value only comes from using that learning to bring about change leading to improvement next time. Really serious, sustained and competitive performance enhancement comes from turning this into a business system that reinforces and compounds the improvements. Capture into the data the learning and the changes, and you have an improvement flywheel that continually extends your differentiation and value.

AI can play a part in systematic learning but it won’t make it happen. It won’t help with the design of that, either. AI is better at making things similar and aligned with what it knows. Perhaps this is where we need humans?