One of the interesting side effects of the explosion in coding intelligence is what it does to enterprise software. The moat of “this is hard to build” is gone. If AI can build it, you need a different moat.
That leaves three categories of enterprise software that survive.
The first handles incredible complexity at scale. Think of something like Epic Systems in healthcare. The complexity and scale required to operate at that level is still an enormous moat, even in this new age.
The second is built on a data asset rather than the software itself. The software might be easy to replicate. The data behind it isn’t. You could build a simple ad targeting engine in a weekend, but without Facebook’s or Google’s data, it won’t work.
The third is software too cheap to bother replacing. Even if someone could build it faster and better, the cost and effort of switching isn’t worth it for what you’d save.
And it isn’t just that building has gotten cheap in the hands of a reasonably skilled team. Most enterprise software, for most companies, comes with a lot of friction, because it was never custom enough for their specific use case.
Now you have the option to build something that fits exactly. If it doesn’t fall squarely into one of the above categories, why wouldn’t you just build it yourself?
