In enterprise software there is an eternal discussion of "buy vs build" and most organizations go through a cycle of:

-- we had a terrible time building something so now we're only going to buy things

-- we had a terrible time buying something so now we're only going to build things

-- repeat...

Either way you can have a brilliant success and either way you fail abjectly, usually you succeed at most but not all of the goals and it is late and over budget.

If you build you take the risks of building something that doesn't exist and may never exist.

If you buy you have to pay for a lot of structure that pushes risks around in space and time. The vendor people needs marketing people not to figure out what you need, but what customers need in the abstract. Sales people are needed to help you match up your perception of what you need with the reality of the product. All those folks are expensive, not just because of their salaries but because a pretty good chunk of a salesperson's time is burned up on sales that don't go through, sales that take 10x as long they really should because there are too many people in the room, etc.

When I was envisioning an enterprise product in the early 2010s for instance I got all hung up on the deployment model -- we figured some customers would insist on everything being on-premise, some would want to host in their own AWS/Azure/GCP and others would be happy if we did it all for them. We found the phrase "hybrid cloud" would cause their eyes to glaze over and maybe they were right because in five years this became a synonym for Kubernetes. Building our demos we just built things that were easy for us to deploy and the same would be true for anything people build in house.

To some extent I think AI does push the line towards build.