> to stave off a "build with Accenture/WITCH+OpenAI/Anthropic" disruption.

idk, i mean you could try to build in house, wait for it to be done, and hope it's correct with respect to your business which i would give about a 5% chance of success or buy saas and concentrate on implementation/migration.

AI or not, unless your business already has the teams and governance in place to manage custom software you're going to be tied to Accenture or some other firm indefinitely which will be expensive. Besides, something of the magnitude of a global ERP is going to be almost impossible even with GenAI. Writing the code and the technical architecture, where GenAI will help the most, is the easiest most straightforward part of a project like that. The real difficulty will be business process definition, alignment, and requirements gathering/refinement same as always. Finally, business doesn't stop while you're implementing something in-house, once it's done (which it will never be truly done) you still have to migrate to it which is another multi-year process. I think where most of these projects will end up is still paying for saas but then for certain processes you use the in-house system. ...So it's the worst of both worlds basically.