29/07/2026
Most ERP AI conversations start with feature lists. The better question: which capabilities hold up once real data, approval chains, and messy records get involved.
One global survey found only 28% of medium-sized businesses report strong data readiness for AI initiatives — which starts to explain why so many rollouts stall regardless of the platform underneath them. AI amplifies what's already there; it doesn't repair a fragmented environment on its own. Fragile data doesn't quiet down once automation sits on top of it. It gets louder.
So where is it actually paying off? Not in the dashboards vendors demo. What we keep seeing hold up across implementations is narrower and less glamorous: compressing the repetitive work that makes ERP expensive to operate — reading attachments, matching records, drafting documents, routing exceptions. That's where the hours were always going.
Meaningful adoption typically takes 6-12 months after go-live — longer than most rollout plans budget for, and a reminder that this isn't a configuration checkbox.
The pattern that keeps holding up: the platform matters less than whether the foundation underneath it can already answer a straight question. Automate before that's true, and you're just automating the mess faster.