The ERP Reckoning: Decades of Customization Debt Blocks Enterprise AI Value

For years, CIOs modified systems from vendors like SAP and Oracle to match specific business processes rather than altering operations to fit the software. [1] While this approach worked historically, it quietly generated significant undisclosed costs. [1] Today, the push for enterprise artificial intelligence is forcing IT departments to confront this legacy customization. [1] Angela Maragkopoulou, chief information and digital officer at Sunlight Group Energy Storage Systems, advises framing a clean core ERP strategy to boards as the price of admission to AI. [1] Grafting new AI capabilities onto an existing, highly modified system means quietly inheriting every flaw within that architecture. [1]

Despite the long-term competitive risks, widespread de-customization remains unpopular among IT leaders. [1] According to August 2025 research from IDC, 51 percent of organizations still prefer a customized approach, representing only a minor decrease from 52 percent in late 2024. [1] Mickey North Rizza, an enterprise software research VP at IDC, notes that moving to standardization is currently a strategic minority move concentrated mainly in large enterprises suffering from technical debt pain. [1] Similarly, James Baker, a partner at Bain & Co., observes that companies are not yet willing to fund wholesale de-customization efforts solely for AI enablement. [1]

The conflict between customized legacy systems and modern AI requirements reveals a looming structural disadvantage for enterprises unwilling to standardize. While leaders advocate for a clean core approach, the data from IDC and Bain & Co. shows that immediate business continuity and cost concerns still outweigh the theoretical future benefits of AI readiness for most organizations. The gap between large enterprises willing to absorb the high cost of de-customizing SAP or Oracle environments and those clinging to their legacy builds is set to widen. This dynamic suggests that the next phase of enterprise AI adoption will be constrained more by foundational IT architecture than by software innovation. What the sources don't address: How vendors plan to help highly customized clients transition to standard cores without disrupting core daily operations.