Enterprise software is entering a new phase of the AI era. The challenge is no longer whether large language models work, but whether ERP systems can support them reliably, securely, and at scale. That question points away from interfaces and algorithms and directly toward infrastructure.
Evan Goldberg, EVP of the Oracle NetSuite Global Business Unit, has been consistent on this point. NetSuite’s advantage, he argues, is not that it adopted AI early, but that it was designed from the beginning as a shared, multi-user system running on a unified data foundation.
“We built NetSuite as a system used by everyone in the company, across geographies, all working off the same database,” Goldberg says during an exclusive interview with ERP Today. “That architecture is what makes advanced AI possible in a business system.”
That architectural discipline becomes far more consequential as ERP vendors push beyond experimentation and into production-grade AI. Autonomous forecasting, anomaly detection, and cross-functional analysis require sustained compute performance, predictable latency, and consistent access to data. These are infrastructure problems before they are software problems.
AI at Scale Exposes Infrastructure Reality
Oracle’s recent expansion of AI-scale capabilities in Oracle Cloud Infrastructure (OCI) illustrates that reality. The company announced major investments in next-generation AI superclusters built to support persistent, high-volume workloads across enterprise applications. Unlike burst-oriented cloud designs, these environments are engineered for continuous operation—exactly what ERP systems demand when AI becomes embedded in day-to-day financial and operational workflows.
For NetSuite, this matters because AI features do not run in isolation. They sit inside transaction-heavy systems that process payroll, revenue recognition, inventory movements, and compliance reporting simultaneously. Infrastructure bottlenecks in those environments translate directly into business risk.
Goldberg frames the value of ERP systems in practical terms. “These systems don’t build your business for you,” he says. “They’re there to get processes out of your way so you can focus on what matters.”
That promise breaks down if AI slows the system, behaves unpredictably, or cannot scale with growth. In this context, cloud infrastructure is no longer just a delivery mechanism. It is designed for consistency, not just peak performance.
What This Means for ERP Insiders
Infrastructure is becoming a competitive differentiator. As AI moves into core financial and operational workflows, ERP end users will need to evaluate where and how those AI workloads run. Providers that rely on loosely coupled or opportunistic infrastructure may demonstrate innovation early, but they face real constraints when customers demand consistency, uptime, and performance at scale. Over time, infrastructure strength will separate those that can industrialize AI from those that can only showcase it.
AI readiness depends on sustained performance, not feature velocity. The market is shifting from pilot projects to production use cases, where AI operates continuously and touches sensitive data. This raises the bar for reliability and governance. ERP providers that optimize for demos or narrow use cases may struggle in the face of customers that expect AI to function as part of everyday operations, not as an occasional assist.
ERP buyers must rethink how they assess AI risk and value. AI risk in enterprise software is less about model accuracy and more about operational impact. Slowdowns, inconsistent behavior, or infrastructure bottlenecks directly affect close cycles, forecasts, and decision-making. Buyers should prioritize vendors whose cloud foundations are designed for transaction-heavy systems, because in the ERP ecosystem, infrastructure shortcomings translate directly into business failures.
