ERP is an operating model, not just an application
SAP installations often connect finance, purchasing, inventory, manufacturing, human resources, and reporting. Over years, organizations configure workflows around their business, link the platform to other systems, and build controls that help keep transactions consistent. The result can be difficult to navigate, but it is also deeply embedded in how work gets authorized and recorded.
In “Why the World Still Runs on SAP,” Eric Zhou and Seema Amble argue that systems of record preserve data, controls, and institutional process. That is a useful explanation for why replacing an ERP platform is more than moving screens or copying tables. A migration can change business operations, reporting, access controls, and the audit trail all at once.
Why replacement carries real risk
Before a migration, teams need to understand custom code, data quality, integrations, user roles, and local workarounds. They must decide which processes to keep, simplify, or retire, then test that the new design handles ordinary transactions and edge cases. If the project misses a dependency, a failure can ripple across billing, inventory, payroll, or financial close. Even a successful implementation requires training and a plan for parallel operation and support.
That is why the right comparison is not “old system versus shiny system.” It is the total cost, risk, capability, and maintainability of the current setup against a credible future-state design. Sometimes replacement is justified; sometimes targeted modernization is safer and more valuable.
AI may add a new way to work with systems of record
AI assistants could help users find information, explain a process, draft a report, or prepare a transaction for review. A controlled interface can make a complex system more accessible while leaving the underlying record, permissions, and business rules in place. This depends on the integration: an AI layer must respect user access, identify its sources, log actions, and request human approval where a decision or transaction requires it.
Organizations should treat an AI layer as a new control surface. Start with read-only queries or low-risk recommendations. Test whether answers are grounded in current records. For any action that changes a record, define approval, rollback, and audit procedures. Keep the authoritative system clear so that employees know where an answer came from and which record governs.
Plan modernization around business value
Inventory critical processes and integrations before choosing a path. Measure support cost, upgrade constraints, security exposure, user friction, and the cost of downtime. Verify the vendor’s roadmap and support deadlines for the exact products and versions in use. SAP’s published maintenance strategy, for example, distinguishes between Business Suite 7 core applications and S/4HANA commitments; dates and options vary by product, so customers should confirm their own contract and release details directly with SAP.
Whether a company keeps, replaces, or layers AI on top of SAP, the goal is the same: reliable operations with controls that fit the work. The systems that persist are not necessarily the newest. They are the ones organizations can trust, govern, and evolve.