SAP announced its Autonomous Enterprise, and Oracle unveiled its Fusion Agentic Applications. Do you see yourself using these to automate cross-functional workflows in the next five years?
More than half of all SAP ECC customers are still waiting to migrate to S/4HANA, a platform released more than a decade ago, and a similar share of Oracle’s ERP installed base is still on legacy solutions like EBS, JDE, and PeopleSoft.
That’s tens of thousands of customers facing real capability barriers.
On-prem installs don’t play well with agentic AI because they’re highly customized legacy systems running on batch jobs, dependent on fragmented integrations, and suffering from questionable data quality. That’s a far cry from the clean, connected, and governed world AI agents expect.
This isn't an on-prem-only problem.
The same problems surface on modern platforms because migrations aren’t greenfield. A PwC & LeanIX study shows 86% of surveyed companies planned brownfield or mixed migrations to S/4 because doing so enables them to retain proven processes and customizations. SAP's own "clean core" push exists precisely because so few customers run anything close to standard.
Migration isn’t the problem here. Readiness is.
Customizations exist for a reason. Usually, it’s because the standard product didn’t meet business needs. User-defined tables and fields allow teams to capture detailed hierarchies, exception codes, and status flags that inform critical steps in workflows. Teams then rely on extraction workarounds, from robotic process automation to clunky macros, to pull data into spreadsheets and BI dashboards.
These workarounds are embedded in daily business operations like setting up new materials, creating POs, receiving goods, and paying vendors. Undoing this isn’t as simple as ‘ripping it out and going standard’. The time and cost required to remove all the custom work is just the floor. Teams also have to re-solve each problem the standard way, redesign workflows, and retrain users. SAP’s CEO, Christian Klein, has already publicly stated that plugging in AI will drive zero value without serious change management.
Even after customers cross these barriers, they’re still limited to automating workflows that SAP or Oracle can see and control.
I have yet to find a cross-functional workflow that respects these system boundaries.
Cross-functional workflows are typically exception-driven, and a lot of this data doesn’t sit neatly in tables. It’s scattered across emails, PDFs, spreadsheets, docs, and Slack/Teams threads. Automating these workflows requires capturing this data and turning it into required context users can trust — in time to drive the next decision, with full understanding of the SOPs, routing, and approvals involved.
This operating reality is the real gate to automating cross-functional workflows.
