SAP Joule now offers more than 40 specialized AI agents and 2,400 skills across finance, HR, procurement, and supply chain. Gartner forecasts that 62% of cloud ERP spending will flow toward AI-enabled solutions by 2027, up from 14% in 2024 (Gartner). The investment is real. The capabilities are advancing fast, and yet, six out of ten companies are not using SAP’s AI agents when migrating to S/4HANA, according to a survey reported by CIO. Only 3% of SAP customers are using Joule in production.
The technology is ready. Most implementation programs are not. AI applied to a broken foundation amplifies the problems rather than solving them.
The Foundation Problem
The promise of AI in SAP is straightforward: automate reconciliation, speed up financial close, predict supply chain disruptions, and reduce manual work across procurement. Gartner predicts that embedded AI in cloud ERP will enable a 30% faster financial close by 2028. SAP’s own Cash Management Agent is estimated to save finance teams up to 70% of the time spent on manual reconciliation. These are not speculative benefits. They are within reach for organizations running clean, well-configured S/4HANA environments.
The problem is that most organizations migrating to S/4HANA are not starting from a clean environment. They are carrying over years of accumulated customizations, inconsistent master data, and ECC workarounds. When AI tools are layered on top of that foundation, they do not produce better decisions. They produce faster, bad ones. A machine learning model trained on inconsistent data will generate inconsistent recommendations with high confidence. An automation agent running against poorly mapped processes will execute the wrong steps more efficiently. In one documented case, a mid-sized manufacturer using SAP’s native AI tools was still spending 30 to 40 hours per week on manual invoice reconciliation, not because the tools failed, but because the underlying data was inconsistent.
McKinsey’s research on the AI-ERP divide reinforces this point: for every $1 invested in developing an AI model, organizations should budget $3 for change management. The technology is the smaller part of the equation. The organizational readiness, including clean data, redesigned processes, and trained users, is where the real work lives.
The Hype Cycle Is Moving Faster Than Readiness
Gartner’s 2026 Hype Cycle for Agentic AI places agentic AI at the Peak of Inflated Expectations. Only 17% of organizations have deployed AI agents, but more than 60% expect to within two years. That is the most aggressive adoption curve among all emerging technologies Gartner measures. At the same time, Gartner predicts that more than 40% of all agentic AI projects will be canceled by the end of 2027, not because of technical limitations but because of escalating costs, unclear business value, and inadequate governance.
For CIOs and Program Directors managing S/4HANA transformations, this creates a specific kind of pressure. Boards and executive teams are asking about AI. Vendors are selling it. But the migration itself is already the most complex initiative most of these organizations have undertaken. Adding AI workstreams on top of a program that has not yet stabilized its core system is a recipe for what Deloitte’s 2026 State of AI in the Enterprise report describes as the pilot-production gap: only 25% of respondents have moved 40% or more of their AI pilots into production.
Where AI Actually Helps in a Transformation
None of this means AI is irrelevant to S/4HANA programs. On the contrary! It means the sequencing matters. AI delivers genuine value when it is applied to a system that is already stable, with clean data, well-mapped processes, and users who understand the baseline before automation changes it.
The key distinction is between AI applied to program execution and AI applied to business operations. The operational benefits, such as automated reconciliation, predictive analytics, and agentic workflows, require a fully configured, production-quality system. The execution benefits, such as automated testing of configured processes, intelligent data validation during migration, and anomaly detection in early integration testing, can start earlier. But only if the system they are running against is real and contains actual business data. These are not the glamorous use cases that appear in keynote presentations. They are the ones that save weeks of manual work in the middle of a program where time is the scarcest resource.
Getting the Foundation Right First
The gap between AI capability and organizational readiness is why the order of operations matters more than the technology selection. An organization that rushes to enable AI agents on an unstable system will generate noise, not insight. An organization that first establishes a clean, working S/4HANA environment, configured with real data and validated by its own users, creates the foundation that makes AI useful rather than distracting.
LeapGreat’s FrontLoad™ is built around this principle: produce a working SAP S/4HANA system with customer data before the traditional program even reaches Sprint 0, then refine it iteratively with stakeholder feedback. Because the system is configured with real data from day one, the data quality and process validation work that AI depends on happens as a natural part of the build, not as a separate remediation project.
AI Is the Accelerant. The Foundation Is the Fuel.
Enabling Joule agents early does not in itself create an advantage. A stable, validated system with clean data does. The 40% cancellation rate Gartner predicts for agentic AI projects has little to do with the technology itself. Most of those projects will stall because organizations are automated on top of foundations that were not ready. We wrote about the broader version of this problem in Why SAP Transformation Programs Discover Problems Too Late.
See how FrontLoad™ creates the foundation AI needs. Get in touch to see how it works with your data, or explore the FrontLoad™ approach in detail.
