Key points
- AI amplifies data and process quality; it does not correct them.
- A clean core means standard processes, trusted data and deliberate extensions.
- Custom code increases the distance between the AI and the real process.
- Start with a readiness assessment, then fix the highest-value constraints.
AI does not fix bad data. It amplifies it.
Every AI use case depends on data and process. If the data is duplicated, incomplete or inconsistently defined, the AI will produce confident answers that are wrong. If the process is full of workarounds, the AI will learn the workarounds.
This is why AI programs that start with a tool demo often stall in production. The model is not the constraint. The operating model is.
What a clean core actually means
A clean core is an SAP estate where the standard processes do the standard work, extensions are deliberate and few, and the data is owned and trusted. It is not a ban on customisation. It is a discipline about why and where you customise.
- Standard SAP processes for commodity capabilities.
- Extensions only where they create competitive advantage.
- Clear ownership of master data and its quality.
- A current release, so new SAP capability is available.
- Governance that protects these decisions through delivery.
The three layers AI depends on
First, the data layer: master data, transactional data and definitions that everyone agrees on. Second, the process layer: how work actually flows, not how it is documented. Third, the architecture layer: which systems hold the data, how they connect and who can access them.
AI readiness is the weakest of these three layers, not the average. A strong model on weak data is still a weak outcome.
Why custom code is an AI problem
Custom code changes the process in ways the standard AI does not understand. It also increases the cost of every upgrade and makes it harder to adopt new SAP capabilities, including Joule.
Reducing custom code is not an aesthetic exercise. It shortens the distance between the AI and the real process, and it lowers the cost of staying current.
A pragmatic sequence
Start with a readiness assessment that covers data, process, architecture, skills and governance. Then fix the constraints that block the highest-value use cases, rather than trying to clean the entire estate first.
That sequence gives the business an early win without pretending the whole estate is ready. It also creates a repeatable method for the next use case.
Where to start
The SAP AI Readiness Diagnostic is a fixed-price way to get the facts. It produces a scorecard, a prioritised use-case list and a costed roadmap. From there, you can decide whether to fix the core, move to SAP Cloud ERP or deliver a specific AI use case first.