From Custom to Embedded: A Practical Roadmap for AI in SAP Environments

AI is in every business conversation today. But when it comes to SAP environments, the journey from ambition to real value can feel like navigating a maze. We see this every day: organizations are eager to harness AI, but the path forward isn’t always clear.

Recent research backs this up.

The SAPinsider 2024 AI Adoption Report found that only 19% of organizations have a formal AI strategy, and more than half haven’t even started shaping their roadmap. Meanwhile, Gartner predicts that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data.

So, what’s the missing link?

Most SAP organizations lack a clear AI roadmap; many AI projects fail due to poor data readiness. In my experience at Atos, it’s a practical, business-driven approach to data and AI — one that’s iterative, not overwhelming, and always aligned to real outcomes.

In this article, I have outlined a dual-track, use-case-driven approach: custom AI for ECC now, embedded AI for SAP Cloud ERP next. Here’s how we can make this work for your organization.

The Data Readiness Engine

Let’s be honest: traditional data transformation first programs are often slow, expensive, and disconnected from business needs.

As Gartner puts it, “data is never AI-ready in general, only for specific use cases.” Other experts, like TELUS Digital and Analytics8, agree that data readiness should be use-case-driven, repeatable, and incremental — not a one-time cleanup.

That’s why we advocate for what we call a data readiness engine. Think of it as a continuous cycle that powers every AI initiative. Here’s how we deploy it in each unique use case:

Assess what data is needed and where it lives.
Fix what’s missing or inconsistent — just enough for the use case
Integrate with modern data platforms like Snowflake, Databricks, or BigQuery.
Ensure alignment with SAP’s clean core principles.

This engine isn’t a one-off project. It’s a scalable, iterative capability. Every time you launch a new AI use case, your data foundation gets stronger, making the next project easier and more impactful.

Two Tracks to AI Value: Customized and Embedded

Now, these tracks can run in parallel, each building on the data readiness engine. Let’s take a closer look at how.

Track 1: Custom AI around ECC — Fast, flexible, and clean-core compliant

Many organizations are still running SAP ECC or hybrid landscapes. For them, custom AI is often the fastest way to unlock value, without the risk of disrupting core systems.

Here’s how it works.

Instead of modifying the SAP core, we build an intelligent sidecar — a modular AI layer that sits alongside your SAP environment. This sidecar connects via APIs like BAPI, OData, IDoc, or events. It then learns from SAP and external data like Snowflake or Databricks, and can execute actions back into SAP processes.

What does this look like in practice? Imagine this:

A spare-part distribution optimizer that predicts demand and reduces inventory costs
An outgoing goods inspection assistant that flags anomalies before they become problems
A supply chain disruption analyzer that spots risks early and suggests mitigation steps
A maintenance decision-support agent that recommends proactive interventions

The beauty of this approach is its flexibility. You can deploy AI agents quickly, see measurable results, and reuse these agents as you transition to SAP Cloud ERP. This protects your investment and reduces future migration effort. And because the sidecar operates outside the SAP core, you maintain a clean, stable system.

Track 2: Embedded AI in SAP Cloud ERP — The destination for intelligent enterprises

The future of SAP is embedded AI, where intelligence is woven directly into business processes. SAP’s own portfolio of SAP Joule, SAP AI Core, SAP AI Launchpad, and SAP Business AI is making this vision a reality.

But it is critical to remember that embedded AI isn’t plug-and-play. Activation requires careful preparation that includes the following:

Provisioning SAP Business Technology Platform (BTP)
Configuring entitlements and identity roles
Mapping to SAP semantic data models
Embedding AI insights into Fiori applications
Establishing governance and human-in-the-loop controls
Monitoring accuracy, drift, and explainability

With more than 27,000 customers using Business AI features, it is critical to note that deeply embedded use cases still require robust groundwork.

The organizations that succeed are those that treat AI as a structural capability, not just an add-on. They invest in the right foundations — data, governance, and integration — so that AI becomes part of their organizational DNA.

Building for Today… and Tomorrow

In my experience, supporting complex SAP transformations has taught me and my Atos team that the best results come from a dual-track strategy, powered by a repeatable data readiness engine. This approach is incremental, robust, and future-proof. It allows your organization to capture value now with custom AI, while steadily building toward the future of embedded intelligence.

As AI continues to evolve, the real question isn’t “How do we start?” but “How do we build AI into the very fabric of our SAP estate, so that every innovation strengthens the next?”

What is your next step on the journey from custom to embedded AI? And how will you ensure your organization is ready for what comes next?

>> Connect with me to discuss how you can optimize your SAP assets for a seamless AI adoption.

>> Find out more about how Atos is working toward a thriving, resilient and innovative future by unleashing the full potential of SAP: https://atos.net/en/services/smart-platforms/sap

Posted 04/12/25