Most mid sized companies in Germany run SAP. Most of them have spent years migrating to S/4HANA, training their teams, and building processes around it. Now AI is everywhere — and the question I keep hearing is: "How do we get AI into our SAP?"
The honest answer is complicated. SAP has AI capabilities. But the path to actually using them is blocked by licensing tiers that most Mittelstand companies simply don't sit in.
The SAP Joule problem
SAP Joule is SAP's generative AI assistant — it lets users query their data in natural language, get explanations of transactions, and automate workflows. It's genuinely impressive when you see it demoed.
It's also restricted to specific S/4HANA Cloud editions and SAP BTP licensing tiers that most mid sized German companies aren't on. If you're running S/4HANA on premise, or on an older cloud edition, Joule isn't available to you without a significant licensing upgrade.
This isn't a secret — SAP's own documentation makes the prerequisite clear. But it creates a practical problem: your ERP has all your business data, and you can't use modern AI tools to query it without paying for a tier you're not on.
What we've been building instead
We've spent the last 18 months working on an alternative: a REST bridge layer that sits between SAP's OData APIs and a lightweight AI layer. The approach works like this:
- Expose SAP data via OData — SAP already has OData service capabilities (both v2 and v4). We configure the relevant business objects as APIs.
- Build a query translation layer — A small service that takes natural language input, converts it to structured API calls against the OData endpoints, and formats the response into human readable output.
- Add a lightweight frontend — Typically a web interface or a chat integration (Teams, WhatsApp Business) so business users can query their data without opening SAP GUI.
The result: business users can ask "What's our open receivables from customers in Bavaria over €50k?" and get an answer in seconds — without learning transaction codes, without opening SAP, and without the Joule licensing requirement.
What this is not
This approach is not a replacement for SAP. It's a read layer — most implementations are query only, with write operations reserved for trained SAP users through the normal interface. We're not bypassing SAP's data integrity controls; we're adding a more accessible entry point for the people who need data but don't need to be SAP power users.
It's also not a magic solution for messy data. If your SAP data quality is poor — duplicate vendors, inconsistent master data, missing classifications — an AI layer will surface those problems loudly. Which is actually useful, but can be uncomfortable.
Who this makes sense for
The pattern works best when:
- Business users regularly ask IT or finance for SAP reports that could be self served
- Management wants dashboards or KPIs but the data lives in SAP and the reporting setup is slow
- You have field teams (sales, logistics, service) who need SAP data on mobile without being trained SAP users
- You're running S/4HANA on premise or a cloud edition that doesn't include Joule
GDPR and data residency
This matters in Germany specifically. Our implementations keep data processing within the EU — typically on your existing infrastructure or German region cloud providers. The AI models used for query translation are accessed via API with no SAP data sent to third party training pipelines. We document this explicitly in the technical architecture so it can be reviewed by your data protection officer.
If you're running SAP and you're curious whether this pattern fits your situation, reach out. A brief technical call is enough to tell you whether it's worth pursuing — and if it isn't, I'll tell you that directly.