Joule, Next-Gen Ariba, or Sap-native – which path for AI in SAP purchasing?
Joule, SAP Business AI, Next-Gen Ariba – anyone navigating the SAP world can no longer avoid these terms. The question on the minds of many purchasing managers and SAP decision-makers is: Which of these are actually relevant to us, and what do they require? According to the DSAG Investitionsreport 2026, only 3 percent of SAP companies actively using AI are making productive use of SAP’s own AI solutions. The reasons are varied: Cloud requirements, complexity of implementation, and licensing issues. This article provides a structured overview: What options are available, what each one requires, and where the limitations are.
Content
- Not all AI scenarios are the same
- AI in SAP purchasing – what's out there?
- Joule and Next-Gen Ariba
- Four reasons why SAP's own AI hasn't caught on yet
- The pragmatic approach — generic AI with SAP integration
- Data integrity & procurement intelligence
- Conclusion: Anyone working natively with SAP today makes decisions that go beyond just a tool
Key takeaways
- SAP Business AI/Joule works only in the cloud, as part of RISE with SAP. If you run S/4HANA On-Premise – as do 56 percent of the companies surveyed by DSAG – this option is not available to you at this time.
- Ariba Next-Generation requires that Ariba be an actively used platform
The AI agents run on BTP, not in the S/4HANA core — and the transition is a standalone migration project, not a feature rollout.
- SAP-native sourcing software delivers AI without an infrastructure project
No BTP setup, no cloud migration, no Ariba rollout required — yet it serves as the data foundation upon which procurement intelligence is built.
Not all AI scenarios are the same
You have five quotations in front of you – formatted differently, with varying payment terms, and one supplier has hidden the shipping costs in line 14. You’d like to sort this out in five minutes instead of an hour. This is exactly where AI is meant to help in procurement – and for tasks like this, it’s already performing reliably today.
Things get more complicated as soon as you need to make a decision based on your supplier history, ongoing framework agreements, or price trends over the past two years. That’s because not every tool that promises “AI in purchasing” actually has access to this data stored in your SAP system.
➡️ Operational AI scenarios like these don't require a vast ERP vontext:
- Summarizing request documents: Long item lists and technical specifications are condensed to include only the requirements relevant to the decision.
- Comparing quotes in a structured way: Prices, terms, and delivery times from quotation documents in various formats are made comparable.
- Flagging price anomalies in the comparison of quotations: Notable deviations from the expected value are automatically flagged—as a note for the purchaser, not as a decision.
- Drafting request texts and communication: Standardized supplier requests, rejection letters, and renegotiation texts are prepared based on guidelines.
As soon as a recommendation involves money or a supplier relationship, things change. When making a contracting recommendation, the AI examines the current quotation – but not the three delivery delays from the previous year, not the ongoing framework agreement in another category, and not the historical price trend from your ERP system. Without this context, the AI makes its assessment in a vacuum.
➡️ These strategic AI scenarios require full, real-time access to your SAP data:
- Spending analyses based on categories and time periods: What areas involve excessive spending, and where is there potential for cost consolidation? These questions can only be answered if the data of purchase orders, contracts, and suppliers is complete and consistent.
- Supplier evaluation based on historical performance: On-time delivery, quality metrics, complaint history — AI-driven assessments are only as reliable as the data from the ERP system. Missing or isolated data leads to inaccurate assessments.
- Benchmarking purchase prices against market data: Is the negotiated price a good one? To answer this question, you need not only the current quotation but also historical purchase prices, volume trends, and external market data for comparison.
- Reviewing contract texts for risks and discrepancies: In what ways does a supplier contract deviate from internal standards? AI-powered contract analysis requires the context of previous agreements, current guidelines, and compliance requirements, not just the document at hand.
AI in SAP purchasing – what's out there?
If you want to integrate Procurement Intelligence into your SAP procurement organization, the first question to consider is one of architecture: Which approach is even feasible in your environment? SAP groups its AI initiatives under the name SAP Business AI – spread across its portfolio: Embedded in S/4HANA, in the SAP Analytics Cloud, and as generative AI via Joule. In addition, there are generic AI tools that many companies are already using, as well as SAP-native sourcing software that operates directly within the S/4HANA data model.
The four approaches differ primarily in terms of the operating model they require and the data on which their AI functions are based:
Procurement intelligence for SAP: A comparison of suitability and limitations
| Approach | Core | Suitability | Limitaiton |
|---|---|---|---|
| SAP Business AI / Joule | Joule on S/4HANA Cloud | You use RISE with SAP or S/4HANA Cloud and want to apply AI across the entire SAP portfolio. | Cloud only (Public or Private Edition / RISE); not available on-premises; AI runs on BTP |
| Next-Generation SAP Ariba | Next-Gen Ariba | Ariba Sourcing is already in productive use at your company, and you want to integrate Agentic AI into the sourcing process. | Ariba must be actively used; AI runs on BTP, not in the S/4HANA core; high implementation effort |
| Generic AI with SAP Integration | Microsoft Copilot, ChatGPT Enterprise and others | You want to quickly implement operational AI scenarios — regardless of the SAP operating model. | There is no native SAP process context; it operates on exported data; no process execution takes place in SAP |
| SAP-native procurement intelligence | FUTURA Smart | You use S/4HANA and want to implement Agentic AI based on your ERP data — without migrating to the cloud and without an upstream infrastructure project. | Specializes in source-to-contract; designed for operational procurement (P2P) using SAP's built-in tools; no additional suite required |
The first three approaches provide AI capabilities and require that the data foundation be created elsewhere. The fourth approach first provides this data foundation – the AI capabilities are the result, not the starting point. This is a structural difference.
Joule and Next-Gen Ariba
SAP has outlined a clear AI strategy for purchasing: Joule, a generative AI assistant for every user, and Next-Gen Ariba, a completely rebuilt source-to-pay platform with embedded AI. Both are part of SAP Business AI – and both require that certain conditions be met in the system landscape
➡️ Joule: Using the SAP AI assistant in daily business
Joule is SAP’s generative AI assistant – embedded in S/4HANA Cloud, Ariba, Fieldglass, and other SAP applications. In purchasing, Joule supports navigation, supplier inquiries, and simple transactions in day-to-day Ariba operations:
- Navigation and Qqeries in SAP Ariba sourcing, supplier management, and Buying
- Bid Analysis Agent: Automated bid comparison, including total cost calculation
- Tender Analysis Agent: Extracting critical Rrquirements from submitted quote documents
- AI-based summaries of supplier feedback
➡️ Next-Generation SAP Ariba: Joule as an integral part
SAP has rebuilt the Ariba platform from the ground up – running on SAP BTP (Business Technology Platform), the cloud platform for integration, data, and AI infrastructure, with Joule as an integral part. The analyst firm, Ardent Partners, describes this as “a complete reengineering of the largest and most entrenched source-to-pay platform in the world.”
In Next-Gen Ariba, AI features such as the Bid Analysis Agent and the Tender Analysis Agent are embedded directly into sourcing workflows and run on a unified BTP data layer. Availability was announced in March 2026, and additional features will be rolled out quarterly throughout 2026 and 2027 – full parity with the previous version of Ariba has not yet been announced.
For existing Ariba customers, the transition is planned as a voluntary migration: Both environments can operate in parallel, but a BTP setup and a process migration are required.
In practice, this means that your sourcing team will maintain two environments in parallel for a transitional period – contract data and supplier history on one hand, and the new AI agents on the other. For every request, the first question is which system currently reflects the most recent status.
Four reasons why SAP's own AI hasn't caught on yet
SAP promises to speed up procurement tasks by up to 50 percent – but the user perspective paints a more sobering picture. According to Heise, barely a third of DSAG members rate SAP’s AI product as “good” or “very good” – and about 30 percent don’t feel confident in forming an opinion at all due to a lack of transparency. The reasons for this are by no means coincidental; they lie in the architecture of Joule and Next-Gen Ariba itself. Four structural hurdles explain why many procurement teams, despite their interest, are not yet using SAP's own AI productively.
1️⃣On-premises is out of the question for now
Joule is available only for S/4HANA Cloud Public Edition and S/4HANA Cloud Private Edition as part of RISE with SAP. At Sapphire 2026, SAP announced that it would make AI products available to on-premises customers in the future, but only in conjunction with a signed RISE migration plan. This is not yet available in production.
56 percent of DSAG survey respondents run S/4HANA on-premises (DSAG Investitionsreport 2026)
Small and medium-sized businesses, in particular, view this shift to the cloud as a source of pressure – for many, the economic viability of a RISE migration is disproportionate to the short-term benefits of AI, according to CloudComputing-Insider.
2️⃣ Ariba must be licensed and actively in use
Using Joule in purchasing requires an active Ariba Sourcing license. The AI features of Next-Gen Ariba also require the new platform itself. SAP Ariba Sourcing is not an extension of S/4HANA – it is a standalone cloud platform with its own data model. This has the following consequences for your day-to-day operations:
- Ongoing integration efforts across two data models
- Double licensing and operation costs for two parallel systems.
- Process discontinuities in day-to-day purchasing when data is not synchronized between the two systems.
For many SAP purchasing organizations, this effort outweighs the functional Benefits – read more about this in our article on SAP Ariba experiences.
3️⃣ Layer between AI and your S/4HANA data
Next-Gen Ariba collects supplier, contract, and transaction data on its own platform (SAP BTP) and makes it available to AI agents there. For you as a purchaser, this means in practice: The supplier evaluation or spend analysis that the AI shows you may be based on data from yesterday or last week – not necessarily on what’s currently in your SAP system. Your ordering history and goods receipts continue to run in S/4HANA and do not automatically end up in this intermediate layer.
For your IT colleagues, to provide some context:
- BTP acts as a middleware layer between S/4HANA and the AI functions; it must be implemented, configured, and maintained on an ongoing basis.
- With the new SAP Business Data Cloud (BDC), SAP is working to integrate S/4HANA data and external sources more closely and make them available through the Generative AI Hub. The architecture is becoming more powerful, but also more complex; BDC is still under development. This is critically examined in the e-magazine article “SAP’s Nonstop ERP Innovations”.
4️⃣ Underestimated complexity of implementation
In the demo, Joule appears to be a simple AI assistant that you can set up in an afternoon. However, setting up Joule Booster involves everything from configuring SAP access rights and user identities to integrating individual transactions and meeting specific minimum technical requirements for the S/4HANA system itself.
In practice, this involves a standalone infrastructure project: With its own budget, its own team, and its own schedule. Experience shows that it takes several months of preparation before the first query works. Anyone who plans this as a quick feature rollout is significantly underestimating the effort involved.
The pragmatic approach — generic AI with SAP integration
Microsoft Copilot, ChatGPT Enterprise, and similar tools are already licensed in many companies. In this context, it’s natural for the purchasers to ask, “Can’t we just use this?” For tasks such as summarizing quotations, drafting text for requests, or researching suppliers, these tools deliver quick, tangible benefits. You ask Copilot to compare five quotations – this works as long as the quotations are available as files.
The structural limitations become apparent as soon as an SAP process context is required:
- No native data access: Generic AI tools process data that has been exported from SAP or transferred via interfaces – with a time lag and a selective data set
- No process execution: Copilot can analyze and formulate, but cannot execute an SAP transaction or initiate a request process
- API layer instead of the SAP core: SAP offers standard integrations, such as Microsoft Copilot for S/4HANA Cloud Public Edition. Even there, Copilot remains outside the S/4HANA data model: The foundation is an API and interface layer, not the native SAP core
- Data sovereignty: Procurement data transferred to external AI services leaves the company – with all the compliance implications this entails for companies in the DACH region
Data integrity & procurement intelligence
AI recommendations in procurement are only as reliable as the database on which they run. Structural data integrity means that AI functions can access the same database on which operational processes run – without replication, without the need for synchronization, and without an additional platform layer in between. This is the foundation upon which procurement intelligence is built.
Procurement Intelligence refers to the ability to make procurement decisions based on interconnected data, end-to-end processes, and business context – supported by analytics, recommendations, and AI. It does not result from an AI model alone, but requires a unified data foundation in procurement.
SAP-native integration: AI without an intermediate layer
The architectural distinction between “SAP-integrated” and “SAP-native”is not specific to AI – it applies to any software used in the SAP environment. In the context of AI, however, this distinction becomes particularly relevant because data quality and data access directly determine the reliability of AI recommendations.
SAP-native sourcing software operates directly within the S/4HANA business logic – supplier evaluation, analysis of requests, and comparison of quotations are based on the same data as purchase orders, goods receipts, and invoice verification.
It doesn't require any separate infrastructure:
✴️ No BTP setup, no identity management
✴️ No data pipeline, no second data model
✴️ No separate AI system – the features are built into the software
Companies that use S/4HANA no longer need to carry out typical integration projects:
✅ No cloud migration required
✅ No Ariba rollout required as a prerequisite
✅ No need to spend months building infrastructure before the first productive use case
FUTURA Smart is SAP-native sourcing software for S/4HANA —designed for companies that want to lay the groundwork for procurement intelligence.
If you would like to know how this works in your specific S/4HANA environment, please contact us.
Conclusion: Anyone working natively with SAP today makes decisions that go beyond just a tool
SAP’s AI roadmap is ambitious. Joule and Next-Gen Ariba will transform the procurement ecosystem – for companies that are moving to the cloud. For the rest, the question is different: It’s not “which AI tool”, but “which architecture allows us to use AI productively today, without having to wait for a migration?”
Anyone who chooses a path is simultaneously making decisions about much more: How dependent they will become upon an external platform roadmap, how quickly they will receive strategic AI recommendations based on a complete dataset, and whether AI in purchasing will remain an infrastructure project or become an operational reality.
For companies using S/4HANA on-premise, SAP-native sourcing software is the only option that doesn’t require an upstream infrastructure project. And the decision has implications beyond the initial use case: Those working in the S/4HANA core today are simultaneously building the data foundation on which Procurement Intelligence is based—without a second system, without a second project.
AI procurement intelligence — directly in S/4HANA
FUTURA is the SAP-native sourcing platform for S/4HANA. It serves as the foundation for procurement intelligence in purchasing.
