SAP Sapphire 2026 started on May 12 and INSI was present at the event following every announcement, every keynote, and every behind-the-scenes conversation. As a strategic SAP partner in Brazil, our role here is twofold: to understand firsthand where the world’s largest enterprise platform is heading and to translate that into concrete decisions for the clients who trust us to lead their digital journeys. This article will be updated daily with the event highlights.
The first day of SAP SAP Sapphire delivered a change that deserves to be read carefully.
The positioning shift
For years, SAP was described as the ERP company that was incorporating AI. In the opening keynote, this logic was reversed. SAP now positions itself as a business artificial intelligence company that operates solutions such as ERP, supply chain, CX, and HCM as application domains. This change is not just conceptual. It redefines how companies should think about architecture, governance, and data-driven decision-making.
At the center of this new positioning is the SAP Business AI Platform, which integrates three main layers:
- AI Foundation
- Business Data Cloud
- Business Technology Platform (BTP)
On top of this foundation, the Autonomous Suite was presented, with 224 agents and 51 active assistants distributed across areas such as finance, supply chain, HR, and customer experience.
The question the client needs to ask is no longer “will SAP have AI?”. It is “how will my operation coexist with a SAP that increasingly assumes decisions by default?”. Why this changes the architecture, not just the interface.
How AI starts to impact corporate architecture
The change presented at the event goes beyond the interface. It changes the systems architecture and the companies’ operating model. Three announcements support this transformation.
AI Agent Hub and agent governance
The AI Agent Hub, expected to go GA in the third quarter, will be the central governance point for SAP and third-party agents, including:
- discovery and cataloging
- verification and compliance
- observability
- lifecycle management
The solution will be included in BTP at no additional cost.
The strategic reading here is simple: SAP understood that an agent without governance becomes a regulatory liability, and it is offering itself as the central tribunal for automated decisions within the corporate environment. Whoever controls this hub controls the auditability of the operation.
Joule Studio 2.0 and AI development
Joule Studio 2.0, now renamed Tools Studio, is the open and model-agnostic development environment. It accepts SAP’s own models or third-party models, including Anthropic’s Claude, which was officially announced as available on the platform. In alpha tests with one thousand developers, SAP reports results ten times faster than traditional code generation approaches, with greater accuracy. Available to the first customers in June.
The central point is the openness of the ecosystem, reducing dependence on a single model provider.
Business Data Cloud and data federation
Business Data Cloud gained federation strength. With the intent to acquire Truven Health Analytics, SAP will enable support for Apache Iceberg, allowing agents to reason over data in any cloud without having to move it. It is the end of the argument that enterprise AI requires prior data consolidation. Semantic consolidation starts to happen at query time, not before it.
The detail few will notice
There is an architectural choice that was somewhat hidden in the keynote and is worth highlighting. SAP embedded NVIDIA Open Shell as a secure execution layer for agents. It is open source, contributed to the community, and serves to restrict what each agent can do, control access to data, and maintain traceability.
Why does this matter? Because the biggest pain point for those testing agents in a corporate environment is not the model’s capability. It is the impossibility of ensuring that it will not do something outside the scope. SAP is saying, in practice, that it has solved a critical piece of the production problem. And it is doing this in partnership with NVIDIA, not internally. It is a sign of maturity: SAP is not trying to own the entire stack. It is owning what matters, which is the business context.
The numbers that matter
Cases presented in the keynote already help separate promise from traction:
- KPMG modernized to S/4 Public Cloud for 270,000 users, with AI deployed in 34 countries and around 20 active agents. The stated goal is to generate 120 million dollars in savings in contract review alone.
- Takeda, operating regulated manufacturing, reported a reduction of up to 25% in revenue loss due to stockouts and up to 5% in safety stock using the Autonomous Suite.
- JPMorgan is migrating its general ledger to SAP and embedding agents with rules from SAP’s control framework, with auditable interventions. The message here is of another order: when JPMorgan puts its general ledger in SAP with agents, the regulatory ceiling of what is acceptable in corporate AI has risen.
And SAP itself announced an investment of more than 100 million euros to build agents in the ecosystem, with open data exchange with third-party agents at no additional charge.
These cases indicate the advancement of AI in highly regulated environments, raising the expected level of maturity for corporate adoption.
What this means for those deciding now
If your company is migrating to S/4HANA or is in RISE/GROW, three movements require decisions in the coming weeks, not in the coming quarters.
First, review the migration case in light of the Autonomous Suite. Classic brownfield remains valid, but with a new detail: each preserved customization is a closed door for a standard agent. The concept of clean core has stopped being technical hygiene and has become an AI prerequisite. One client presented at the event reduced from 18,000 custom code changes to 19 objects in the latest upgrade. That is the size of the difference between those who prepared and those who did not.
Second, treat the AI Agent Hub as an architecture decision, not as a tool. Whoever defines agent governance today defines tomorrow who can run what in the environment. It is a power movement within the company.
Third, look carefully at Rise and Grow with SAP. SAP is contractually committing to AI activation and adoption, including in ECC, with a package of more than 20 assistants activated at no additional cost for new GROW customers. This changes the economics of the decision for those still postponing the move away from legacy.
Day 2: from vision to execution
If the first day presented the destination, the second made the path clear.
Without a solid ERP foundation, the promise of autonomous AI does not hold. The central concept that connects this discussion is Clean Core.
Clean Core: from technical concept to an AI prerequisite
One of the main barriers to AI adoption in ERPs lies in the complexity accumulated over the years. At the event, it was highlighted that upgrades occur, on average, every eight years, and that each cycle can require up to six months of testing and remediation.
In this scenario, any AI agent ends up operating on an unstable foundation, with inconsistent data and unreliable processes.
Clean Core emerges as the answer to this problem, but with a new framing. It is not about eliminating customizations, but about governing how they exist. The goal is to rebalance IT investment, currently concentrated on maintenance, toward innovation in the processes that truly generate business value.
The five principles of Clean Core
SAP structured the Clean Core concept around five principles that guide this transformation. The first is Fit-to-Standard, which proposes keeping operations as close as possible to SAP standard, with adherence levels between 90% and 95%, extending only when there is clear business value.
Another key element is decoupled extensibility, which separates extensions from the ERP core to reduce upgrade risks and avoid the accumulation of technical debt. This decision shifts from being reactive to becoming strategic.
Data governance emerges as a foundational element. Reliable data is no longer just an operational requirement but becomes the basis for automated decision-making with AI. Without this principle, the quality of recommendations is compromised, regardless of the model being used.
Integration through APIs and events reinforces the need for a flexible and resilient environment, reducing the risk of disruptions with each update. Finally, operational efficiency emerges as a direct outcome of this structure, enabling risk reduction and faster adoption of innovation.
Clean Core has moved from technical hygiene to an AI prerequisite.
One customer reduced from 18,000 customizations to 19 objects in the latest upgrade.
AI-assisted migration
Day 2 also brought concrete advancements in how migration to S/4HANA is being approached.
The highlight was the use of the Data Management Assistant, an AI assistant that operates directly in ECC before the transition. Its role is to prepare the data foundation for a more standardized and governed target environment.
In practice, this means automating data quality assessment, performing deduplication, proposing corrections, and monitoring the environment after migration to ensure stability.
With this type of approach, migration is no longer an isolated event and becomes part of a continuous evolution process. Organizations arrive at S/4HANA with cleaner data, more consistent processes, and a foundation prepared to support AI agents in production.
What this means in practice
For companies migrating to S/4HANA or evaluating RISE and GROW with SAP, some decisions are no longer optional and begin to shape the entire strategy.
Clean Core becomes a prerequisite, not only for technical evolution but also to enable AI adoption. Agent governance becomes an architectural decision, with direct impact on control, compliance, and operations. And the economics of migration itself change as AI becomes embedded as part of the solution’s foundation.
SAP Sapphire 2026: two days, one clear message
SAP Sapphire 2026 built a consistent narrative across the two days. On the first, it presented the vision of the autonomous enterprise. On the second, it detailed the foundation required to make it possible.
The conclusion is straightforward: AI in SAP will not be limited by technology, but by the quality of the foundation companies build.
Without Clean Core, governed data, and standardized processes, the potential of AI is significantly reduced.
For companies running SAP, the coming months will be decisive in defining their level of maturity and competitiveness.
If your company is evaluating a migration to S/4HANA or wants to build an AI-ready foundation, the next step is to understand how these decisions apply to your scenario. Talk to our experts.