Finally together again: SAP TechEd 2025 – Day 1

Missed what happened on days 2 and 3? Read our report on SAP TechEd 2025 day 2 and SAP TechEd day 3.

What stood out this year was the keynote at the end of the day. That worked well. We were able to immediately relate the announcements to what we saw during the day in hands-on sessions and roadmap sessions.

Day 1: the experience

Beforehand, there was a lot of shuffling around with sessions. The hands-on sessions were full in no time. On site, it was old-fashioned busy. Long lines for badges, rooms that closed quickly. The acoustics in some rooms were mediocre. That made the first content blocks less sharp than desired.

Nevertheless, the feeling of “finally learning together again” prevailed. The quality of the content was strong.

In the morning, we dove into SAP Business Data Cloud. What stuck with us: data as a product is the guiding principle. The Data Product Studio makes creating and managing standard and custom data products a reality. We saw a strong use case in which Master Data Governance uses external sources to automatically create and enrich new business partners.

This was followed by a full session on Joule implementation and contextualization. We recognized the challenges that were discussed. Our approach proved to be fully in line with SAP’s best practices. In the corridors, we heard that improvements will be made quickly in a few areas of concern.

In between activities, we chose a hands-on session that offered little that was new. The lesson of the day was clear: anything with “AI” or “Joule” in its name is immediately full and provides the most new insights.

The keynote in context

The day ended on a high note with a keynote speech that underscored SAP’s AI momentum.

Key points

SAP Build will be directly integrated with development tools such as Cursor, Claude Code, and Visual Studio Code. Developers will be able to move from idea to working flow more quickly.

Joule Studio expands with more options for modeling and orchestrating your own agents. Includes links to tools such as n8n for process automation.

Business Data Cloud is being expanded with Snowflake integration, in addition to existing connections with Databricks and Google Cloud. This simplifies hybrid data architectures.

SAP introduced SAP-RPT-1, an enterprise relational foundation model focused on predicting business outcomes such as delivery delays and payment risks.

SAP is committed to making 12 million professionals AI-ready by 2030. This will be achieved through training courses and partnerships with Coursera, among other initiatives.

What we had already sensed during the day was confirmed in the keynote: AI is moving toward agents with governance. Joule is shifting from “copilot” to “contextual agent” that follows Sense–Reason–Act. First, reliable context, then explainable reasoning steps, and only then controlled execution.

This line is consistent with what we observed in the labs: document grounding, roles and rights as a basis, and full traceability for audit and EU AI Act compliance.

The common thread: data, agents, clean core

Three trends connect everything together.

1. Data as a foundation – Without data products, lineage, and access control, AI cannot be release-safe or explainable. The combination of Business Data Cloud, Snowflake and Databricks integrations, and a Data Product Studio gives organizations the means to make data reliable and reusable.
2. Enterprise-scale AI – At BTP, we saw how Generative AI Hub, AI Core/Launchpad, and connectors to enterprise data together form an orchestration layer for models, prompts, and tools. In development practice, an ABAP AI SDK and an ABAP client are added on top of this. ABAP teams deploy AI in a capable and controlled manner within existing CI/CD and quality controls.
3. Clean core and extensibility – SAP is refining the extensibility model with an A–D classification and clear guardrails. Anything that can be done side-by-side belongs on BTP (CAP, events, API-first). Anything that needs to be on-stack works within ABAP Cloud via RAP and whitelisted APIs. The goal remains the same, but the route is more concrete: lower TCO, predictable upgrades, and faster benefits from new S/4 and BTP releases.

What this means for SAP users

Three clear steps if you want insight into trends and direction.

1. Anchor clean core in the architecture – Explicitly define what is allowed on-stack within ABAP Cloud rules. Move the rest to side-by-side extensions with event-driven integration.
2. Make AI adoptive – Start with defined Joule skills and one or two agents in processes with demonstrable business value. Think of credit checks in order-to-cash or intake-to-purchase-order. Ensure governance via Control Tower-like policies with masking, RBAC, logging, and cost monitoring.
3. Invest in the data layer – Get data quality and lineage in order. Determine the division of roles between Business Data Cloud, Snowflake, or Databricks within the existing landscape.

In addition, costs and performance require explicit attention. Agents who combine generative models and external tools need FinOps discipline: budgets, alerts, model selection, and efficient prompt and tool flows.

Link this to explainability and risk assessments. This allows you to demonstrate to the EU AI Act what an agent does, with which data, and under which safeguards.

In development, we see ABAP Cloud and CAP as the standard set. RAP services for release-safe on-stack scenarios and CAP services for side-by-side. The effect is less technical debt and faster implementation of upgrades.

How we translate this into action

SAP is clearly moving toward a combination of clean core, agent-based AI, and a solid data foundation. That is why we are aligning our course with SAP’s own guardrails and roadmap.

For extensibility, we follow the A–D model: side-by-side on BTP as the preferred route with CAP, events, and API-first, and on-stack only within ABAP Cloud with RAP and whitelisted APIs.

In AI, we are adopting SAP’s “compliance-first” approach: adoption via Joule Studio and the Control Tower, with document grounding, roles and permissions, auditability, and cost observability, as emphasized in the keynote and labs.

When it comes to data, we focus on Business Data Cloud and data products, with the announced Snowflake connection alongside Databricks and existing connections. Lineage and access control are guaranteed.

For development, we follow the SAP tooling direction: integrations with Cursor, Claude Code, and Visual Studio Code in Build, as well as the ABAP AI SDK and ABAP client for the Generative AI Hub. This fits within existing CI/CD and quality controls.

In short: we only scale up where SAP’s reference models and governance are ready. We start with defined scenarios in high-value processes. And we measure the effect on release safety, quality, costs, and compliance in line with SAP’s direction.

What follows

Day 1 in Berlin was intense and promising. The timing of the keynote underscored what we already sensed in the labs: SAP’s course connects data reliability, agent-based AI, and clean core into an actionable strategy.

For us, the direction is clear. Those who focus on data products, release-safe extensibility, and AI with governance will gain speed without upgrade pain. You maintain control over costs and risks.

Tomorrow, we will continue to explore Joule Studio and the AI lifecycle at BTP. We will apply these insights to concrete assessments, pilots, and modernization projects for our customers.