In the first part of this blog series, I described why the Autonomous Enterprise has become tangible after SAP Sapphire 2026. The vision is set. But there’s something essential that lies between a powerful keynote and real business value: the homework. Autonomous agents are only as good as the architecture, data, and processes on which they run. And that’s exactly what this follow-up article is about—not the destination, but the journey to get there.
1. Homework: Architecture, Transparency, and Governance (LeanIX)
You can’t manage what you can’t see. And before long, many companies will have hundreds of agents running—from SAP, Microsoft, Google, or their own frameworks. Without an overview and clear rules, this will quickly turn into a new shadow IT problem—only with significantly more autonomy. Governance in the Autonomous Enterprise is therefore not an optional extra, but the first essential task.
Technically, this is addressed by the SAP AI Agent Hub, which is built on SAP LeanIX: a vendor-agnostic system of record for every agent—regardless of who built it. The SAP AI Agent Hub inventories agents, links them to applications and business capabilities, guides them through a clear lifecycle from “proposed” to “decommissioned,” gives each one its own identity with defined access rights, and ensures that only verified agents run in production.
The fact that SAP is including it in the SAP Business AI Platform at no extra cost (General Availability, GA, scheduled for Q3 2026) and that the EU AI Act will take full effect in August 2026 makes one thing clear: There’s no getting around governance—especially in finance, HR, and procurement processes.
However, the tool alone isn’t enough; the work behind it is organizational. There are two specific things you should address now:
- Make approval processes agent-ready. Take a look at how requests and authorizations are approved today—and who is involved. Will these processes and people hold up when, in the future, it’s not individual employees but agents who are submitting an ever-increasing number of requests? Where someone manually approves requests today, you’ll need clear, partially automated rules tomorrow.
- Document your IT landscape and strategy transparently. Only those who have clearly documented their architecture and roadmap can integrate agents effectively. Check whether there are binding guidelines for integration and expansion—keyword: Clean Core—or whether every integration today is decided on a case-by-case basis.
In short: Create transparency regarding architecture and agents, define approval and authorization rules that can scale with your needs, and take inventory of your agent portfolio today.
2. Homework: The Path into the Cloud
Most new AI features are “cloud-first.” This changes the business case for transformation: It’s no longer just about the end of support in 2027 or cost avoidance, but about access to AI capabilities—a much stronger argument. Cloud ERP is thus less of an IT project and more of a ticket to autonomy, making it a key element of the Autonomous Enterprise.
The key is the “how”: Don’t migrate via “lift and shift”; instead, modernize with a Clean Core so that extensions coexist seamlessly with the standard and don’t block upgrades. On-premises solutions generally offer the same capabilities for integrating your own AI agents. The crux of the matter is the availability and functionality of the standard interfaces. This leads to an often-underestimated task: Make your key OData services Clean Core-compliant so that your own data fields and custom logic from SAP BTP can also be accessed seamlessly.
3. Homework: The Data Foundation (SAP BDC)
SAP CEO Christian Klein summed it up at this year’s SAP Sapphire: “No AI agent can compensate for a broken data model.” That is precisely why the SAP Business Data Cloud (BDC) is at the heart of the preparation. It creates a semantic data layer that spans SAP and non-SAP data—no silos, no data sprawl. And before an agent outputs anything, it verifies identity and access rights.
The SAP Knowledge Graph provides the managed ontologies that enable agents to understand processes, relationships, and context. It is this context alone that transforms a plausible agent into a correct one. The task at hand is uncomfortable but unavoidable: Consistently align your data strategy with the capabilities and growing importance of the BDC—ensure data quality, clarify data ownership, and reliably connect your data products. Those who cut corners here are building agents on sand.
4. Homework: Understanding Processes and Controlling Agents
Before you automate, you need to know what you’re automating. Process mining with SAP Signavio reveals how processes actually work—including the deviations that are often overlooked.
On the path to the Autonomous Enterprise, the key task is to prioritize correctly. Don’t look for processes where AI could be applied; instead, look for where the need is greatest. Where are error rates the highest? Where is scaling impossible due to staff shortages? Where are long turnaround times costing you actual revenue? That’s where the first agents come in—not with the most exciting use case.
Once the agents are running, you need to monitor them: Agent Mining reveals their behavior, compliance, and business value. Give each agent its own identity, clear access rights, and measurable KPIs right from the start. This way, governance isn’t a roadblock—it’s the prerequisite for deploying agents in critical processes.
Foundation before glory: If you do your homework, you can turn your vision into reality
The above-mentioned tasks may lack glamour, but… may lack glamour, but they have a lot of substance—because they determine success or failure. My clear recommendation: don’t start with the most spectacular use case, but with the foundation. First, ensure transparency regarding architecture and agents; then move to the cloud, establish a clean data foundation in SAP BDC, and implement a governance model that’s in place from the very beginning. These are precisely the steps our valantic experts guide you through—from enterprise architecture and data strategy to processes and agent governance. The vision is set. Now it’s all about who does their homework.
In the next part of our blog series on the Autonomous Enterprise, we’ll take a closer look at AI Agents & SAP Joule 2.0. Stay tuned!
Turn your vision into reality: Female Digital Impact
