The main enterprise AI agent trends visible at SAP TechEd 2026 are open interoperability, governance sold as a product, business context as the differentiator, freedom of model choice with usage-based cost, and agents extending into physical operations.
The enterprise AI agent trends below combine SAP’s announcements with Datavapte’s reading of them. The announcements are facts from SAP’s sources, and the conclusions are interpretation.
Which enterprise AI agent trends stood out at SAP TechEd 2026?
| Trend | Evidence from TechEd | Planning implication |
| Open interoperability | Third-party agents invoke Joule Agents through a governed gateway; A2A and MCP supported | Design for mixed-vendor agents |
| Governance as product | Enterprise AI Gateway, AI Agent Hub, runtime verification, AI Governance Assistant | Budget for policy, not only models |
| Business context | Knowledge Graph, Ontology Manager (beta), Databricks inside Business Data Cloud | Invest in data definitions |
| Model choice and cost | 70+ models in Joule Studio; usage-based AI units for ABAP AI | Forecast cost by usage |
| Physical AI | Intel collaboration on autonomous robots; SAP-RPT planned for the International Space Station | Expect AI beyond screens |
Why are open protocols becoming the default?
SAP supports A2A and MCP, so an agent from another vendor can request work from a Joule Agent, as the TechEd announcement states. Joule Work also connects to Office365, Claude Cowork, Glean and ServiceNow. For buyers, the practical result is that no single vendor will host every agent, so identity, access and logging must work across vendors.
Is governance now part of the product?
Yes. SAP described agents as ISO and SOC compliant, runtime enforced and traceable through SAP AI Agent Hub, and said agents not verified there cannot be invoked, according to the keynote summary. The AI Governance Assistant produces an audit-ready risk classification for EU AI Act purposes. This lowers the effort of evidence gathering but does not remove the need for an internal owner for each agent. The agentic governance playbook sets out how to assign those owners.
Why does business context separate good agents from poor ones?
SAP’s executives argued that public models do not know a company’s processes or master data. The Knowledge Graph and Ontology Manager aim to supply that meaning, and SAP described combining records from several systems into a golden master record. Context is only as good as the records beneath it, which is why practitioners are returning to ownership and validation. Resources such as the economics of SAP data quality white paper and the ongoing SAP data governance use case address that layer.
How will model choice and cost reshape planning for enterprise AI agent trends?
With more than 70 models available, selection becomes an evidence task. SAP’s own claim of 50% lower cost per task from clearer requests suggests that how work is specified affects spend, though it is a vendor test result. In the ABAP area, SAP is moving from per-user billing to usage-based AI units. Finance teams will need monthly usage reviews, and the SAP AI playbook is one reference for structuring pilots.
What is the physical AI signal among enterprise AI agent trends?
SAP announced a collaboration with Intel to link enterprise workflows with autonomous robots, and plans to run SAP-RPT onboard the International Space Station, with no date given. These are early signals, not purchasable products today. Their relevance is that SAP expects agents to act on events outside the screen, which raises the bar for data accuracy and reconciliation, as in this reconciliation and audit readiness use case.
For groups that want a governance primer, Datavapte’s data governance guide is a starting point, and Datavapte’s note on governance decisions covers program choices.
FAQs
1. What are the main enterprise AI agent trends in SAP?
Open protocols for agent interoperability, governance built into the platform, business context from graphs and ontologies, wider model choice with usage-based cost, and early physical AI.
2. What are A2A and MCP?
A2A is a protocol for agents to communicate with each other. MCP is a standard for connecting models to tools and data.
3. Can enterprises govern agents from several vendors?
SAP offers a gateway and an agent hub for its own and verified agents. Cross-vendor governance still needs enterprise policy and ownership.
4. Does physical AI matter for SAP customers today?
Mostly as a signal. The Intel and International Space Station items were announced without dates.
5. How should enterprises plan agent costs?
Model costs by usage per use case, review actuals monthly, and request SAP’s pricing after the free period ending December 2026.
Yogi Kalra
CEO, DataVapte
Yogi Kalra is the CEO of DataVapte and a leading SAP migration expert with over 28 years of experience delivering zero-risk SAP transformations. He specializes in preventing data disasters during complex S/4HANA transitions and is the author of more than eight books on various modules of SAP ECC and S/4.
LinkedIn Profile
