From Agent Hype to Enterprise AI Architecture
Enterprise AI is the use of advanced machine learning and language models to automate, coordinate, and optimize core business processes across applications, data, and infrastructure in a way that is reliable, auditable, and aligned with organizational goals. At SAP Sapphire, much attention went to the Autonomous Enterprise, Joule Studio, and AI agents embedded into workflows, but SAP Labs US is already planning beyond this initial agent wave. Yaad Oren, Global Head of Research & Innovation and Managing Director of SAP Labs US, describes agents as an important phase rather than the final destination for enterprise AI architecture. Inside SAP’s research agenda, agents sit within a broader push that connects future AI models, new data foundations, evolving user experiences, and cloud ERP evolution. For CIOs and architects, the question is shifting from “How do I add an agent?” to “What enterprise AI architecture will support many agents, new models, and governance at scale?”
Post-Transformer Models: The Next AI Platform Shift
Today’s enterprise AI stack relies heavily on transformer-based large language models inspired by the 2017 “Attention Is All You Need” paper. SAP Labs US now expects another architectural shift. Oren says SAP is working with universities such as Stanford and the Technical University of Munich on post-transformer architecture that could power the next generation of enterprise AI capabilities. These post-transformer models aim to improve reasoning, efficiency, and multiagent orchestration beyond what current generative AI can provide. While not yet ready for customers, they are already shaping how major vendors think about future products and cloud ERP evolution. The implication for enterprises is strategic: AI roadmaps must be flexible enough to absorb new model classes, rather than locking all innovation into today’s transformer-centric designs or a single vendor’s agent framework.
AI Governance Frameworks for an Agent-Rich Enterprise
As AI agents spread across business functions, governance is becoming as important as model performance. SAP’s Business AI Platform focuses on building agents, giving them context and reasoning, and governing them end to end. According to ERP Today’s interview with Yaad Oren, SAP is investing in tools like Agent Hub from LeanIX to give companies a single registry of agents, including those not built by SAP. Signavio adds “agent mining” capabilities so enterprises can trace agent behavior, track exceptions, and audit generated data. Oren describes the need for a “blanket over all their agents” so organizations know which agents exist, who created them, and how they behave. AI governance frameworks will therefore include cataloging, monitoring, and policy control that extend across heterogeneous systems, rather than being limited to one platform or business unit.
Agentic Data Platforms and the Future of Cloud ERP
Enterprise data platforms are evolving into agentic data platforms designed to support fleets of autonomous services. In Oren’s six-part research agenda, the “future of data” recognizes that current industry platforms will not be enough. Enterprises will need synthetic data generation to train agents, improved data quality tools, richer metadata intelligence, and new ways to understand data created by agents themselves. This shift intersects directly with cloud ERP evolution: as SaaS applications change to accommodate agents, cloud architectures must optimize for agent orchestration, latency, and observability. SAP sees this as part of a larger move where SaaS is “not dead, but evolving,” with AI-native data services embedded into the core. For customers, that means decisions about data platforms today should consider how easily they can support agent telemetry, lineage, and governance tomorrow.
Preparing Infrastructure for the Next Enterprise AI Wave
SAP Labs US is tracking six technology areas that together outline the next enterprise AI wave: future AI architectures, data, user experience, physical AI and robotics, quantum computing, and cloud architecture. This broader lens signals how infrastructure must change to support more sophisticated systems than stand-alone agents. Robotics will need tight links to ERP processes so tasks are executable and auditable. Quantum computing, which SAP’s CEO has associated with the decade after AI, is already being explored for large-scale optimization problems in supply chain and logistics. Meanwhile, the future of user experience anticipates AI-native workers who expect conversational, immersive, and emotionally aware interfaces. For enterprise vendors and customers, preparing now means aligning AI strategies with these converging trends so today’s agent projects become stepping stones toward a more autonomous, data-intelligent, and AI-native application landscape.
