
Meet the Authors
SAP Joule Studio supports agent development using more than 70 AI models, with new controls to monitor agent behavior, model usage and spending.
SAP’s planned Enterprise AI Gateway will centralize authentication, policy enforcement, routing, monitoring and auditing of enterprise AI traffic.
Select customers can run custom agents free on the SAP-managed Joule Studio runtime through December 2026; development and testing tools remain free under fair-use limits.
SAP used its TechEd conference to announce new tools for building AI agents, along with controls to track their activity, model use and spending. The announcements cover Joule Studio and Joule Work, as well as the business data those agents use. SAP Business AI Platform provides the tools for developing and governing AI across SAP and third-party systems.
Building Agents in Joule Studio
With Joule Studio, developers and business users can build their own agents or extend existing ones using no-code and pro-code tools. SAP says they can choose from more than 70 AI models. Customers in the Joule Studio early adopter program can also build and run n8n workflows, with broader access to follow. SAP has added controls that let developers and administrators see how agents behave, which models they use and what they spend. Joule Studio is accessible through Joule Work, where SAP also announced capabilities designed to plan and coordinate the steps in a business workflow. The company cited Agent2Agent, or A2A, and Model Context Protocol, or MCP, as part of its work to connect agents and systems. Select customers can deploy and run custom agents on the SAP-managed Joule Studio runtime for free through December 2026. The tools for building and testing agents and n8n workflows with AI-assisted development will remain free under fair-use limits, according to SAP’s announcement.
Controlling How Agents Work Across Systems
SAP said third-party agents can call Joule Agents through its agent gateway, which enforces access controls. It also announced SAP Enterprise AI Gateway, built on SAP Integration Suite that is intended to handle more complex requirements, including limits on request volumes, routing, protocol translation and advanced access control. SAP said it would centralize authentication, policy enforcement, monitoring, usage tracking and auditing of AI traffic. SAP also plans to make AI Governance Assistant available to SAP Early Adopter Care customers in November. The assistant uses existing information and guides agent owners through additional questions to produce an AI risk classification. SAP describes it as a tool to help customers meet EU AI Act compliance requirements.
Bringing More Business Data into AI
SAP is also expanding the data that agents can draw on. Announced changes to SAP Knowledge Graph will incorporate additional customer-specific information from SAP and non-SAP sources. SAP Business Data Cloud Ontology Manager, now in beta, lets customers add industry-specific concepts to SAP Knowledge Graph or build ontologies describing how their data and business concepts relate. Other data announcements include expanded lakehouse capabilities with Dremio, with support for Apache Iceberg and Apache Polaris. A federated query engine is intended to let customers access data across sources without moving or copying it. SAP’s expanded Databricks partnership includes deeper SAP Business Data Cloud integration, a new SAP Application Data Foundation and integration between Joule and Genie.
SAP also announced two models for working with structured data. SAP-RPT-2.0 will add anomaly and outlier detection and analysis through embeddings to its classification and regression capabilities. SAP-RPT-TS is intended to support forecasts of business metrics over time in SAP Integrated Business Planning without custom machine learning.
What This Means for SAPinsiders
Agent governance is becoming an architecture decision that needs a clear owner. SAP is adding a governed gateway for third-party agents to invoke Joule Agents, and its planned SAP Enterprise AI Gateway would centralize authentication, policy enforcement, routing and auditing for AI traffic. Organizations already running their own API or AI gateways should settle which layer keeps the authoritative audit trail for agent-initiated ERP transactions before multi-vendor agents reach production.
The free runtime window makes today’s pilots a cost-measurement exercise. For select customers, running custom agents on the SAP-managed Joule Studio runtime stays free only through December 2026. Teams should use Joule Studio’s new observability controls to track agent behavior, model usage and spend now, so 2027 budgets reflect measured consumption once the promotion ends. Teams evaluating Joule Studio should also assess whether an existing SAP agent covers the workflow, what needs extending and what requires custom development. Compare those options against integration requirements, governance responsibilities and ongoing runtime costs.
Agent accuracy will rest on master data work outside the developer team. SAP is building agent grounding on SAP Business Data Cloud and SAP Knowledge Graph, with a forthcoming Ontology Manager and Reltio master data management for customer, supplier and product records. Agents reasoning across domains inherit whatever duplicates and inconsistencies those records hold, so data owners should audit them and assign ontology ownership before agents scale beyond a single process.



