Meet the Authors

Key Takeaways What you need to know
  1. SAP built its TechEd 2026 announcements around Joule Work, and Ted Way says it starts from what people want to accomplish and not from the tool they have to learn.

  2. The SAP Knowledge Graph, due in the first quarter of 2027, and the Ontology Manager are designed to draw on customer-managed data products, so the data work done now is what those tools will have to work with.

  3. Governance should be set up centrally and early, Way advises, and human approvals protect a business only when the people reviewing them examine what they approve.

SAP centered its announcements at SAP TechEd 2026 on Joule Work, its AI interface for getting business work done across SAP and non-SAP systems. A new agentic capability lets it plan and carry out multi-step tasks, and SAP is rolling out Joule Work in phases.

Ted Way, vice president and chief product officer of SAP Business AI Product Engineering, tells SAPinsider the thinking behind Joule Work starts with what people want from their software. “People don’t want to learn a tool. They want to accomplish a goal.”

SAP framed TechEd around moving enterprises from AI experimentation to autonomous execution. The change for customers, Way says, is that AI “stops being something you configure and wait for.” Agents carry out the work, with people in control.

Explore related questions

Work Starts With the Goal

Way came to SAP after 15 years at Microsoft, where his teams shipped some of the first generative AI use cases in business applications and “always started by partnering with the customer and digging deep until we understood the true pain points.”

Way describes what sets SAP apart. “At SAP, the stakes are different. Our AI is embedded in systems that run a company’s finances, supply chains, procurement, HR, and sales. Getting something wrong means the books are affected or production is delayed.”

The first lesson of his time at SAP has been to work from the user’s intent, which he calls “exactly the idea behind Joule Work.” Way points to Spaces, a Joule Work feature that SAP CTO Philipp Herzig demonstrated in his TechEd keynote. A Space builds an interface around what a user needs when they ask for it. Someone could create one to track revenue, share it with colleagues, and return a week later to find it updated with current data.

Way sees the old model, in which users depended on screens from software vendors that “somebody thought would be useful,” as “basically coming to an end.”

Joule Studio, SAP’s environment for building and extending agents, brings that goal-first thinking to developers. A developer states what they want to build, Way explains, and Joule Studio “knows the business context of where the solution fits in the business process,” then identifies the data and authorizations involved. With that context built in, work that previously took two days can come down to an hour, in Way’s estimate.

At TechEd, SAP announced role-based access in Joule Studio that gives IT teams more control over how contributors take part in building agents.

Context Decides What an Agent Can Do

Giving agents more work to do also means giving them enough information to understand the business. Way names that as one obstacle to autonomous execution, because much of what agents need to know is “scattered across systems.”

Access to the data does not settle it.

Way’s example is “critical supplier,” a term that “may mean different things across different companies.” An agent has to know which meaning applies before it can act on the information, and that is why, in his words, “you don’t want generic AI, you want useful AI.”

Way points to the SAP Knowledge Graph as part of SAP’s answer.

It supplies the business context that Joule and its agents draw on. The graph initially held data mainly from specific SAP applications. At TechEd, SAP announced plans to extend it to data unique to each customer, from SAP and non-SAP sources, with general availability planned for the first quarter of 2027.

Where the SAP Knowledge Graph supplies the broad business context, the SAP Business Data Cloud Ontology Manager is meant to focus that context on a particular job. Developers will be able to choose what a custom agent needs to know and publish views built around a specific business problem, drawing on how their company operates.

Way says the Ontology Manager, along with lakehouse and master data capabilities in SAP Business Data Cloud, is meant to give agents ‘a unified, trusted, semantically rich foundation to work from.”

Autonomy Grows as Trust Is Earned

Once agents collaborate with other agents and outside tools, security, compliance, and cost become “difficult to manage project by project,” Way says. Agents will also need to work across vendors because customers “do not run wall-to-wall SAP.”

At TechEd, SAP announced new agent gateway capabilities that let third-party agents invoke Joule Agents through a governed entry point and take part in the same business process under SAP’s access controls

Way recommends setting governance up centrally and right away. If every project builds its own security and cost controls, he warns, “you’ll never catch up when multiple agents are collaborating across your environment.” SAP plans to release the SAP Enterprise AI Gateway in the first quarter of 2027 as a single place to apply those controls.

Central controls still depend on the people reviewing what agents do. The human-in-the-loop checkpoints in Joule Work’s orchestration capability are only as good as the person standing at them. “Approval fatigue is real, and a human that blindly clicks the ‘Approve’ button is the same as not having any human-in-the-loop at all,” Way cautions.

As agents take on more authority, he argues, “human-in-the-loop needs to be truly human,” and the controls around them have to work together. SAP also announced a built-in sandbox in Joule Studio for validating agent behavior before it reaches production.

Bringing employees along matters too, Way adds. Companies should find the people already using AI to multiply their impact and help them bring colleagues with them.

As a technology company, Way notes, “you would think we only get excited about all the latest innovations in AI that we leverage in our products. Of course that’s true, but what’s more important is understanding the user and governing the AI.”

What This Means for SAPinsiders

Agent projects work best when they start from the goal. Joule Work and Joule Studio are built so that an employee or developer states what they want done and the software works out the steps. Teams that frame each use case around the outcome keep it tied to work people already want done.

Business context decides how useful an agent is. Way recommends building a modern data estate, with unified data products and clean master data, because agents reason about a business only as well as its data allows. The SAP Knowledge Graph and Ontology Manager, both due in the first quarter of 2027, are designed to draw on customer-managed data products, so the data work done now is what those tools will have to work with.

Employees already using AI well can lead adoption. Way advises finding the people using AI to multiply their impact and helping them bring colleagues along. Colleagues learn from the people they already work with, which gives adoption a practical starting point.

 

Events

07Dec
SAPinsider Technology Executive ForumPhoenix, Arizona, United States
View All