SAP’s AI Stack Puts a Knowledge Graph Under Its Agents

SAP set out its AI strategy on one slide of its third-quarter 2026 earnings presentation, published on 21 October 2026. The slide is titled “Powering the Autonomous Enterprise”. It stacks AI agents on top of SAP’s apps and data, and puts the “SAP Knowledge Graph” between that data and what SAP calls the “SAP Business AI Platform”.

Chief executive Christian Klein tied the quarter to the same idea in SAP’s statement: “Our Autonomous Enterprise strategy is not just resonating, it is being adopted.” Our results piece covers the numbers. This piece maps the stack on the slide, layer by layer.

Business Pill · CONTEXT ENGINEERING

A short Business Engineer explainer on context engineering: giving an AI model the right business context before it acts. It teaches the general idea only and says nothing about SAP or any company in this piece.

Sap Q3 2026 explained · video

The quarter in about a minute and a half, from our published pieces (21 Oct 2026).

The Stack, Layer by Layer

The table follows the slide from top to bottom. The labels are SAP’s own, as printed on the slide.

A bracket on the slide labels the agents, apps and data layers together as the “SAP Autonomous Suite”. SCM is supply chain management, HCM is human capital management, meaning HR, and CX is customer experience.

What the Knowledge Graph Does

In plain terms, a knowledge graph is a map of how a company’s records connect: which customer placed which order, which invoice paid for it, which supplier shipped it. On SAP’s slide it sits under the agents and over the AI platform. As we read it, that placement is where an agent would draw business context from.

SAP’s statement says customers are investing “from the cloud and data foundation to our Business AI”, and that “Joule Work makes the business case for their transformation more attractive than ever”.

SAP’s statement also lists recent deals. On 6 July SAP completed the acquisition of Dremio, “an open, high-performance data lakehouse platform”. On 17 July it completed the acquisition of Prior Labs, “the pioneer of Tabular Foundation Models (TFMs)”. The statement also says SAP agreed to acquire TechWolf, “a provider of a leading AI work intelligence platform”.

How It Compares With Microsoft

Our Microsoft stack piece set out how Satya Nadella’s annual letter, also published on 21 October, describes Microsoft’s stack as “model-forward and agent-first”. The table sets the two descriptions side by side. It is our framing, from the two companies’ own documents as our pieces report them.

How We Read It

As we read it, SAP’s slide names no AI model and no chip. The layers it does name are its own: business data, the knowledge graph that links it, and agents built for finance, spending, supply chain, HR and customers.

As we read it, that puts the weight of SAP’s pitch on knowing how a customer’s business runs. Our harness piece reported Nadella’s view that models become interchangeable; as we read it, SAP’s slide leaves the model layer unnamed, although SAP bought Prior Labs, a model maker, in July.

The Structural Read

As we read it, the two companies’ stacks differ at the bottom: Microsoft’s letter starts with datacenters and chips, while SAP’s slide starts with its Business AI Platform and business transformation management.

As we read it, the deals SAP’s statement lists, Dremio and Prior Labs in July and the agreement to buy TechWolf, add data and AI pieces to that stack.

As we read it, the slide names no model provider, although SAP’s statement says it bought Prior Labs, a maker of tabular foundation models, in July.

Three Implications

FOR SAP CUSTOMERS As we read it, the slide asks customers to keep their business data inside SAP’s stack, which the slide draws under the agents.

FOR MODEL MAKERS As we read it, the slide gives model makers no named place in SAP’s stack.

FOR INVESTORS The slide gives no revenue figure for the agents or the AI platform. As we read it, that is the disclosure to watch.

The Business Engineer Lens

This story maps onto the Business Engineer framework Enterprise AI: From Software to Substrate.

The framework’s core line: “Enterprise vendors don’t sell software anymore—they sell irreversibility. The product is the interconnection.”

As we read it, SAP’s slide draws that idea as a stack, with a knowledge graph linking a customer’s business data under its agents. The framework does not discuss this slide.

What Is Not Established

The slide gives no revenue figure for the agents, the knowledge graph or the AI platform. SAP’s statement gives no count of customers running the autonomous agents in production. Our layer descriptions come from a single slide and SAP’s statement; SAP’s fuller product documentation may describe the layers differently.

Connecting the Coverage

Our results piece covers SAP’s first quarter above €10 billion of revenue and its €24.1 billion current cloud backlog. Our Microsoft stack piece and our harness piece cover Nadella’s letter, and our cost-to-outcome piece covers his idea that buyers will weigh token cost against outcomes.

What’s new here: SAP’s own picture of its AI stack, with a knowledge graph under the agents and no model layer on the slide.

The Bottom Line

SAP’s third-quarter presentation draws its AI stack as agents on top of its apps and data, with the SAP Knowledge Graph connecting that data to its Business AI Platform. As we read it, the slide puts the customer’s business data, not a named model, at the centre of SAP’s AI pitch.