SAP pitched its process knowledge as the context layer for agents at Connect 2026. Forrester says CIOs must first decide which work intelligence SAP should supply. They must also decide which context debt to repair before agents execute. CIOs must decide who owns what agents learn from corrections and overrides. They must decide who measures value and bears failure risk for agent actions. SAP’s architecture splits engagement, work organization, and execution across its tools. SAP Knowledge Graph maps fields, tables, APIs, and data products to provide context. Forrester warns that vendor-reported scope does not guarantee semantic accuracy. It advises keeping company-specific definitions under enterprise control. SAP allows other AI experiences to call its agents but offers connectivity, not portability. Forrester recommends classifying agent failures to expose context debt. It advises repairing defects that repeatedly block valuable work or create control risk. SAP plans a company-memory capability to capture reasoning from overrides and conversations.
What it unlocks
Using SAP process knowledge to ground and coordinate AI agents across enterprise applications.
- 7 million fields mapped
- 500,000 tables mapped
- 50,000 APIs mapped
- 400 data products mapped
What you need to act on it
- SAP Joule Work, AI assistants, or agents deployment
- SAP Business Data Cloud or SAP Knowledge Graph access
- Enterprise control over governance, identity, and data
Sources