AI-Ready Data — Quality, Semantics and Lineage

What you will learn

• Explain why 'AI-ready' resolves into four checkable properties — quality, semantics, lineage, authorisation — and reject 'we ran it through AI' as a quality control
• Name real SAP data-quality mechanisms (Data Quality Management microservices for location data; Reltio-based master-data harmonisation in Business Data Cloud) and the failure mode each addresses
• Check on the Catalog's Semantic Enrichment tab that every measure an agent will quote carries a description, a glossary term and, for a KPI, a formula
• Use SAP Datasphere's Impact and Lineage Analysis, distinguishing Data Analysis from Dependency Analysis, to trace a column back to its source systems
• State why an agent's data access must inherit a human user's space-level authorisation rather than a broader service identity

Module overview

Who this is for. You know how to model a Datasphere space or a BW/4HANA InfoProvider; now a steering committee wants to know whether that data is "AI-ready" before Joule or an agent is allowed near it. This module gives that word a definition you can defend: not a feature flag, but three checkable properties — quality, semantics and lineage — plus the authorisation boundary that decides who, human or agent, may read what. It builds on AI & LLM fundamentals and feeds directly into the semantic layer as LLM context, natural-language query and vector search on analytic data.

Prerequisites

• AI & LLM Fundamentals for SAP Consultants or equivalent working knowledge of grounding and RAG
• Working knowledge of at least one SAP data product (Datasphere space, BW/4HANA InfoProvider or S/4HANA embedded analytics)
• Familiarity with the Datasphere Catalog and space-based authorisation, or willingness to explore a trial tenant

Outcomes

• Score a real data product against a four-point AI-readiness checklist (quality, semantics, lineage, authorisation) and justify each score.
• Trace a column's provenance through Datasphere's Impact and Lineage Analysis and explain what it would and would not show under Dependency Analysis.
• Write a business glossary term (name, definition, formula, synonym) and attach it to the right analytic-model column so that it appears under Related Objects.
• Explain to a client, without hand-waving, why an ungoverned data product produces confidently wrong agent answers.