Analytics Legends logoAI & Analytics Legends
Menu
Sign inSign up
Guide

AI-ready data in SAP: what readiness actually requires

As of 2026-11-06

Edited by · How this page is made

An agent is only as right as the data it is grounded on, and SAP's grounding path runs through Business Data Cloud data products, Datasphere semantics and the SAP Knowledge Graph. Readiness therefore means three things you can test: governed data products for the objects the agent needs, a semantic model the graph can read, and master data clean enough that two agents do not disagree. The Ontology Manager and the data product creation agent are announced, not available; readiness cannot wait for them.

Looking for a ? Cédric Mary, SAP AI & Analytics Architect & Team Lead, takes SAP Business AI, Joule and Business Data Cloud missions.

Grounding path from sources to agents in SAP: S/4HANA, BW, non-SAP sources; Business Data Cloud data products; Datasphere semantics and knowledge graph; Ontology Manager and data product creation agent marked announced.
The data an agent stands on, and which parts are still announced. Sources: SAP TechEd in 2026: SAP Tells Developers You Can Build the Autonomous Enterprise (SAP News, 28 October 2026)
Readiness decision tree: are data products installed and owned, is there a semantic model, is master data reviewed, do custom fields reach the graph.
Four readiness tests before the first agent. Sources: SAP TechEd in 2026: SAP Tells Developers You Can Build the Autonomous Enterprise (SAP News, 28 October 2026)

Why it matters now

Data readiness sat behind many TechEd 2026 questions even when the word was not used: 131 of the 956 knowledge units concern the knowledge graph, ontologies or semantics, 88 concern data products, and 76 of the 531 audience questions asked about them. People asked where a data product is physically created, whether it is materialised and how it refreshes, whether a derived data product costs extra in Business Data Cloud, how the knowledge graph ingests custom fields from S/4HANA Private Edition, and whether open standards such as data contracts or quality checks are planned.

SAP's TechEd innovation guide of 27 October 2026 gives dates: the data product creation agent in early adopter care in Q4 2026 and general availability in the first half of 2027 (another slide of the same deck gives a different quarter). The Ontology Manager is listed as beta planned for November 2026, a date we could not confirm elsewhere. Knowledge graph extensions to customer-managed data: GA planned for Q1 2027. SAP Reltio, for multidomain master data, was presented as part of Business Data Cloud. The master data discipline of the last twenty years has become the AI prerequisite.

The market is still paying for the plumbing: Datasphere is the most requested skill in live SAP analytics adverts (33.1 % of roughly 1,900 adverts with skills, 16 October 2026) and the knowledge graph appears in 0.5 %. The readiness work is where consultants are, and the graph is where they will be.

Decision guide by situation

You run S/4HANA Cloud with Business Data Cloud — Install the SAP-managed data products for the objects your first agent needs; check what the SAP-managed knowledge graph already covers for them; start the master data review on the same objects.

You run S/4HANA Private Edition with custom fields — Custom extensions are the TechEd question with the least complete answer. Inventory the custom fields the agent must understand and test whether they reach the graph; if not, the agent must be scoped to standard fields.

You have Datasphere but no Business Data Cloud — Datasphere semantics and the catalogue are the grounding layer you have; the graph engine on HANA Cloud is reachable. Build the semantic model as if the graph were coming, and keep the catalogue clean.

You have BW and no cloud data layer — Readiness starts with which BW objects become data products; the BW Data Product Generator is the tool for SAP-managed scenarios. Do not ground an agent on BEx queries.

You have non-SAP data in Databricks, Fabric or Snowflake — BDC Connect brings it in as data products, with Databricks the most mature path. A TechEd question asked whether a Fabric data product can feed an SAP agent; SAP's answer was that federated access is a roadmap item. BDC Connect for Microsoft Fabric itself is in restricted public preview with general availability planned for Q3 2026 by SAP, not yet confirmed shipped.

Options compared

Ground on data products plus the SAP-managed knowledge graph — Effort: install and curate · Trust: highest for SAP standard objects · Limit: custom and non-SAP semantics until the Q1 2027 extensions.

Ground on Datasphere models and catalogue — Effort: modelling you may already have · Trust: high within the model · Limit: no SAP-managed graph over custom objects yet.

Ground on a vector index of documents — Effort: low to start · Trust: variable, answers drift · Limit: no business semantics, no authorization model unless you build it.

Ground on application APIs only — Effort: low · Trust: exact for the record returned · Limit: no cross-object reasoning, heavier per-call cost.

What shipped and what is announced

Available: 300+ SAP Data Products (generally available); BW Data Product Generator (BDC path) (generally available); BDC Connect — Databricks Enterprise (generally available); SAP Reltio in Business Data Cloud (generally available).

Announced, not available: SAP Business Data Cloud Ontology Manager (beta planned for November 2026); SAP Knowledge Graph on customer data (GA planned for Q1 2027); SAP Reltio data product capabilities (GA planned for Q1 2027); Dremio-based lakehouse in SAP Business Data Cloud (GA planned for Q1 2027). A date here is a plan, not a commitment.

Slipped or unconfirmed: 500+ SAP Data Products (unconfirmed (target Q4 2026)).

Outside the SAP status registry: Data product creation agent: early adopter care Q4 2026, general availability first half 2027 (SAP TechEd 2026 Innovation News Guide); Iceberg and Polaris in the lakehouse: one SAP page says Q1 2027, a feature page of 29 October 2026 says supported now (SAP pages read on 5 Nov 2026).

To verify in your tenant: what the lakehouse exposes today, and whether derived data products are billed beyond the Business Data Cloud capacity you already hold. Both were asked at TechEd and neither has a published general answer.

Pitfalls

Waiting for the Ontology Manager. Semantics live in Datasphere models today; the manager will consume them, not create them.

Grounding a pilot on a document index because it is fast, then discovering that the answer ignores authorizations.

Treating installed data products as curated. Installation is a step; ownership, refresh and quality are the readiness.

Skipping master data because it is unglamorous. Two agents reading two customer records give two answers, and the business will blame the AI.

Assuming custom fields are in the graph. Test it on the Private Edition tenant before scoping.

What to do Monday morning

Pick the first agent and list the five business objects it needs. For each: data product installed or not, owner named or not, master data reviewed or not.

Run the SAP-managed knowledge graph coverage check on those objects; write down the gaps.

Open the custom-field inventory with the application team if you are on Private Edition.

Put the announced dates in the plan as dependencies to verify, not as milestones.

Read our module on SAP Domain Models and the knowledge graph for agents, and the concept on data products as AI context, both linked below.

What we cannot assert

We could not confirm the date SAP gives for the Ontology Manager outside its TechEd guide. Whether a tenant has Iceberg and Polaris access today is contradicted between two SAP pages. The metering of derived data products is not published.

Frequently asked

Is Datasphere inside Business Data Cloud or under it?

Datasphere is a component of Business Data Cloud, as SAP stated at TechEd 2026; it is not a separate layer underneath.

Do we need the knowledge graph to ground an agent?

No; data products and Datasphere semantics ground agents today. The graph adds cross-object reasoning for SAP data now and for customer data according to SAP's plan (GA planned for Q1 2027).

Does a derived data product cost extra?

Asked at TechEd, not answered publicly. Business Data Cloud is sold in capacity units; check your contract and your tenant metering before building many derived products.

Can the knowledge graph read custom fields from S/4HANA Private Edition?

Partially, and this was the least settled TechEd question. Test on your tenant; scope the first agent to standard fields if the test fails.

Engage the editor

This guide is written by . 27 years of SAP — Datasphere, Business Data Cloud, SAC Planning, Joule. Freelance or permanent · Hybrid / remote · Anywhere in EMEA.

Profile and contact

What this page is built on

External sources

Read next

The other eleven challenges

Get new SAP Business AI & Joule missions every week

No account needed. We first ask you to confirm by e-mail; nothing is sent without that confirmation, and every alert has a one-click unsubscribe link. Privacy

Related guides

These guides cite this page as a source.

Open missions

115 live roles and contracts in SAP AI & Joule right now — see the SAP AI & Joule missions · freelance SAP contracts.