Sunil Singh

Senior Manager — SAP Analytics Consulting, Deloitte Consulting GmbH

November 2026

21 min read

Analytics-lens analysis Architecture + Strategy For architects, data leaders, practitioners 7 official SAP sources · every claim mapped

SAP TechEd 2026: Lakehouse, Context, and Agents — A Sourced Analysis for SAP Analytics

SAP TechEd 2026 ran at Messe Berlin and online from October 27–29, under the keynote banner "Built for the Ones Who Build." SAP's CTO, Philipp Herzig, summed up the thesis as moving "from AI learning your business to running it," and the commercial headline matched: general availability of Joule Work, Joule Studio, and the Autonomous Enterprise assistants and agents is now underway in phases. If you own the data and analytics layer, though, the announcements that matter most sit underneath the agents: an open lakehouse for SAP Business Data Cloud built on the Dremio acquisition, new tools to curate business context for AI, master data turning into data products, and a growing family of tabular AI models. This is my analytics-lens read of the event and a follow-up to my Sapphire 2026 analysis. Every product fact is traced to an official SAP source listed at the end, and I keep a hard line between what is generally available today and what is planned for Q4 2026 or Q1 2027.

How to read this page

I separate three things on purpose, as I did for Sapphire. Facts are what SAP announced, each traceable to an official SAP source in the references. Interpretation is my own read, in the shaded boxes. Timing is where TechEd needs the most care: SAP's innovation news guide mixes capabilities that are generally available now, capabilities whose general availability is "underway in a phased approach," and capabilities planned for Q4 2026 or Q1 2027. I use SAP's own words for status, give every item the date SAP gave it, and say so where SAP gave none. TechEd also built openly on the two earlier 2026 events — SAP Sapphire in May and SAP Connect in October — so where something was first announced elsewhere, I say where.

The headline: general availability, in phases

SAP's TechEd 2026 innovation news guide, published October 27, 2026, opens with the status line that matters most for planning: general availability for the Autonomous Enterprise assistants and agents, Joule Work, and Joule Studio "is now underway in a phased approach," with general availability for other innovations starting in October and further capabilities to follow. Joule Work and Joule Studio were both announced at Sapphire in May, and SAP's Connect guide (October 6, 2026) added that the Autonomous Suite assistants and agents, Joule Studio, and Joule Work are delivered and activated as SAP-managed experiences through the SAP for Me portal. TechEd was where SAP showed what comes next on top of them.

Joule Work is the experience where a person describes a goal and AI orchestrates the agents, applications, and data needed to reach it, across SAP and non-SAP systems. At TechEd it gained an agentic thinking capability to reason, plan, and execute complex multistep work, with audit trails, role-based permissions, and human-in-the-loop checkpoints built in (phased GA); a secure gateway through which third-party agents can invoke Joule Agents under enforced access control (phased GA); and richer, customer-specific SAP Knowledge Graph context (GA planned for Q1 2027). SAP's press release frames the interoperability around open protocols, including Agent2Agent (A2A) and the Model Context Protocol (MCP).

Underneath sits the SAP Business AI Platform, which SAP describes as one place to build, contextualize, and govern AI across SAP and third-party environments; the new solution architect certification organizes it into three pillars — build; contextualize and reason; and govern. One passage in the press release deserves a data leader's attention: SAP restated that customer data remains the customer's, and that SAP does not share it with anyone to train or improve large language models.

My read: "Phased approach" is the phrase to plan around. It means parts of these products are generally available for some customers now; it does not mean everything demonstrated in Berlin is in your tenant today. My working rule is simple: before any TechEd capability goes into a plan, find its status sentence in SAP's guide, then confirm it for your own landscape through SAP for Me. The platform framing is also a useful reminder of where analytics effort pays off — "contextualize" is the pillar data and analytics people own, and almost every data announcement below lands there.

The analytics core: SAP Business Data Cloud

SAP describes Business Data Cloud (BDC) as the trusted data foundation for the SAP Business AI Platform: a business data fabric that connects data, semantics, processes, and policies while preserving the business context that gives data meaning. Irfan Khan, president and chief product officer of SAP Data & Analytics, set out the TechEd additions in an October 29 post. Four of them change how I would draw a target architecture; a fifth changes how you run the database underneath.

An open lakehouse, built on Dremio

SAP completed its acquisition of Dremio on July 6, 2026, and TechEd showed what it becomes inside BDC. According to SAP's guide, new lakehouse capabilities will provide native support for Apache Iceberg and a distributed SQL query engine: a native Iceberg catalog based on Apache Polaris, integrated with the BDC catalog, and an engine that can query a variety of lakehouse catalogs and relational databases. BDC already supports Delta Lake; Iceberg is the addition. Khan's post adds the working detail — federated queries across lakehouses and relational databases, including SQL Server, MongoDB, Oracle, and Postgres, without moving data, and Dremio's Reflections technology, which learns from how data is used and automatically precomputes and optimizes the results workloads need, adapting as those workloads change. Through Joule Work, business users will be able to question this data in natural language. SAP plans general availability of the lakehouse capabilities for Q1 2027.

Dating note: SAP's own sources differ in tense here. Khan's October 29 post says BDC "now supports" native Apache Iceberg and a Polaris-powered universal catalog; SAP's innovation news guide, published two days earlier, plans general availability of the lakehouse capabilities for Q1 2027. I plan against the guide's date. Until the capability shows up as generally available for your landscape, treat Iceberg in BDC as roadmap, not as a dependency.

Business context you can curate: BDC Ontology Manager and SAP Knowledge Graph

The second theme is context — the thing SAP keeps saying agents lack. SAP Knowledge Graph already provides business context about a customer's SAP systems; the new SAP Business Data Cloud Ontology Manager adds an AI-assisted way to curate domain-specific ontologies on top of it: focused, reusable business context for particular applications and agents. Developers select the context relevant to a custom agent and publish a domain view for a specific business problem — SAP's examples include understanding customers, investigating equipment failures, and recommending next best actions. Khan's example makes it concrete: a collections agent needs customers, invoices, payment terms, and credit risk, so a team publishes a collections view, and because views inherit definitions and governance from the Knowledge Graph, the same semantics can be reused for cash flow or order management without being redefined. Ontology Manager will work with both SAP data products and customer-managed data products, and SAP says reuse helps reduce time and token use. Its beta is planned for November 2026.

The Knowledge Graph itself is widening. It has traditionally drawn on data from specific SAP applications; new capabilities will add information in SAP-managed data products, customer-managed data products in BDC, and data in extensions of SAP Autonomous Suite offerings, so Joule can ground its answers in each customer's own landscape. SAP's press release adds that users will be able to query and explore this customized data in natural language. General availability for all features is planned for Q1 2027.

Master data as data products: SAP Reltio

SAP Reltio — multi-domain master data management based on technology from Reltio, which SAP acquired earlier this year — has been generally available in BDC since SAP Connect in October. Its AI-based entity resolution matches and reconciles records across SAP and third-party systems into unified views of customers, suppliers, products, and other entities, and those trusted profiles can be published as governed data products in BDC for Joule and Joule Agents to use. At TechEd, SAP added support for building unified data products; together with SAP Master Data Governance in BDC, SAP says this will deepen native master data management inside the fabric. General availability of the SAP Reltio data-product capabilities is planned for Q1 2027.

SAP Databricks gets more of the Databricks platform

SAP also expanded its Databricks partnership, bringing Lakeflow, Agent Bricks, Unity Gateway, and Lakebase to SAP Databricks inside BDC. Alongside it, a new SAP Application Data Foundation makes data from SAP applications available for consumption in the BDC engine of your choice, including SAP Databricks, and the resulting SAP data products can then be shared zero-copy through SAP BDC Connect. SAP also announced an integration between Joule and Databricks' Genie. Neither the press release nor Khan's post gave availability dates for these items, so I treat them as announced, with timing to be confirmed.

Joule for SAP HANA Cloud administration

Finally, new Joule agentic capabilities in SAP HANA Cloud will give database administrators a conversational interface to monitor system health, troubleshoot issues, optimize database and SQL performance, and manage lifecycle operations, with AI-assisted diagnosis and remediation. SAP's rationale is one every platform team will recognize: AI and agent workloads raise query volumes and make resource consumption harder to predict. General availability is planned for Q4 2026.

My read: The lakehouse is the most consequential announcement of the event for data architects, because it changes what BDC is. Until now the pitch was a governed SAP fabric that reaches other platforms through zero-copy connectors; with native Iceberg, a Polaris-based catalog, and a federated SQL engine, BDC also becomes a place where open tables are stored, cataloged, and queried in place. That is good news for estates that standardized on open table formats — and it raises questions to get answered before Q1 2027: which catalog is authoritative for which tables, how BDC governance applies to federated sources, and how automatic precomputation shows up in your capacity consumption. On context, Ontology Manager is the right design instinct: it is the semantic-layer pattern we already use — subject-area views on a shared, governed model — applied to agents. Treat the November beta as a place to learn, not something to build production dependencies on.

Joule Studio for data teams

Joule Studio is SAP's AI-first development environment for building and extending enterprise agents, and it is natively integrated with Joule Work, which SAP calls its primary engagement layer. Business users and developers get one place to create custom agents, extend existing Joule Agents, and orchestrate workflows through SAP's partnership with n8n, with no-code and pro-code tools and a choice of more than 70 AI models. The pieces that matter most for data and analytics teams:

BDC data products inside Joule Studio. Development teams will be able to create and access BDC data products directly in Joule Studio, and customers can choose to give agents runtime access to trusted business data within controls IT has already established. SAP says this automates what was previously a manual process. General availability is planned for Q1 2027.
Intent-based development. A developer enters intent, and Joule Studio helps identify the processes, data, and APIs involved, map system relationships, and account for authorizations, drawing on SAP LeanIX, SAP Signavio, SAP Knowledge Graph, and SAP Domain Models. In one step, the Joule Studio CLI can configure MCP connections, development skills, and authorizations. Phased GA.
Testing and evaluation. A built-in pre-deployment sandbox lets teams run test queries, inspect traces, and use auto-generated unit tests (phased GA). Offline evaluations will assess accuracy, A2A call correctness, and resistance to prompt injection against auto-generated scenarios, so changes to prompts, models, or data sources can be verified in isolation first; general availability is planned for late Q4 2026.
Runtime and oversight. Agents deploy on the Joule Studio runtime, where they become discoverable in the SAP AI Agent Hub and usable in Joule Work conversations; runtime-level isolation with configurable policies and guardrails comes from the Joule Studio runtime and NVIDIA's Open Agent Safety Platform (both phased GA). SAP also added observability and administrative controls covering agent behavior, model usage, and spend.
n8n workflows. Customers in the early adopter program can already design and run n8n workflows in Joule Studio; general availability is planned for Q1 2027.

On cost, SAP is running a time-bound promotion: for select customers, deploying and running custom agents on the SAP-managed Joule Studio runtime remains free through December 2026, and the design-time tooling for building and testing agents and n8n workflows remains free under fair-use limits. A new Joule Studio developer certification on the SAP Learning site is free through December 31, 2026, after which the regular cost applies.

My read: For analytics teams, the date to circle is Q1 2027, when BDC data products become available inside Joule Studio. That is the difference between agents grounded in governed, versioned data products and agents wired to whatever API a developer found first. Until then, spend the free window on discipline rather than demos: sandbox testing and offline evaluations are what turn an agent into something an auditor will accept. If you are one of the select customers on the free runtime, measure your runtime consumption before the promotion ends in December, so that 2027 budgets rest on your numbers rather than on estimates.

Tabular AI: the SAP-RPT family grows

SAP's argument for tabular AI is that large language models handle unstructured content well but struggle with the structured data that business decisions run on, and that tabular models can make precise, explainable predictions on that data without custom training or feature engineering. SAP launched SAP-RPT-1 in 2025, and TechEd added three models:

SAP-RPT-TS (generally available now). Time-series forecasting for time-indexed business metrics — demand forecasting in SAP Integrated Business Planning is SAP's example — without custom machine learning. Developers can use it through SAP HANA Cloud or SAP AI Core.
SAP-RPT-Decisions (GA planned for Q4 2026). Decision automation in applications and agentic workflows, such as approving a customer order in SAP S/4HANA, routing supply chain exceptions in SAP IBP, or selecting the approval path for a purchase requisition in SAP Ariba solutions. Available through SAP AI Core.
SAP-RPT-2 (GA planned for Q1 2027). Extends SAP-RPT-1.6 beyond regression and classification to anomaly and outlier detection, structured-data embeddings for AI-assisted analysis, and native multi-table processing — for example, surfacing unusual transactions in SAP S/4HANA Finance. Available through SAP AI Core; the press release styles it SAP-RPT-2.0.

SAP and HPE also announced the deployment of SAP-RPT on the HPE Spaceborne Computer-2 aboard the International Space Station, where SAP says the model is planned to run onboard for anomaly triage, failure prediction, reconstruction of missing sensor values, and mission-health reporting without sending data back to Earth. Herzig tied the work to TabPFN from Prior Labs, the tabular foundation model company whose acquisition SAP completed in July 2026.

My read: SAP-RPT-TS is the model analytics teams can use now, and it has an obvious home in planning and forecasting. SAP-RPT-2 is the more interesting one for analytics — anomaly detection and multi-table processing go straight at finance analytics and data-quality monitoring — but it is a Q1 2027 marker, not a present-tense capability. Whatever you pilot, benchmark it against your current forecast or rule baseline on your own history before you switch: "no custom training" removes the build effort, not the need to validate. The space deployment is a credible edge-computing story; your own backtest is still the only benchmark that counts.

Governance: a gateway, an assistant, and a knowledge repository

The governance announcements matter to analytics leaders because they decide who can reach your data through an agent, and how that use is tracked and paid for.

SAP Enterprise AI Gateway (GA planned for Q1 2027). Part of the SAP Business AI Platform and built on SAP Integration Suite, the gateway will centrally manage AI traffic — MCP, API, A2A, and LLM calls — and consolidate authentication, policy enforcement, routing, monitoring, usage tracking, and auditing in one managed layer. Developers get a standardized MCP gateway and developer hub to discover and subscribe to MCP servers from any provider and to reuse existing APIs and integration flows as agent-ready capabilities; platform teams get identity-based authentication, policy-driven authorization, and consumption monitoring.
AI Governance Assistant (GA planned for Q4 2026). A Joule assistant, part of the SAP AI Agent Hub, that supports EU AI Act compliance from the enterprise architecture inventory you already maintain: it extracts content from uploaded system documentation, identifies gaps, and proposes a risk classification — prohibited, high, limited, or minimal risk — with citations to the relevant requirements, then triggers reassessment when a fact sheet changes materially. SAP frames every output as a draft for legal and compliance review.
Enterprise knowledge repository (GA planned for Q1 2027). Connects architecture decisions, standards, landscape data, and documentation into governed knowledge that agents can ingest and reuse.

My read: The gateway is the control point most agent architectures are missing today: one place where identity, policy, and consumption meet before an agent reaches a model or a data API. If your agents will touch BDC data products, plan for that traffic to pass through it once it is generally available, and size your 2027 governance work around it. The governance assistant is useful precisely because SAP is clear about its scope — it drafts classifications for legal review; it does not make the legal call. My EU AI Act orientation for SAP analytics covers what that means for analytics systems.

Beyond the analytics lens (brief context)

The rest of the event in short form — important, but not the focus of this site.

SAP Integration Suite. Generally available now: expanded distributed tracing for advanced event mesh with SAP S/4HANA support, unified real-time integration monitoring in SAP Cloud ALM, an additional, fully managed edge integration cell tenant for RISE with SAP customers' private landscapes, AI-assisted resolution of message-processing errors, AI detection of irregular request payload sizes, deeper GitHub integration, and RosettaNet support. Automatic construction of message-mapping expressions from sample payloads is planned for Q1 2027.
SAP Document AI. A new SAP-DOC-1 model that extracts information without up-front configuration, a document-splitting node, custom prompting, and deeper integration with Joule Work and Joule Studio; general availability for all enhancements is planned for Q1 2027.
SAP Sovereign Cloud on AWS. SAP AI Core (with curated LLMs through the generative AI hub), Joule, and SAP Datasphere are being added for regulated industries, with general availability planned for Q4 2026 in the United Kingdom, Australia, and Canada — relevant if you run Datasphere under sovereignty requirements.
Skills. A new SAP Business AI Platform solution architect certification and learning journey are generally available; candidates may use Joule and SAP documentation during the exam. My guide to SAP analytics certifications shows where it fits.
Partners. A collaboration with Intel on physical AI that links enterprise workflows to autonomous robots, and new SAP solution extensions with Docusign, Parloa, and Redwood.

GA vs. planned: the timeline

These are the two tables I would pin above any 2027 roadmap: one for data, context, and AI models, one for agents, development, and governance. Status and timing use SAP's own words from the TechEd guide and press release; "phased GA" means SAP says general availability is underway in a phased approach.

An architect's playbook for the next two quarters

This section is my judgment, not SAP's announcement. Most of what SAP showed in Berlin lands between now and the end of Q1 2027, so this is what I would do in that window.

Decide your open-table-format posture before the lakehouse ships. BDC supports Delta Lake today and plans general availability of its Iceberg-based lakehouse capabilities for Q1 2027. If your non-SAP estate has standardized on Iceberg, write down now which catalog will be authoritative for which tables and which engine governs access, so the Q1 2027 release becomes an implementation decision rather than an architecture debate.

Curate context on paper before you have the tool. Ontology Manager enters beta in November 2026, but the hard work is not the tool. Pick the two or three domains where agents will act first — collections, cash flow, and order management are SAP's own examples — and document the entities, relationships, and rules each one needs. That work carries over to whatever tooling you end up using, and it is the same work a good semantic layer needs anyway.

Use the free window to build evaluation discipline. If you are among the select customers on the free Joule Studio runtime, use December to measure runtime consumption and to build the evaluation habit — sandbox tests now, offline evaluations when they arrive in late Q4 2026. Design agents that need governed data around BDC data products in Joule Studio (Q1 2027), not around ad hoc API access.

Pilot SAP-RPT-TS where you already forecast. It is generally available, and planning is where its value is easiest to test: backtest it against your current baseline on your own history. Keep SAP-RPT-2 on the roadmap as a Q1 2027 marker for anomaly detection in finance and data-quality monitoring.

Put one governance choke point in the 2027 plan. Inventory your AI systems and agents now, because the AI Governance Assistant (Q4 2026) works from the architecture inventory, and plan to route agent traffic through the SAP Enterprise AI Gateway once it is generally available (Q1 2027). Register agents in the SAP AI Agent Hub from day one.

Key takeaways

General availability of Joule Work, Joule Studio, and the Autonomous Enterprise assistants and agents is underway in phases; most of the new capabilities SAP showed at TechEd are planned for Q4 2026 or Q1 2027.
For analytics, the center of gravity is SAP Business Data Cloud: a Dremio-based lakehouse (native Iceberg, a Polaris-based catalog, a federated SQL engine; GA planned for Q1 2027), BDC Ontology Manager (beta planned for November 2026), customer-specific SAP Knowledge Graph context (Q1 2027), and SAP Reltio data-product capabilities (Q1 2027) on top of SAP Reltio's general availability in BDC.
SAP Databricks gains Lakeflow, Agent Bricks, Unity Gateway, and Lakebase, alongside a new SAP Application Data Foundation and a Joule–Genie integration; SAP gave no dates for these.
Joule Studio brings BDC data products into agent development in Q1 2027; the SAP-managed runtime is free through December 2026 for select customers, and design-time tooling is free under fair-use limits.
Tabular AI: SAP-RPT-TS is generally available; SAP-RPT-Decisions is planned for Q4 2026 and SAP-RPT-2 for Q1 2027.
Governance gets a control plane: the SAP Enterprise AI Gateway (Q1 2027) and an EU AI Act–focused AI Governance Assistant in the SAP AI Agent Hub (Q4 2026). Inventory now; route agent traffic through the gateway when it ships.

Freshness and scope note

This is a time-sensitive page about a launch event (SAP TechEd, Berlin and virtual, October 27–29, 2026). Most items are planned rather than generally available, and SAP's scope and dates can move; statuses were compiled in November 2026 from SAP's own materials. Where SAP sources differ in tense — the lakehouse, which SAP's October 29 post describes in the present tense while SAP's guide plans general availability for Q1 2027 — I use the dated, more conservative statement. Connector-by-connector dating for SAP BDC Connect is in my Sapphire 2026 analysis. This page is independent editorial analysis from an analytics and data-architecture point of view, not official SAP product documentation; verify any release-dependent detail against current SAP materials before high-stakes decisions. Sections marked "My read" are interpretation, not SAP claims.

Sources

Attribution: Product facts on this page are drawn from the seven official SAP communications listed above (July–October 2026), accessed and verified in November 2026. Status and timing follow SAP's TechEd 2026 innovation news guide unless another source is named. The analysis in the "My read" boxes and the playbook is the author's own.