Content
SAP spent Connect 2026 showing the Autonomous Enterprise at work: Joule Work, a wave of agents, and the SAP Business AI Platform. The harder question for your business is whether any of it can act on your processes, and that answer lives in the integration between SAP and the rest of your systems.
SAP Connect 2026 ran October 5-7 in Las Vegas, building on the Autonomous Enterprise vision from Sapphire in May. The difference this time was tone. SAP talked about rollout: Joule Work reaching customers, assistants spreading across finance, supply chain, HR, procurement and customer experience, and more proof points than a roadmap slide.
This recap covers the main keynote and the published announcements. We previewed the event with the questions enterprise buyers should ask. Here is what shipped, what is still roadmap, and the integration work that decides whether Business AI answers questions or does the job.
The headline from SAP Connect 2026
SAP said the Autonomous Enterprise has arrived. Read it more carefully and Connect was a rollout checkpoint for the stack introduced at Sapphire, with more of it moving from slide to customer.
The architecture has three parts. Joule Work is how people engage, an intent-led workspace that replaces clicking through one application after another. SAP Autonomous Suite holds the assistants and agents across five business domains. The SAP Business AI Platform is where you build, add context and govern.
Joule Work got the most stage time. It runs on the SAP Knowledge Graph, which maps more than 7 million data fields, and SAP says it is live for 110,000 of its own staff with 20% productivity gains in finance, HR and procurement. Those are SAP’s own internal numbers, so weigh them as a vendor claim, not an audited result.
The real change is direction. These agents are meant to do the work, not only answer, and that holds only on connected, governed systems underneath.
The announcements that matter
One question sorts the noise: what can you use now, and what is still a promise on a roadmap?
| Announcement | Status | What it means for you |
|---|---|---|
| Joule Work | Customer rollout starting; GA timing varies | The visible layer. Its value depends on what is safely connected beneath it. |
| SAP Autonomous Suite (5 domains) | Many assistants land Q4 2026 / Q1 2027 | A portfolio direction, not a uniformly available autonomous stack. |
| SAP Business AI Platform | Active; SAP BTP now sits inside it | The build, context and governance foundation beneath the agents. |
| SAP Reltio (Business Data Cloud) | GA | Trusted master data, so agents act on coherent customer, supplier and product records. |
| SAP Pay | GA in the US and UK | AI moves from insight to a controlled payment executed inside Cloud ERP. |
| API-Centric Integration + MCP Gateway (SAP Integration Suite) | Delivered in Q3; more on the roadmap | Turns APIs into governed tools that agents can discover and call. |
| A2A + MCP interoperability | Adopted; some patterns still transitional | Connects agents to other agents and to governed enterprise tools. |
| Consumption pricing | 0.02 AI Units per agent action | Cost scales with how much the agents actually do. |
Two things stand out. The generally available pieces, Reltio, SAP Pay and the Q3 SAP Integration Suite work, are infrastructure rather than the headline agents. Many marquee assistants carry Q4 2026 or Q1 2027 dates, so “the Autonomous Enterprise has arrived” fits the architecture better than the full catalog. For any internal deck, put the real status next to each capability and let no keynote line stand in for an availability date.
Enterprise impact: SAP BTP integration and API integration
An agent is only as useful as the systems it can reach. That makes API integration the real story of Connect, and it explains where SAP BTP integration went.
BTP did not vanish, it moved up a level
SAP BTP is still here. Its capabilities now sit inside the SAP Business AI Platform, grouped around three jobs: build, contextualize and reason, and govern. For architects that continuity matters, because existing BTP integration, extension and security skills still apply. “BTP was renamed” is the wrong summary. It became one part of a bigger platform.
Integration Suite as the control plane
SAP Integration Suite is turning from classic middleware into a control plane for both application and agent integration. API-Centric Integration pulls routing, transformation, security, rate limiting and lifecycle into a single API artifact, so API management and integration live in one place. MCP Gateway then publishes selected APIs, flows and data sources as governed tools an agent can find and call, with customer-managed OAuth identity, quotas and tracing.
The shift is real. Traditional enterprise application integration had a developer wire an endpoint into a fixed interaction. Now an agent decides at runtime whether to call a tool, which still needs bounded permissions, approvals and an audit trail.
SAP splits the work across two open protocols:
| Protocol | SAP’s job for it | The question you still own |
|---|---|---|
| MCP | Agent-to-tool: expose APIs, integrations and data as governed tools via MCP Gateway | Which operations, at what granularity, with which identity, scopes and approvals? |
| A2A | Agent-to-agent: let Joule and external agents discover, delegate and exchange results | How are agent identity, delegation, trust boundaries and accountability handled? |
Open protocols reduce proprietary glue, but they leave the design work with you: tool definitions, identity propagation, versioning and monitoring. The payoff is real money when it is done well. SAP cites a 368% ROI and an eight-month payback for Integration Suite customers, measured by IDC, plus a sixth year as a Leader in Gartner’s iPaaS Magic Quadrant.
Gaps, trade-offs and open questions
Judge the announcements by whether you can deploy them, not by the demo. Five tensions sort progress from noise.
- Open standards, controlled access. SAP backs A2A and MCP, yet its API Policy limits autonomous AI from running API-call sequences outside SAP-endorsed pathways, and DSAG has asked for clarity. This is not a ban on third-party AI; it routes that traffic through governed pathways. The question for buyers: is Integration Suite a neutral control plane or a required tollgate?
- Maturity and evidence. The marquee numbers, 20% productivity and up to 50% less trade-classification effort, come from SAP. Ask for the baseline, the sample size and the review load before they go in a business case.
- Agent identity. You need an agent identity model now: named-user delegation versus service principal, least privilege, segregation of duties, revocation. SAP’s Access Governance and Security Assistant is a Q1 2027 item, so the design sits with you until then.
- Runtime economics. At 0.02 AI Units per agent action, a multi-step task runs up a variable bill. Budget by completed outcome rather than by seat.
- The migration clock. SAP marked PI/PO migration as delivered in its Q3 roadmap. With SAP Business Suite 7 mainstream maintenance ending in 2027, cleaning up legacy integration is now part of Business AI readiness.
Is your integration layer ready for SAP Business AI?
We get your SAP and non-SAP estate agent-ready – governed APIs, controlled integrations, and clear ownership – so autonomous processes run in production, not just in a demo.
Foundation first. Agents second.
Foundation first. Agents second.
Practical next steps
Start with your processes and interfaces, not an agent catalog. The order of work decides whether agents reach production or stall in a pilot.
A bounded path we use with SAP clients:
- Build an announcement register. For each capability that matters, record the product, region, edition, prerequisite, real status and an owner.
- Inventory your integrations. Integration Suite, PI/PO, point-to-point APIs, custom OData, events and any non-SAP iPaaS. You cannot govern what you have not mapped.
- Classify your APIs. Confirm which endpoints are published and whether an agentic use needs an SAP-endorsed pathway.
- Pick one cross-system process with real friction. High volume, clear rules, reversible actions, a visible exception backlog.
- Expose business-level tools, not raw endpoints. Separate read, propose and execute, and set an autonomy level for each step.
- Pilot through governed integration and measure cost per outcome. Test the failure paths, unavailable systems, stale data, rejected approvals, before you scale.
The pattern repeats on every platform: get the data and integration foundation ready first, then put agents on top of it.
Summary: five takeaways from SAP Connect 2026
- The headline was the Autonomous Enterprise; the working story is integration. Joule Work only feels seamless when the APIs, events, identity and data beneath it are connected.
- BTP still matters, it just moved into the Business AI Platform. Existing BTP skills and investments carry forward.
- MCP and A2A make agents interoperable in theory, not production-ready by default. Tool design, permissions and lifecycle stay with you.
- Availability is uneven. Reltio and SAP Pay are GA, while many headline assistants land in Q4 2026 or Q1 2027.
- Measure the next SAP AI case per completed process, counting AI Units, integration traffic, retries and human review.
Connect made the Autonomous Enterprise more concrete, and the integration dependency clearer with it. Business AI does not remove integration work. It raises the stakes, from connecting applications to governing how agents find, decide and act across them.
Related services: AI Data Integration Solutions · Data Integration Consulting · SAP + Dell Boomi Integration · EAI Services
Sources
- SAP Puts the Autonomous Enterprise to Work | SAP News
- SAP Connect 2026 Innovation News Guide
- A2A and MCP for Interoperability | SAP Reference Architecture
- SAP Integration Suite | iPaaS
- SAP Business AI pricing (AI Units)
- SAP API Policy (PDF)
- The Register: SAP user group slams ‘uncertainty’ in API policy
