At Google Cloud Next 2026 in Las Vegas on April 22, SAP and Google Cloud announced an expansion of their partnership focused on deploying multi-agent AI across enterprise marketing and customer experience workflows. The integration connects SAP Joule — SAP’s own AI assistant embedded across its ERP and CX portfolio — with Google’s Gemini Enterprise, creating a shared environment where agents from both platforms can coordinate on complex business tasks.

This is not a feature update. It is a statement about where enterprise AI is going: agents from different vendors, working together, across the software stack a business actually uses.

What Was Announced

The technical foundation of the partnership rests on two open standards that have been gaining traction across the AI industry.

The first is the Agent-to-Agent (A2A) protocol, which allows agents built on Google Cloud to communicate directly with SAP Joule. Rather than requiring custom integrations or a third-party middleware layer, the protocol gives both sets of agents a shared language for passing tasks and context between platforms.

The second is the Model Context Protocol (MCP), which provides agents with secure, real-time access to data stored in SAP systems via the ABAP SDK for Google Cloud. When a Google Cloud agent needs current inventory data, an open sales order, or a customer record to complete a task, it can retrieve that information directly from SAP without batch exports or manual hand-offs.

The data layer underneath both is the SAP Business Data Cloud (BDC) Connect for Google and BigQuery, which enables bidirectional, zero-copy data access between the two platforms. This means neither SAP nor Google Cloud needs to maintain duplicate copies of enterprise data — agents query the source of record, in real time.

The first live use case targets marketing. Through integrations between SAP Engagement Cloud, SAP Customer Experience, and Gemini Enterprise, joint customers can deploy agents that access unified data across both ecosystems to execute marketing strategies from a high-level goal. A marketing team defines the outcome; the agents work out the execution steps across the underlying systems.

The marketing use case is expected to reach customers in H2 2026, with the multi-agent model designed to extend across the broader SAP CX portfolio over time.

A Real Example

The announcement was accompanied by a real deployment. Smyths Toys, a European toy retailer, built a multi-agent resolution system connecting a frontend built on Google Kubernetes Engine and BigQuery with a backend powered by SAP systems and Google Cloud AI capabilities. The result: over 23% of incoming customer queries now handled automatically, without a human in the loop.

That number matters not as a headline but as a calibration point. A complex customer query — one that requires checking stock levels, order status, and account history across different systems simultaneously — is exactly the kind of task where multi-agent orchestration earns its cost. Single agents working within a single platform cannot do this efficiently. Connected agents that share a common data layer and a shared protocol can.

Why This Matters

SAP is not a niche platform. It runs the ERP, finance, and CX workflows for the majority of large enterprises globally. Google Cloud is rapidly becoming the AI infrastructure layer for those same enterprises. When these two platforms announce that their agents can now work together through open standards, the implication is significant: the era of siloed enterprise AI agents is ending.

Until recently, the challenge with AI agents in large organizations was coordination. An agent embedded in your CRM could not see what was happening in your ERP. An agent in your cloud data warehouse could not take action in your customer platform. These silos created a situation where agents could make recommendations but not complete work across systems.

The A2A protocol, MCP, and integrations like this SAP-Google partnership are dismantling those silos systematically. This is not happening through a single vendor building everything — it is happening through open standards that allow different vendors’ agents to interoperate.

For businesses currently evaluating AI agent strategies, this shift matters for a practical reason: the value of AI agents compounds when they can coordinate. A single agent answering one type of question delivers incremental value. A network of agents that can pass tasks, share context, and complete multi-step workflows across your entire software stack delivers something closer to an autonomous operational layer.

What This Means for Business

If you are a business running SAP for your core operations and Google Cloud (or considering it) for AI workloads, the partnership announced today is worth tracking closely. The marketing integration arriving in H2 2026 is the first evidence of what this looks like in production, but it is not the ceiling.

The broader pattern here reflects a maturation in how enterprise AI is being built. Large vendors are increasingly committing to open protocols rather than proprietary lock-in, because interoperability is what enterprise buyers are demanding. That shift benefits businesses — it means AI agents you deploy today can connect to new platforms tomorrow without a full rebuild.

The companies best positioned to benefit are those that have already invested in clean, accessible data foundations. The Smyths Toys example illustrates this directly: the 23% automation rate is not primarily a function of which AI model is running. It is a function of having real-time access to reliable data across systems. If the data is fragmented, delayed, or inconsistent, even the best multi-agent architecture cannot help.

For organizations that want to move toward a connected AI agent workforce — where agents across your CRM, ERP, marketing stack, and communication channels work together rather than in parallel — this is the direction the industry is building toward. The infrastructure is arriving. The question for most businesses is whether their data layer is ready to support it.

For a deeper walkthrough of tools like this and how they fit together, the free Working With Claude field guide covers the ecosystem end to end. Get the guide.