Sapphire 2026 Changed Your S/4 Roadmap. Here's What Actually Matters.

Welcome to Between the Hype. A biweekly newsletter on where enterprise systems and AI actually intersect. Not the keynote version. The version you need when you're back at your desk on Monday.

I spent the last few days going through everything that came out of SAP Sapphire in Orlando. Dozens of announcements. Partner press releases. Analyst reactions. Demo recordings. It's a lot.

So here's what I'd tell you if we had 15 minutes over coffee: four things changed this week that actually affect your roadmap. Everything else is noise — or at least, noise you can catch up on later.

What SAP announced (the version that matters)

1. The Autonomous Enterprise

Christian Klein's headline: SAP is no longer a software company. It's a “business AI company.” Constellation Research analyst Holger Mueller called it the first genuine ERP vision on this side of the millennium. The product concept is called the Autonomous Enterprise — a unified platform where AI agents handle end-to-end business processes while humans supervise, approve, and handle exceptions.

Practically, this means SAP is shipping more than 50 Joule Assistants and over 200 specialised agents across finance, supply chain, procurement, HR, and customer experience. These sit on top of a new SAP Business AI Platform — which consolidates BTP, Business Data Cloud, and Business AI into a single platform with three layers: a context layer (data and knowledge), a build layer (Joule Studio 2.0), and a governance layer (SAP AI Agent Hub). Different components are shipping on different timelines — Joule Studio 2.0 starts in June, the AI Agent Hub goes GA in Q3 — but the architecture is defined and the rollout has begun.

Is it production-ready everywhere? No. But the architecture is defined, the development tools are shipping, and the partnerships are signed. This is a timeline now, not a roadmap slide.

2. Joule Studio 2.0 — the agent builder

This is the one with the most immediate practical impact. Joule Studio 2.0 went from a basic agent builder to an intent-based development environment. You describe a business outcome, and it generates the product requirements, technical specs, workflow logic, and evaluation criteria. It orchestrates multiple agents. It's model-agnostic — you can use Claude, or other models, depending on the task. And it works with both SAP and third-party environments through MCP and A2A protocols.

The part that matters most: SAP is offering all customers and partners twelve months of free design-time access to Joule Studio, including AI-assisted development capabilities under fair-use limits. First customers start receiving access in June. They've also invested €100 million in their partner ecosystem specifically for agent development.

If you have BTP and a use case in mind, the barrier to building your first agent just dropped to near zero. Test it before you build anything custom.

3. The AI partnership web

This is where it gets complicated — and where most of the coverage I've read gets it wrong.

SAP didn't pick one AI partner. They built a web. Anthropic's Claude is positioned as the primary reasoning and agentic capability across Joule agents — powering decision-making in finance, HR, procurement, and supply chain. But Joule Studio 2.0 is explicitly model-agnostic. SAP's AI Hub also supports OpenAI GPT models, Google Gemini, and Mistral. Microsoft and Google Cloud both have deep integrations for agent-to-agent interoperability. NVIDIA's OpenShell provides the secure runtime for agent execution within the SAP Business AI Platform.

And the Joule-Copilot integration announced at Sapphire is genuinely interesting. It's bidirectional — Copilot Studio acts as an orchestration layer consuming Joule services through the A2A protocol, and vice versa. In practice, that means an agent can pull data from Microsoft's stack (Outlook, Teams, Azure AI Search) and execute actions in SAP's stack (Ariba, S/4HANA). Two ecosystems, one workflow.

For customers running SAP on Azure with Microsoft 365 — which is a lot of you — this matters. Your AI assistants are starting to talk to each other across platforms. That's new.

4. Company Memory — the one that deserves more attention

This is the announcement I keep coming back to.

Company Memory is built on the Signavio acquisition. It's a knowledge layer that continuously learns from your policy documents, process models, Slack and Teams messages, email approval chains — and extracts what SAP calls “process atoms.” Small, structured rules that agents can follow. Including the exceptions.

Here's why this matters. Think about a typical manufacturing company running SAP. The senior buyer has a mental list of which suppliers consistently ship invoices with the wrong tax code. The AP team knows that anything from a specific vendor needs a manual GL adjustment before posting. The approval thresholds on paper say €5,000 — but in practice, anything from three specific cost centres goes straight to the plant controller regardless of amount, because of an incident years ago that nobody ever formally documented.

That knowledge lives in people's heads. When those people leave, it goes with them. Anyone who's been through an enterprise transformation has hit this wall — the process documentation says one thing, the actual process does something else, and the gap between the two is where automation projects die.

Company Memory is SAP's attempt to capture exactly that. The tacit knowledge, the workarounds, the “everyone knows you do it this way” processes. Whether it works as described remains to be seen — it's early. But the concept is the most practically important thing announced at Sapphire, because it addresses the problem that has killed more automation projects than any technology limitation.

What this means for your roadmap

If you're running SAP, here's what changed this week — practically.

If you're already on cloud S/4 or RISE: You're in the best position. Evaluate Joule Studio 2.0 as soon as you get access (June onwards). It's free. It's model-agnostic. Test whether it can handle a real use case before you invest in a custom agent framework. If you use Signavio, start understanding how your process models feed into Company Memory.

If you're mid-migration: Your business case just got stronger on the destination side — the AI capabilities landing in cloud S/4 are substantial. But the journey hasn't gotten easier. Consulting capacity is tight, rates are up, and two-thirds of the customer base is competing for the same migration resources. The practical move: keep your migration timeline, but carve out time for your team to see what's actually available. Most S/4 customers I talk to haven't seen a live demo of Joule in their own environment.

If you're still on ECC: The hard truth — most of what was announced at Sapphire requires cloud. But Joule Studio's free tier is worth exploring regardless, because it helps you understand what AI-enabled processes look like before you commit to the full migration economics. And documenting your institutional knowledge (the Company Memory exercise) is something you can start today, on-prem, with zero technology investment.

Regardless of where you are: Brief your team leads on what's actually shipping. I keep coming back to something I've seen in every organisation I've worked inside: the platform capabilities are months or years ahead of customer awareness. Before you evaluate new vendors or new AI tools, understand what you're already paying for. I put together enterprise AI guides that walk through the capabilities of SAP, Microsoft, and Oracle module by module — no vendor language, just what's there and what it does. Worth 30 minutes of your team's time before your next steering committee.

The keynote and the Monday morning

The Sapphire demos were impressive. Agents handling end-to-end processes, cross-platform orchestration, Claude reasoning over live business data. The technology is real. I'm not dismissing any of it.

But demos run on clean data, clear authority, and well-defined processes. They don't need to know which vendor always gets the tax code wrong. They don't need to navigate unwritten approval rules that exist because of an incident nobody documented. They don't need to account for the GL adjustment that everyone does but nobody wrote down.

Company Memory is the feature that's supposed to close that gap. Joule Studio 2.0 is the tool that makes agents accessible. And the Business AI Platform is the architecture that holds it all together.

For the first time, I think SAP has a credible answer to the question at the heart of this newsletter: how do you get from “the AI features exist” to “the AI features are actually changing how we work?”

The answer isn't just better technology. It's better institutional memory. And the organisations that start building that memory now — formally, deliberately, whether they use SAP's tools or a shared document — will be the ones that actually get value when the agents arrive.

That's still a human exercise. And it starts on Monday.

If this issue was useful, forward it to someone navigating the same decisions. That's how Between the Hype finds its audience.

Dive deeper: platform-by-platform guides

Everything mentioned above — the embedded capabilities, the activation steps, the honest limitations — is covered in detail in my practitioner guides. Each one walks through what's actually available today, how to enable it, and where the gaps are.

Executive overview: the strategic decisions that matter →