On May 11, 2026, SAP became a business AI company.
Every traditional SAP firm has been rebranding around that moment ever since. You are right to be skeptical of all of them.
DEBCOR started building production AI against SAP data in 2024 — mostly running outside the SAP system, in our own pipelines, against extracted SAP data. Data cleansing. Validation. Process discovery. Classification. Real workflows in real client environments. By the time SAP announced Joule Studio 2.0, Joule Work, the Autonomous Suite, and the AI Agent Hub at Sapphire 2026, those patterns had been running in our clients' SAP environments for over a year.
Multi-model by design — Anthropic Claude, OpenAI, Amazon Bedrock, and open-source models orchestrated for the job, because no single model is right for every SAP problem and the landscape has shifted every quarter for two years.
Our founder, Gareth de Bruyn, wrote the first ABAP textbook ever published, in 1996. He is the architect who has been building it since 2024.
Here is the build sheet.
The Build Sheet. We've been on the road early.
Multi-model AI orchestration layer with persistent memory
A production AI operating system running across enterprise SAP environments. Routes between Claude, GPT-class, and specialized models per task. Persistent agent memory via vector store — agents remember context across sessions, projects, and quarters. Built by SAP-certified Generative AI Developers on Anthropic Claude and Amazon Bedrock.
SAP Business AI Platform + Joule Studio 2.0
Unified platform announced as the foundation for all SAP agents, merging the AI foundation layer with Business Data Cloud and BTP. Joule Studio 2.0 enters customer hands beginning June 2026, with broad availability rolling through Q3 and Q4.
Cross-vendor agent registry, audit trail and policy guardrails
Every agent action — tool call, model invocation, data access — is logged, tagged, and policy-checked before execution. Works across SAP and non-SAP agents because it sits above the model layer. DEBCOR leadership holds the SAP LeanIX Enterprise Architecture Consultant certification — the same platform SAP's AI Agent Hub is built on.
SAP AI Agent Hub (built on LeanIX)
Central command center to discover, manage, and govern SAP and non-SAP agents. Included in Business AI Platform at no additional charge. Targeted GA in Q3 2026 — currently the only announced cross-vendor governance layer in the SAP ecosystem.
A memory layer that gets smarter every quarter
Our agents continuously learn from the documents, processes, specifications, and conversations that flow through an enterprise — accumulating context that compounds over time rather than resetting with each session. Patent pending. Already deployed in client environments.
SAP Company Memory
Built on the Signavio foundation. Knowledge management and context graph layer that learns from policy documents, process models, and team communications including chats and emails. Operationalizes tacit knowledge — but timeline to broad customer GA was not disclosed at Sapphire.
AI-native integration that adapts instead of breaks
Agents discover, negotiate, and execute integrations against SAP and non-SAP systems on the fly — replacing brittle point-to-point mapping with adaptive contracts between agents and the systems they touch. Patent pending. Running in production today across multiple enterprise integration scenarios.
Joule Work — MCP & Agent-to-Agent (A2A) support
Joule Work adds an agentic harness with computer and file access, MCP, and A2A protocol support — the technical scaffolding for multi-agent orchestration across heterogeneous environments. Partnerships with Microsoft and Google announced for bidirectional agent interoperability.
Multi-model, multi-cloud — Bedrock, Anthropic direct, Vercel AI
We were never single-model. Production workloads run across Amazon Bedrock (for enterprise governance and data residency), Anthropic's API directly (for Opus reasoning), and Vercel AI SDK for edge inference. Foundation model flexibility is the architecture, not a future option.
SAP Business AI Platform — partnerships across the board
SAP announced foundation model partnerships with Anthropic (Claude as a reasoning layer for Joule), AWS (zero-copy data integration with Athena), Google Cloud and Microsoft (A2A interoperability), plus Mistral and Cohere for sovereign deployments. NVIDIA OpenShell as the secure runtime.
n8n-driven AI workflow factory — already in client tenants
Our internal n8n pipelines power live sales intelligence, proposal generation, integration test orchestration, and client onboarding. Same engine SAP announced as a partner. Difference: ours is wired into production data, not a Sapphire demo loop.
n8n inside Joule Studio
Announced at Sapphire 2026: n8n becomes the visual AI workflow orchestration layer inside Joule Studio. A meaningful endorsement of n8n as enterprise-grade — and it confirms the architecture we already operate on.
Codex, Claude Code & Opus inside the delivery loop
Our senior architects ship faster because they're paired with the best coding agents in the world — Claude Code for repository-aware engineering, Codex for parallel exploration, Opus for architectural reasoning. ABAP, BTP CAP, Node, Python, React. Same senior engineer, 10× output.
Joule Studio 2.0 PRD → code generation
The Sapphire demo showed Joule Studio 2.0 generating a PRD, technical specifications, workflow logic, evaluations, and orchestrating multiple agents — model-agnostic and contextualized to SAP semantics. SAP is investing €100M in the partner ecosystem for agent development.
Claude Code + Codex pair-programming ABAP at 10× output
Our senior architects ship modernized ABAP faster because they're paired with the best coding agents in the world. Claude Code handles repository-aware refactors, Codex runs parallel exploration, Opus reasons through architectural decisions. Every commit reviewed by an SAP-certified senior before it hits production. Same senior engineer, ten times the throughput.
SAP-ABAP-2 foundation model
SAP announced SAP-ABAP-2, a foundation model trained on SAP's own ABAP corpus, shipping Q3 2026. Coding assistance grounded in SAP semantics. Once GA, clients will need to integrate it into existing workflows and retrain teams on the new tooling.
AI Data Cleansing in production today
Patent-pending AI agent for SAP master data — machine-learning validation, standardization, and deduplication. Running today in client environments, cleansing legacy data before it reaches Migration Cockpit, after migrations, and on continuous cadences. Reduces design effort by 65–80 percent. Built on Anthropic Claude and Amazon Bedrock for enterprise governance.
SAP Domain Models for data reasoning
SAP Domain Models — foundation models trained on SAP's own metadata and business processes — reach GA in Q3 2026. They will improve how agents reason over SAP data structures. Native autonomous data cleansing tooling has not been announced as a standalone product.
Production agent audit plane — built before SAP announced theirs
Every prompt, tool call, and agent decision in our client environments is logged, attributed, and replayable. Built before SAP announced AI Agent Hub. Governs Claude, GPT-class, and specialized models alongside MCP servers, with session-level observability and root-cause traces for failed agent runs. The same instrumentation we use internally — already deployed for clients.
SAP AI Agent Hub on LeanIX
SAP AI Agent Hub is GA now as a vendor-agnostic command center to inventory and govern agents, LLMs, and MCP servers. Built on SAP LeanIX. Full observability — session health, goal completion, tool-call correctness, root-cause analysis via SAP Cloud ALM — reaches GA Q3 2026.
Readiness diagnostics that start a layer earlier
Proprietary instruments — already live on debcor.com — that diagnose an enterprise's true readiness to absorb agentic AI across data, process, governance, and culture, and classify organizations by adoption profile. The work begins before any agent is deployed, helping clients prepare the ground so SAP's autonomous suite lands well when it arrives.
Complements the SAP roadmap
SAP's frame is the suite and the agents that operate inside it. Our readiness work runs alongside — preparing the people, data, and governance posture clients need so that, when each Sapphire announcement reaches GA, adoption is fast and the outcomes hold.
How this compounds when SAP ships.
We already know the stack SAP just announced. Joule Studio, AI Agent Hub, Company Memory, Business AI Platform — every layer of it. Our senior architects — each holding the full set of SAP architect-tier credentials, an unusual depth even for senior SAP firms — have been working in these architectural patterns for over two years. So when SAP's roadmap hits GA, there's no learning curve for our team. But we don't stop there. The stack we've built sits alongside SAP's — it gets your organization to those outcomes sooner than SAP's promise dates, and it lives beyond them as a complement that compounds your value realization quarter after quarter.
The stack we're fluent in, today.
Every layer of the SAP Business AI Platform announced at Sapphire 2026 has a counterpart we already operate with. As SAP delivers, we connect the two — bringing your environment from where it is now to the autonomous enterprise SAP described, without a learning curve in the middle.
DEBCOR is the product-grade SAP partner • Three patents pending • Every architect-tier SAP credential SAP offers
What CIOs are asking after Sapphire 2026.
What were the most significant AI announcements at SAP Sapphire 2026?
Three threads run through every SAP Sapphire 2026 keynote. First, Joule moved from copilot to autonomous agent orchestrator — Joule Studio 2.0 was announced and a fully managed runtime is targeted for Q3 2026 GA. Second, SAP committed publicly to open agent interoperability as a founding contributor to the A2A (Agent-to-Agent) protocol alongside Google Cloud. Third, the Generative AI Hub on BTP opened to a wide pool of foundation models, including Google Gemini 2.0 Flash, Flash Lite, Gemini 1.5, and Text Embedding Gecko, accessed through a single harmonised API. The animating concept tying it together is the Autonomous Enterprise — SAP's positioning that within three years, AI agents will run finance, supply chain, procurement, and HR processes end-to-end while humans focus on exception handling and strategy.
What is SAP's 'Autonomous Enterprise' vision — and is it real?
The vision is real and the platform is maturing — but it is not a single product you buy. It is a multi-year architecture: SAP Joule (the orchestrator), the SAP AI Agent Hub (governance and discoverability), the SAP Knowledge Graph (grounded business context), Joule Studio (custom agent authoring), MCP and A2A (open connectivity protocols), and SAP's Generative AI Hub (model access). Pieces are GA today (original Joule Studio, MCP support, A2A on BTP, Gemini in Generative AI Hub), others are in Early Adopter Care (managed Joule Studio), and some are still preview (full bidirectional A2A). The honest read is that the foundation works now for selected use cases, and the platform deepens through 2026–2027. The risk for SAP customers is not vapourware; it is sitting still while peers start building.
How is the SAP-Google Cloud partnership changing what's possible on BTP?
Three concrete things. First, model choice: Google Gemini 2.0 Flash, Flash Lite, Gemini 1.5, and Text Embedding Gecko are now selectable via SAP's Generative AI Hub harmonised API alongside SAP-hosted and other partner models — meaning Joule and custom agents can pick the right model per use case without lock-in. Second, multimodal RAG: Joule integrations with Google Video Intelligence (on-screen text detection) and Speech-to-Text (audio transcription) unlock retrieval over video and audio — for example, finding the specific moment in a training video that answers a Joule user's question. Third, open agent interoperability: SAP is a founding contributor to the A2A (Agent-to-Agent) protocol that the two companies launched together, ensuring agents built on either side can collaborate across vendor boundaries.
Do I need to wait for Joule Studio 2.0 to start building?
No — and waiting is the bigger risk. Original Joule Studio is GA, MCP support is GA, and A2A on BTP works today using SAP's open-source joule-a2a-agent-toolkit (it scaffolds an A2A-compliant agent in LangGraph or CAP, deploys to Cloud Foundry, and registers with Joule in one command). What Joule Studio 2.0 specifically adds is bidirectional A2A improvements and a fully managed runtime targeted for Q3 2026 GA — neither of which is required to start delivering value now. The teams that build their first Joule + custom-agent integration on the current stack will be the first to ship managed-runtime production scenarios when it lands.
How is DEBCOR already running what SAP just announced?
DEBCOR has been operating the open-source equivalent of SAP's announced stack — LangGraph for orchestration, MCP for connectivity, n8n for workflow design — in production with SAP customers since roughly 18 months before Sapphire 2026. That is the central thesis of this page: we are not selling a future you have to wait for. The same Generative AI Hub, AI Core, AI Agent Hub, A2A, and Knowledge Graph that SAP described are layers we have engineering experience integrating, governance frameworks designed around, and Catalyst-platform tooling to accelerate. When SAP Joule Studio 2.0 managed runtime is GA, our clients will migrate their existing agent estate onto it — they will not be starting from a slide deck.
What should I do now if I watched the Sapphire 2026 keynote?
Three moves, in order. First, audit your RISE with SAP entitlement — most organisations have Joule activation included and unused, and unlocking it is the lowest-friction starting point. Second, pick one high-volume, structured process (AP invoice matching, dispute resolution, IDoc routing, master data maintenance) and scope a 90-day pilot for a Joule + A2A agent grounded in your actual SAP data. Third, set up a governance baseline now — input validation, output policy, audit logging — so that when agent volume scales, the controls are already in place rather than retrofitted under pressure. DEBCOR offers all three as packaged engagements. The right time to start was the day before Sapphire 2026 ended.
SAP Sapphire 2026 changes how every customer runs.
The teams already building in this stack will move first.
If you watched the keynote and asked “who do we trust to take us there?” — that's the conversation we have every day.