SAP's Autonomous Supply Chain: Five Pillars and a Copilot Named Joule
Since the industrial revolution, companies have been trying to solve the same underlying supply chain problems: managing forecast uncertainty, gaining real-time visibility, orchestrating information across departments, delivering more value to customers, and constantly adapting to a shifting technological and geopolitical context. At the Tech For Industry Show, Sylvain Chatonnier, Solution Advisor Supply Chain (SAP France) presented the vendor's current answer to these challenges: the autonomous supply chain.
Five Pillars Covering the Entire Product Lifecycle
SAP structures its strategy around five pillars, covering what the vendor calls "design to operate" — the full cycle from design to end of life, relevant to both discrete and process manufacturing:
- Long- and short-term planning, with the challenge of synchronizing shop-floor variations with their impact on long-term planning.
- Product definition, including eco-design and accounting for the impact of bill-of-materials changes across the entire supply chain.
- Manufacturing, covering every production mode (engineering to order, make-to-stock, repetitive manufacturing).
- Logistics, both internal (managing stock and components) and external (distribution, transport).
- Industrial asset management, including for companies operating "equipment as a service" models — the example cited being a compressor manufacturer selling compressed-air capacity rather than a machine.
At the center of this ecosystem sits the S/4HANA ERP, surrounded by complementary solutions (long-term planning, MES for manufacturing execution, collaborative predictive maintenance) and cross-cutting elements like the SAP Business Network, a collaborative platform for exchanging with suppliers and subcontractors on a shared, consistent data foundation.
Joule: The Copilot at the Center of SAP's AI Strategy
Joule is SAP's new interface — an evolution from historical transactions and web applications toward natural-language interaction similar to market AI assistants. Within the autonomous supply chain, Joule provides assistants for each major domain (design, logistics, manufacturing), each acting as a "team lead" overseeing a multitude of operational agents handling specific tasks.
The Technical Foundations: Knowledge Graph and Governance
To make these agents' reasoning reliable, SAP relies on two essential building blocks. First, the SAP Knowledge Graph, a structure linking all business objects across the SAP ecosystem — customers, orders, suppliers, carriers — and capitalizing on decades of process history. Second, an agent governance tool, compared by the speaker to an "HRIS for agents" (by analogy with a tool like SuccessFactors for human employees): it defines which agent is allowed to do what, which systems it can interact with, and orchestrates both SAP agents and third-party ones.
SAP rounds out this architecture with a so-called "Business Data Cloud" layer, making it possible to integrate third-party data (weather, geopolitical or climate data, for example) and structure it according to SAP semantics.
An Agent Catalog ("Activate") Organized by Pillar and by Industry
SAP provides a catalog of ready-to-use agents, organized both by functional pillar (design, manufacturing, planning, asset management) and by vertical scenarios specific to certain industries — renewable energy, agribusiness, life sciences — co-developed with specific clients, with availability for all SAP clients announced between the end of August and the end of the year.
Four Levels of Demonstration, from Planning to Maintenance
The session illustrates the growing sophistication of these agents through several demonstrations.
Resolving a Stockout in Long-Term Planning
In a first scenario, a demand planner for the Nordic pharmaceutical market finds, every morning, a list of alerts already pre-analyzed by agents that ran overnight: identifying a coverage shortfall, suggesting sourcing part of the missing volume from a nearby distribution center, supplementing with semi-finished products available at another European site — with full traceability of the reasoning (expiry dates, transit times, manufacturing lead times) and an alert on the potential impact of this decision three weeks later on another market.
Launching a Manufacturing Order with a Simple Prompt
The second scenario illustrates Joule Work, a command-prompt-style interface for launching a manufacturing order in natural language, without navigating traditional transactions. The agent asks the user about necessary choices (for example, which manufacturing version to use), performs availability and capacity checks, then executes the action directly — inheriting the user's own permissions.
Analyzing the Impact of a Bill-of-Materials Change with Multiple Agents
The third scenario illustrates multi-agent orchestration: on simple request, the system identifies pending engineering change requests, checks the quality of associated data (planning types, lead times), then simulates the impact of the new configuration on production times, lead times and safety stock compared to historical data — chaining together several specialized agents.
A Predictive Maintenance Scenario Triggered by IoT
The final scenario covers maintenance asset management: an industrial pump connected to vibration IoT sensors reports a critical alert. The system automatically consolidates the context (equipment history, associated notifications), creates a notification then a maintenance order after checking there's no duplicate, assigns it to a qualified, available technician, who can then check the equipment's history and required parts directly from their mobile app.
Building Your Own Agents, Without Necessarily Coding
Beyond the catalog, Joule also allows the development of custom agents, applications or workflows. One example presented: a client historically using an IBM AS/400 system for warehouse management, wanting to migrate to SAP Extended Warehouse Management without disrupting its logistics teams' habits. From a simple prompt describing the need, the system generates a complete functional specification (personas, mobile screens, technical elements), allows a quick usability test, then delivers a mobile app retaining visual cues familiar to operators while natively relying on SAP EWM.
Security, Governance and AI Model Choices
On the security side, SAP offers access to several market AI models (chosen by the client, or automatically selected by SAP based on context), but these models remain encapsulated within SAP's Business AI Platform rather than being queried via public access. The vendor also has a proprietary model dedicated to generating synthetic data, useful when historical data is missing within an organization.
The agent governance tool tracks the privileges granted to each agent, the models used, associated consumption, as well as the process steps requiring human validation — visually represented within the tool. It also includes automatic detection of duplicate agents, to avoid redundancy. Finally, SAP Cloud ALM allows usage to be monitored across the entire application architecture, with a detailed audit trail of actions carried out by agents.
Key Takeaways
The vision presented by SAP rests on the idea that supply chain autonomy can't be built without a trusted infrastructure: a foundation of structured knowledge (the Knowledge Graph), rigorous agent governance, and a constant principle of keeping humans at the center of strategic decisions, even as many operational steps become progressively automated.
Watch the full session, along with the other Tech For Industry Show replays, on our dedicated page.
