Key Takeaways
- Intelligence Center X (ICX) is Siemens' industrial AI orchestration platform, announced June 1, 2026. It combines Mendix, Graph Studio, and AI Studio under one set of governance controls.
- Siemens built the context layer first: a knowledge graph and industrial ontologies, ahead of the agents. That order is the point.
- Mendix is the action layer, where an agent's recommendation becomes a task, an approval, and a record.
- As of September 28, 2026, Siemens hasn't published ICX pricing or a general-availability date. Mendix 11.12 already includes Agents Kit 2, Model Context Protocol (MCP) support, and human-in-the-loop controls.
- Whatever you adopt, the outcome depends on shared definitions, agent permissions, and a named human at every checkpoint.
Siemens built the context layer first. Before the agents and before the apps, it built a knowledge graph and a set of industrial ontologies.
We've been saying for a while that structure has to come before AI. It's good to see a company with Siemens' weight build its platform in that order.
If you run Mendix, Intelligence Center X (ICX) puts your platform in a specific job: the place where an AI recommendation turns into governed work.
Here's what ICX is, what you can build on Mendix today, and what to settle before an agent touches a real process.
What is Intelligence Center X?
Intelligence Center X is Siemens' industrial AI orchestration platform, announced June 1, 2026 at Realize LIVE Americas.
It combines Mendix, Graph Studio, and AI Studio so that data context, AI models, and applications work together under one set of governance controls. Siemens hasn't published pricing or a general-availability date.
Siemens describes it as software where "people and AI agents work together with shared context, workflows and lifecycle intelligence." It comes in three deployment patterns:
- Layered on Siemens' own AI products
- Standalone for asset-intensive companies that run other vendors' operational technology
- As a pure agentic-enterprise platform for financial services, insurance, healthcare, government, and retail
Let’s clear up a term first: An ontology is a shared map of what things in your business are and how they connect. A part links to its supplier, the plant that built it, and the quality issue filed against it.
Why did Siemens build it?
Because AI pilots stall when the AI doesn't know the business and nothing governs what it does.
Mendix calls the result the AI value gap: the distance between what AI can do in a demo and what it delivers inside a real process.
Tony Hemmelgarn, president and CEO of Siemens Digital Industries Software, put the fix in one line: "AI only delivers real value when it is embedded in how work gets done."
Picture a model that flags a likely quality problem. The flag has no value until someone connects it to the product configuration, the supplier batch, and the plant asset involved. Then it needs an owner, an approval rule, and a corrective action. Context and workflow do that work. The model alone can't.
Where does Mendix fit in Intelligence Center X?
Siemens organizes ICX in four layers: Context, Intelligence, Action, and Trust.
- Graph Studio builds the knowledge graph across your systems.
- AI Studio builds and governs the machine learning models.
- Mendix is the action layer, where teams "build agents and intelligence applications, and orchestrate them into action," in Siemens' words.
- Governance runs across all of it.
Siemens has been building toward this since it acquired Mendix in 2018. We covered why Siemens is betting on Mendix as the digital thread earlier.
In practice, that means an agent's recommendation shows up in a Mendix app as a task with an owner, an approval step, and an audit trail. That's the orchestration we care about: a specific quality investigation or vendor onboarding queue, with a person accountable at the end of it.
Verdantix analyst James Prestwood notes that Siemens is emphasizing governance as "one of Intelligence Center X's strongest differentiators, brought to it through Mendix."
Vivix Vidros Planos, a flat-glass manufacturer in Brazil, is Siemens' lead example. It built nearly 30 Mendix applications connecting operational and IT data across SAP S/4HANA, Siemens Industrial Edge, and Snowflake, plus an AI "Virtual Engineer" built on ICX with Amazon Bedrock and Claude.
Siemens reports an 85% reduction in production-issue resolution time and 6,000 hours of manual work recaptured in a year.
Those are Siemens' numbers for one customer. Your results will depend on your data, your systems, and how well people adopt the workflow. Read the description closely, though: most of it is connected applications and data. The agent is one piece.
What can Mendix teams build today, and what depends on Siemens?
Most of the action layer is available now. Mendix 11.12, the long-term-support release from June 30, 2026, includes Agents Kit 2, MCP support, and human-in-the-loop controls. Which parts you can use depends on your Studio Pro version and license.
The context layer depends on Siemens' rollout, which has no published date.
| Capability | In Mendix's published docs today | What ICX adds, per Siemens (availability unconfirmed) |
|---|---|---|
| Build agents | Agents Kit 2 and the Agent Editor in Studio Pro 11.12 and later | Agents grounded in a shared knowledge graph |
| Connect to tools | MCP client and server modules | Industrial ontologies for engineering and manufacturing data |
| Choose models | Claude, OpenAI, Mistral, Gemini, Amazon Bedrock, and others | Custom ML models built and governed in AI Studio |
| Keep humans in control | Human-in-the-loop controls and approval flows | A platform-wide layer for policy, traceability, and oversight |
We've covered the build side before, in how to build AI agents in Mendix without AI expertise and what Maia offers development teams.
MCP deserves a closer look. It's the open standard that lets an AI agent call outside tools and data. A Mendix app can act as a client, using other tools, or as a server, letting an outside agent use its functions.
The server side is worth thinking about now. More companies are putting their own AI front end on top of the platforms they run and skipping the platform's own screens. If an agent becomes the way people reach your Mendix apps, each app becomes a set of tools that agent calls.
A tool is only as safe as its data model and permissions. Clean entities and tight roles were nice to have when a person clicked through a screen. When an agent calls the app directly, they're the control surface.
What should you settle before you put an agent in a Mendix workflow?
Five things, and none of them are glamorous.
- Shared definitions. Do "product," "asset," "supplier," and "batch" mean the same thing in every system? An agent inherits every inconsistency you have.
- Agent permissions. Decide what it may read, recommend, and execute. Build that into tiered permissions inside the app, not into a policy document nobody opens.
- A named human at each checkpoint. Someone approves. Someone handles exceptions. Both get recorded. Mendix's own approach to AI leans on checkpoints rather than black boxes, and your workflows should too.
- Traceability. You should be able to trace any recommendation back to its sources and the rules that shaped it.
- One process, one measure. Pick a single decision-heavy process and track resolution time or manual hours. Model accuracy alone won't tell you whether the work got better.
This is what we mean by HI × AI = FIRE. The model supplies speed. Your people supply judgment. The platform records both.
Get your Mendix apps ready for AI agents
You don't need ICX to start. You need clean definitions, tight permissions, and a human at every checkpoint, and you can put those in place on the Mendix version you run today. We outlined the path in building intelligent applications on Mendix.
We're a Mendix Gold Partner, and we build this way for a reason: we are the experts, so our clients don't have to be. If you're deciding where an agent belongs in your Mendix environment, schedule a Mendix consultation. We'll look at your data model, your permissions, and one process worth starting with. An App Checkup is another place to start, and our AI services cover the agent work. No pressure, no pitch.
Frequently Asked Questions
Does Intelligence Center X replace Mendix?
No. Mendix is one of ICX's components, the layer where agents and applications are built and orchestrated. Nothing in Siemens' published materials suggests existing Mendix apps get replaced. How much an app benefits from ICX depends on its data, its integrations, and your license.
Is Intelligence Center X available yet?
Siemens announced ICX on June 1, 2026, and as of September 28, 2026 it hasn't published pricing, licensing terms, or a general-availability date. Ask Siemens or your Mendix representative for current terms for your deployment.
Do I need Mendix 11.12 to build AI agents?
Agents Kit 2 requires Studio Pro 11.12 or later. Agents Kit 1 runs on Studio Pro 10.24 and later. Mendix 11.12 is the first long-term-support release since 10.24, which makes it a stable version to plan around.
Is Intelligence Center X only for manufacturers?
Siemens' focus is industrial: engineering, manufacturing, supply chain, quality, and service. It also describes a pure agentic-enterprise deployment for financial services, insurance, healthcare, government, and retail. Its published customer examples so far lean industrial.
