What Is Agent iXi? How Agentic AI Systems Work in Automation

Agent iXi is Bosch Rexroth’s answer to a simple question: what happens when an AI tool stops answering and starts doing? The industrial automation company laid out the system on November 5, 2026. It takes a natural-language task and a drawing, then builds axes, kinematics and an IEC-61131-3 PLC program inside the ctrlX AUTOMATION environment — a concrete look at how agentic AI systems work in automation.

That is a different animal from a chatbot. Agent iXi plans a sequence of engineering actions and calls the software functions it needs along the way. Bosch Rexroth says it can configure inputs and outputs on its own, with access still tied to the permissions of the logged-in user.

🔸What Agent iXi Changes in Industrial Automation
💡Who Is Building the Agentic Automation Platform
✨What Agentic AI Adds to Engineering Workflows
💡Agentic AI Numbers: What the Available Figures Show
⚠️Security Controls and Adoption Risks
🔷What Happens Next for Bosch Rexroth's AI Architecture
❓FAQ: How Agentic AI Systems Work in Automation

What Agent iXi Changes in Industrial Automation

From answering questions to carrying out engineering steps

Bosch Rexroth already sells Ask iXi, a chatbot that answers questions, explains documentation and generates code. That is the assistant model: a person asks, reviews the result and finishes the technical work.

Agent iXi goes further. It turns a requested outcome into individual steps and calls the functions it needs in the automation environment. Bosch Rexroth calls the shift assisted code generation to agentic engineering.

What the agent can produce

Through interfaces to ctrlX PLC Engineering and ctrlX I/O Engineering, the agent generates IEC-61131-3 code in the engineering context and configures I/Os automatically. It draws on the ctrlX AUTOMATION software development kit, open APIs and the tools engineers already use to build automation applications.

The reach spans much of the engineering process. Bosch Rexroth lists component and topology selection, motion and drive setup, I/O and HMI design, programming, system monitoring and documentation among the possible tasks.

The three-axis handling example

A three-axis handling system shows how the workflow runs. Conventionally, an engineer designs the axes, configures the kinematics, sets the coordinate system and motion limits, and writes the PLC program.

With Agent iXi, the user hands over a drawing containing the important details plus a natural-language description of the required motion. The agent creates the axes and kinematics and generates the PLC program for that task. The example illustrates the intended interaction.

Who Is Building the Agentic Automation Platform

Bosch Rexroth and ctrlX AUTOMATION

Bosch Rexroth is wiring agentic AI into the platform itself rather than shipping it as a separate chatbot. Agent iXi reaches the engineering tools, applications, documentation, data and functions associated with ctrlX AUTOMATION according to the context of the request.

The platform includes ctrlX OS, the applications that run on it and the ctrlX Data Layer. Bosch Rexroth describes the data layer as a way to provide continuous access to data and functions within the platform while respecting the relevant controls.

The executive view of the user experience

Steffen Winkler, Senior Vice President Sales in Bosch Rexroth’s Automation & Electrification Solutions business unit, said the user’s work changes because the user increasingly describes the desired result instead of performing every technical step separately. He also said the agent handles the engineering steps in a way that remains transparent to the user.

That design places the human at the level of the intended outcome while the agent handles the sequence underneath. The practical value depends on whether the generated configuration, code and machine behavior meet the requirements of the specific application.

Tools, standards and hardware partners

Model Context Protocol, or MCP, provides the connection between the AI agent or language model and the software tools. Bosch Rexroth is using MCP interfaces for ctrlX PLC Engineering and ctrlX I/O Engineering so that Agent iXi can work with those systems rather than only describe how a person might use them.

Bosch Rexroth is also working with AMD on a hardware base intended to support local execution of high-performance AI models with ctrlX OS. The company describes that local option as part of its future direction, alongside cloud-based inference.

What Agentic AI Adds to Engineering Workflows

The user describes the result

For users, the main change is the level at which they interact with the automation system. Instead of manually carrying out every configuration step, they can provide a target, a drawing and the relevant requirements, then review the work produced through the connected tools.

Technical judgment does not disappear. It moves: from entering every parameter individually toward specifying the result, checking the generated engineering work and confirming that the machine behaves as intended.

Skills give the agent task-specific context

Bosch Rexroth uses skills to supply the knowledge and rules required for a particular task. A motion application, for example, can use a curated skill that tells the agent how axes and kinematics are created within ctrlX AUTOMATION.

The company describes the architecture as “one agent – many skills.” Rather than loading every possible instruction into every interaction or creating a separate agent for each job, the central agent can call the capability that matches the task.

An ecosystem that can expand through applications

Each application can bring its own skills. An improvement to an app can therefore expand what the central agent can do, while third-party applications can make their own domain expertise available to AI.

This approach makes the application ecosystem part of the agent’s capabilities. It also creates a requirement for those skills to be properly curated and validated, because the agent’s output depends on the instructions and functions it is allowed to use.

Agentic AI Numbers: What the Available Figures Show

Business leaders and expected returns

According to the reference material, 92% of leaders expect agentic AI to deliver measurable return on investment within two years.

The expectation helps explain why companies are looking beyond chat interfaces. If an AI system can coordinate a complete workflow rather than produce one answer, businesses may assess it against resolution time, manual effort and the number of systems involved in a process.

Time spent searching for information

Deloitte estimates that a typical employee can spend nearly 40% of the workweek searching for information. In a fragmented enterprise environment, an agent that can retrieve context and act across approved systems is intended to reduce that coordination burden.

For industrial automation, the equivalent problem is the effort required to locate documentation, understand a machine configuration and transfer information between engineering tools. Agent iXi’s platform integration targets that handoff rather than only improving the wording of an answer.

Software cost pressure

Gartner has reported enterprise software-as-a-service renewal costs rising by 10% to 20% or more, ahead of IT budget growth. That pressure can make automation attractive, but it also raises the standard for proving that an agent improves a workflow rather than adding another software layer.

Neither the cited figures nor Bosch Rexroth’s announcement supplies a cost saving, deployment count or productivity improvement for Agent iXi. What exists is a description of the architecture and intended uses, not a measured business outcome.

Security Controls and Adoption Risks

Permissions follow the logged-in user

MCP communication is encrypted and authenticated, Bosch Rexroth says. The agent also takes the permissions of the logged-in user into account, which is intended to limit the data and functions available to it.

That control matters because an agent connected to engineering tools has a wider operating range than a chatbot that only returns text. The practical boundary is set by the user’s permissions, the configured integrations and the functions exposed through the platform.

The data boundary inside ctrlX AUTOMATION

The ctrlX Data Layer provides secure, continuous access to data and functions inside the platform, per Bosch Rexroth. This gives the agent a structured route to the information and operations required for a task rather than unrestricted access to every system in an enterprise.

Companies still need to decide which applications, data and actions should be exposed. A useful agent depends on access to enough context to complete the job, while a safe deployment limits what can be changed and under which account.

Cloud models, local models and technology lock-in

ctrlX AUTOMATION is not tied to one large language model or one AI provider, according to Bosch Rexroth. Companies can connect different models and third-party agents according to their existing infrastructure, while choosing between cloud-based inference and the future local option described by the company.

This model is intended to reduce technological lock-in. It also leaves implementation choices with the company using the platform, including which model receives engineering information and where that processing takes place.

What Happens Next for Bosch Rexroth’s AI Architecture

Local inference is the next platform direction

Bosch Rexroth plans to enable local execution of high-performance AI models on ctrlX OS with hardware developed with AMD. Processing sensitive engineering and machine data on site could reduce the need to send that information to a cloud service.

No launch date or performance figures for the local execution plan have been given.

Choice remains part of the adoption model

Companies can choose the agents and models they connect, decide where inference takes place and use the AI feature only if it fits their needs, Bosch Rexroth says. Users who do not want the feature can choose not to install it.

The immediate story is therefore not that every automation system is becoming autonomous. Industrial platforms are being redesigned so an AI agent can interpret a goal, select task-specific skills and work through approved engineering tools.

FAQ: How Agentic AI Systems Work in Automation

What is the practical difference between an AI assistant and an agentic system?

An assistant mainly returns information, explanations, code or other generated content. An agentic system connects an intended outcome to a sequence of actions and uses integrated tools to carry out that sequence under configured permissions.

What does Model Context Protocol do in Bosch Rexroth’s platform?

MCP provides a standardized connection between an AI agent or language model and software systems. Bosch Rexroth uses MCP interfaces to let Agent iXi interact with ctrlX PLC Engineering and ctrlX I/O Engineering.

Why does Agent iXi use skills?

Skills provide the instructions, rules and task-specific knowledge needed for a particular job. In a motion application, a curated skill can guide the agent in creating axes and kinematics inside ctrlX AUTOMATION.

Can Agent iXi connect to more than one AI model?

Yes. Bosch Rexroth says ctrlX AUTOMATION is not limited to one underlying model and can connect different models or third-party agents based on a company’s AI infrastructure and requirements.

Where can the AI inference take place?

The architecture supports cloud-based models, while Bosch Rexroth is working with AMD toward local execution of high-performance models through ctrlX OS. The company has not supplied a date or performance specification for that local capability.

What can Agent iXi access?

Depending on the user’s permissions and the configured platform connections, the agent can work with engineering tools, ctrlX OS, applications running on it, and data and functions available through the ctrlX Data Layer.