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

Agent iXi represents Bosch Rexroth’s response to a fundamental shift: moving beyond AI that merely answers queries to technology that actively executes tasks. Unveiled by the industrial automation firm on November 5, 2026, the system processes a drawing and a natural-language task to construct axes, kinematics, and an IEC-61131-3 PLC program directly within the ctrlX AUTOMATION environment—offering a practical demonstration of agentic AI in automation.

This approach diverges significantly from traditional chatbots. Agent iXi maps out a series of engineering steps and invokes necessary software functions along the way. According to Bosch Rexroth, it can independently configure inputs and outputs, though its access remains governed by the active user’s permissions.

What Agent iXi Changes in Industrial Automation

From answering questions to carrying out engineering steps

Bosch Rexroth already offers Ask iXi, a chatbotive tool that handles inquiries, explains documentation, and writes code. This represents a traditional assistant framework where a human prompts the system, reviews the output, and completes the technical labor manually.

Agent iXi takes this a step further by converting a desired goal into individual actions, executing them directly inside the automation environment. Bosch Rexroth characterizes this evolution as a transition from assisted code generation to agentic engineering.

What the agent can produce

Utilizing interfaces for ctrlX PLC Engineering and ctrlX I/O Engineering, the agent generates IEC-61131-3 code within the native engineering framework and automatically sets up I/Os. It leverages the ctrlX AUTOMATION software development kit, open APIs, and established tools that engineers already rely on for automation projects.

The system’s capabilities cover a broad spectrum of the engineering lifecycle. Bosch Rexroth highlights component and topology selection, drive and motion configuration, HMI and I/O design, programming, documentation, and system monitoring as potential tasks.

The three-axis handling example

A three-axis handling system demonstrates how this workflow operates in practice. Typically, an engineer must design the axes, configure kinematics, establish motion limits and coordinate systems, and write the corresponding PLC program.

With Agent iXi, the operator provides a drawing featuring essential details alongside a plain-text description of the required movement. The agent then sets up the axes and kinematics while generating the appropriate PLC program. This scenario outlines the intended collaborative model.

Who Is Building the Agentic Automation Platform

Bosch Rexroth and ctrlX AUTOMATION

Rather than packaging agentic AI as a separate chatbot, Bosch Rexroth integrates it directly into the platform itself. Agent iXi interacts with ctrlX AUTOMATION’s engineering tools, applications, documentation, functions, and data depending on the context of the prompt.

The underlying platform encompasses ctrlX OS, its hosted applications, and the ctrlX Data Layer. Bosch Rexroth defines this data layer as a mechanism providing continuous access to platform functions and data while maintaining appropriate security controls.

The executive view of the user experience

Steffen Winkler, Senior Vice President of Sales for Bosch Rexroth’s Automation & Electrification Solutions division, noted that user responsibilities shift toward defining desired outcomes instead of executing every isolated technical procedure. He emphasized that the agent manages these engineering workflows transparently to the user.

This architecture positions the human operator at the level of high-level objectives while the agent handles underlying sequencing. Ultimately, the practical value relies on whether the resulting machine behavior, code, and configurations align with application specifications.

Tools, standards and hardware partners

The Model Context Protocol (MCP) serves as the bridge connecting language models and AI agents to software utilities. Bosch Rexroth integrates MCP interfaces into ctrlX PLC Engineering and ctrlX I/O Engineering, enabling Agent iXi to directly manipulate these systems instead of merely outlining instructions for humans.

Additionally, Bosch Rexroth is collaborating with AMD to develop a hardware foundation capable of supporting local, high-performance AI model execution via ctrlX OS. The firm presents this local deployment option as a future path alongside traditional cloud-based inference.

What Agentic AI Adds to Engineering Workflows

The user describes the result

For end-users, the primary adjustment lies in how they interact with the automation ecosystem. Rather than manually configuring every parameter step-by-step, operators supply a target objective, a technical drawing, and relevant criteria, then validate the output generated through connected applications.

This does not eliminate technical expertise; rather, it shifts it. Engineers move away from entering individual parameters toward defining goals, auditing generated work, and verifying correct machine operation.

Skills give the agent task-specific context

Bosch Rexroth employs modular “skills” to supply necessary rules and domain knowledge for specific assignments. For instance, a motion control application can utilize a dedicated skill instructing the agent on how to build axes and kinematics inside ctrlX AUTOMATION.

The company outlines this framework as “one agent – many skills.” Instead of burdening every interaction with exhaustive instructions or deploying a separate agent for every task, a central agent invokes the specific capability required for the job.

An ecosystem that can expand through applications

Because each software application can introduce its own skills, updates to an app can naturally broaden the central agent’s capabilities, while third-party solutions can integrate their own domain-specific expertise into the AI.

This methodology embeds the application ecosystem directly into the agent’s skill set. At the same time, it necessitates rigorous curation and validation of these skills, given that the agent’s performance relies entirely on the functions and instructions it is permitted to execute.

Agentic AI Numbers: What the Available Figures Show

Business leaders and expected returns

Reference data indicates that 92% of corporate leaders anticipate measurable return on investment from agentic AI within a two-year timeframe.

This outlook clarifies why organizations look beyond conversational interfaces. When an AI system manages an entire workflow rather than simply answering a prompt, enterprises evaluate its utility based on manual effort reduction, resolution speed, and the volume of systems integrated into a process.

Time spent searching for information

According to Deloitte estimates, typical employees may spend close to 40% of their workweek hunting for information. Within fragmented corporate structures, an intelligent agent capable of cross-referencing context and operating across validated systems aims to minimize this coordination overhead.

In industrial automation, the parallel challenge involves the time required to trace documentation, grasp machine configurations, and move data between various engineering utilities. Platform integration in Agent iXi targets this friction point rather than just refining textual explanations.

Software cost pressure

Gartner reports that enterprise SaaS renewal costs have climbed by 10% to 20% or more, outpacing standard IT budget growth. While this pressure makes automation appealing, it also raises the bar for proving that an agent genuinely streamlines workflows rather than adding another software expense layer.

Neither Bosch Rexroth’s announcement nor the cited figures provide concrete deployment totals, cost reductions, or productivity gains for Agent iXi. Available documentation details platform architecture and intended use cases rather than verified business metrics.

Security Controls and Adoption Risks

Permissions follow the logged-in user

Bosch Rexroth notes that MCP communication is authenticated and encrypted. Furthermore, the agent inherits the permissions of the currently logged-in user, restricting the data and software functions it can access.

This safeguard is critical because an agent linked to engineering software possesses a much broader operational scope than a text-only chatbot. Practical limitations are enforced by user credentials, integrated permissions, and platform-exposed functions.

The data boundary inside ctrlX AUTOMATION

According to Bosch Rexroth, the ctrlX Data Layer supplies secure, ongoing access to internal platform data and tools. This offers the agent a structured pathway to necessary information and operations, avoiding unrestricted access to an entire corporate IT infrastructure.

Organizations must still determine which data, actions, and applications to expose. A functional agent requires enough contextual access to complete tasks, while secure deployment mandates strict limitations on modifications and user account privileges.

Cloud models, local models and technology lock-in

ctrlX AUTOMATION is designed independently of any single large language model or proprietary AI vendor, per Bosch Rexroth. Enterprises can integrate diverse models and third-party agents that fit their current infrastructure, choosing between cloud inference and future local execution capabilities.

This flexibility aims to prevent technological vendor lock-in. Simultaneously, it places implementation responsibility on the platform user, who must decide which models handle engineering data 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 directly on ctrlX OS utilizing hardware co-developed with AMD. Processing sensitive machine and engineering data on-premises could decrease reliance on external cloud services.

To date, no specific launch timeline or performance benchmarks for local model execution have been released.

Choice remains part of the adoption model

Bosch Rexroth emphasizes that companies retain the freedom to select integrated models and agents, determine inference locations, and deploy AI features only when aligned with operational needs. Operators who prefer not to use the functionality can simply omit its installation.

Consequently, the immediate narrative is not that all industrial automation systems are turning autonomous. Instead, industrial platforms are evolving so AI agents can interpret objectives, pull relevant skills, and operate through authorized engineering environments.

?FAQ: How Agentic AI Systems Work in Automation

01What is the practical difference between an AI assistant and an agentic system?

An assistant primarily delivers information, explanations, or code outputs. An agentic system links a stated goal to a sequence of actions, utilizing integrated software tools to execute that sequence within established permission limits.

02What does Model Context Protocol do in Bosch Rexroth’s platform?

MCP establishes a standardized connection between software environments and AI agents or language models. Bosch Rexroth leverages MCP interfaces to enable Agent iXi communication with ctrlX PLC Engineering and ctrlX I/O Engineering.

03Why does Agent iXi use skills?

Skills supply the rules, instructions, and task-oriented knowledge required for specific assignments. For example, a specialized skill can direct the agent on setting up kinematics and axes within ctrlX AUTOMATION for motion tasks.

04Can Agent iXi connect to more than one AI model?

Yes. Bosch Rexroth states that ctrlX AUTOMATION remains open to multiple foundational models, allowing companies to link third-party agents or different models according to infrastructure needs.

05Where can the AI inference take place?

The architecture supports cloud-based model execution, while Bosch Rexroth collaborates with AMD on local processing capabilities via ctrlX OS. No official performance metrics or release dates have been published for local processing.

06What can Agent iXi access?

Subject to user permissions and active platform integrations, the agent can interact with engineering software, ctrlX OS, hosted applications, and functions or data accessible through the ctrlX Data Layer.