Siemens, SUPCON, and INOVANCE Are All Going All-in on “AI Agents” — Can Traditional Automation Still

Gartner has ranked Agentic AI as the top strategic technology trend for 2025, forecasting that by 2028, 15% of everyday business decisions will be made autonomously by Agentic AI, a leap from virtually zero in 2024.

Agentic AI is not your traditional, reactive, command-driven AI assistant. Instead, it represents a new class of AI agents that are goal-driven, autonomous, capable of learning over time, and adept at executing multi-step tasks. These AI agents don't just follow orders—they can interpret goals, plan execution paths, call tools, manage resources, and even collaborate with other agents in complex environments.

To put it simply: Agentic AI is the kind of AI that gets the job done—on its own.

In contrast, generative AI (what we’re most familiar with today) behaves more like a helpful consultant: great at answering questions and drafting content, but limited to a single back-and-forth. The task ends when the chat ends.

Agentic AI, however, is different. Once it receives a task or goal, it will gather information, execute actions, make decisions, and control devices as needed—until the job is completed.

The year 2025 marks what many are calling the “pre-explosion phase” of Agentic AI. This transition didn’t happen overnight. The AI agent journey evolved from Conversational AI (2020–2023) to Workflow Agents (2024), and now in 2025, we are entering the era of General-Purpose Agents.

According to market research, 2025–2027 will be the key adoption window for Agentic AI in standardized applications such as programming, customer service, and data analytics. The global AI agent market is projected to surpass $300 billion.

This surge is driven by a confluence of technological progress and rising enterprise demand:

• Large Language Models (LLMs) like GPT-4, Claude 4, and Gemini have reached a level of cognitive sophistication necessary for complex reasoning and task decomposition.

• Tool use and function calling capabilities allow AI to interact with databases, APIs, and real-world control systems—essentially allowing the AI to act instead of just talk.

• Advances in reinforcement learning and memory systems enable agents to improve through experience and make increasingly optimal decisions over time.

2. How Is Agentic AI Reshaping the Industrial Landscape?

In the industrial automation sector, Agentic AI is quietly launching a revolution. With autonomous decision-making, task decomposition, and environmental interaction capabilities, AI agents are elevating traditional industrial control systems from passive execution tools to collaborative partners.

According to Gartner, by 2025, 25% of companies that use generative AI will deploy AI agents—reshaping the fundamental logic of industrial software, control systems, and operations.

In industrial scenarios, the application of Agentic AI has significant practical and strategic value. These environments are typically complex, time-sensitive, and safety-critical. Traditional automation systems rely heavily on preprogrammed logic and manual intervention, making them inadequate for dynamic optimization or unexpected disruptions. Agentic AI is now changing that.

Schneider Electric has partnered with Microsoft to launch an industrial Copilot system that integrates Microsoft’s Azure AI Foundry with Schneider’s advanced and secure automation systems to enhance productivity and workforce efficiency.

Another industrial giant, Rockwell Automation, has teamed up with Microsoft to develop AI agent solutions tailored to manufacturing via Azure OpenAI services.

3. Three Fast-Track Pathways for Deploying Industrial Agentic AI

Despite its momentum, deploying Agentic AI in industry still faces a pressing challenge: How can enterprises extract value from AI without overhauling existing systems or requiring massive technical teams?

Based on real-world deployments from leading companies, three practical and efficient strategies have emerged to lower technical barriers and implementation risks—offering fast, scalable pathways for turning AI agents from concept to production value.

Path 1: Retrofitting Intelligence into Legacy Production Lines

This is the most accessible and cost-effective approach. Without modifying existing equipment, enterprises can install a lightweight AI agent module to collect real-time data, analyze operational status, identify potential risks, and push alerts or recommendations—sometimes even triggering automated responses.

Path 2: Embedding AI Agents in Back-Office Cloud Systems

Many companies now use cloud-based ERP, WMS, or TMS platforms, which often come with rich API support. Using services like Microsoft Dynamics 365 or AWS Bedrock Agents, it’s entirely feasible to build a “Procurement Decision Agent”—one that automatically checks inventory, compares supplier quotes, considers lead times and production schedules, and then recommends or even places orders autonomously. These solutions require no factory floor hardware changes, and deployment is fast with quick ROI.

Path 3: Robots as the Physical Embodiment of AI Agents

This approach is the most visually impactful. With the rapid progress of AMRs (Autonomous Mobile Robots) and humanoid robots, developers can now use simulators like NVIDIA’s Isaac Sim to train AI agents in virtual environments—teaching them to grasp, move, or avoid obstacles—and then deploy the trained agents on real-world machines. This "robot-as-agent" model is tangible, visible, and easy to evaluate.

At the 2025 China Chain Expo, NVIDIA CEO Jensen Huang emphasized that the next wave of AI is robotic systems—machines that can reason, execute, and understand the physical world. In the next decade, factories will be driven by software and AI, with coordinated robot teams producing intelligent products under AI direction.

A Final Word: Agentic AI Is Not the Next Software—It’s the Next Worker

Beneath the excitement, it’s important to stay realistic. The large-scale deployment of Agentic AI still faces critical challenges:

Security and control: Granting AI increasing levels of autonomy introduces risks. Misjudgments or malicious use could result in serious incidents. Strict permissions, audit trails, and override mechanisms are essential.

Integration complexity: Industrial environments are full of heterogeneous legacy systems. Ensuring AI agents can seamlessly integrate with diverse protocols and data formats is a major technical hurdle. 

But one thing is clear: In tomorrow’s smart factories, Agentic AI won’t just be the next generation of software—it will be the next generation of workers.

Humans will no longer operate machines directly. Instead, they will orchestrate fleets of intelligent AI agents capable of thinking, collaborating, and optimizing in closed feedback loops.

The next chapter of the industrial revolution will be written—not by operators at control panels—but by autonomous, decision-making agents.

And today, whoever issues them their ID badge first will reap the rewards of tomorrow’s production capacity.

Tags: Siemens, automation parts