By Sursha Wu
As generative AI (Gen AI) matures, the spotlight is now shifting to its next evolution—Agentic AI (AI Agents).
This article distills insights from leading global AI experts, McKinsey’s 2025 industry research, and PowerArena’s hands-on experience deploying AI Agents on manufacturing floors to help manufacturers understand the near-future potential and practical roadmap of AI Agents.
This Article Covers:
- Why are AI Agents seen as the next key evolution after Gen AI? What market forces will drive adoption in 2025–2026?
- How can AI Agents be applied on factory floors to improve decision-making and management efficiency?
- What steps can manufacturers take today to prepare for AI Agent adoption?
Who Should Read This:
- Manufacturing leaders exploring AI Agent deployment
- Plant managers and digital transformation heads looking to boost efficiency and decision quality
- Strategy planners aiming to maximize ROI from smart manufacturing
What is AI Agent?
We are moving from passive AI, which only responds, toward agentic AI, which can plan, reason, and take actions to achieve goals. —— Andrew Ng, BUILD 2024 Keynote: The Rise of AI Agents and Agentic Reasoning, YouTube, Nov. 2024.
Dr. Andrew explains that traditional LLMs rely on zero-shot prompting—a simple process where a person gives a prompt and the model generates a quick response.
This interaction works like a one-off Q&A, aimed at meeting an immediate need.
By contrast, an AI Agent is built to solve problems, not just answer questions. It brings higher-level abilities in planning, reasoning, and action:
- Proactivity and Multi-step Planning: An AI Agent can break down a complex task into multiple steps and continuously adjust its strategy to reach the end goal.
- Interaction and Tool Use: It can autonomously call external APIs, databases, or software, and refine its process based on execution results.
- Continuous Feedback: Unlike one-off answers, an AI Agent can learn from feedback during the process, making corrections and optimizations along the way.
AI Agents for manufacturing in 2025-2026
1. From Passive Response to Autonomous Decision-Making
Like any truly disruptive technology, AI agents have the power to reshuffle the deck. Done right, they offer laggards a leapfrog opportunity to rewire their competitiveness. Done wrong—or not at all—they risk accelerating the decline of today’s market leaders.
— McKinsey & Company, Seizing the Agentic AI Advantage, June 13, 2025
The autonomy of AI Agents is the key to moving beyond the limits of traditional AI tools. Instead of just assisting human decisions, AI Agents can actively perform tasks and act as “digital employees,” driving deeper automation and smarter decision-making.
Unlike earlier systems built for narrow, task-specific use, AI Agents can process and combine text, data, and images as inputs for decisions, and directly interact with enterprise systems and workflows.
This cross-system, multi-task ability allows them to be applied flexibly in complex areas like manufacturing, supply chains, and finance, helping businesses respond to market changes and risks with greater speed and agility.
2. Unlocking Trillions in Economic Value: $2.6 trillion to $4.4 trillion
McKinsey estimates that the technology (Gen AI) has the potential to unlock $2.6 trillion to $4.4 trillion in additional value on top of the value potential of traditional analytical AI….However, this enthusiasm has yet to translate into tangible economic results. More than 80 percent of companies still report no material contribution to earnings from their gen AI initiatives.
— McKinsey & Company, Seizing the Agentic AI Advantage, June 13, 2025
McKinsey notes that although generative AI was expected to deliver significant benefits since 2022, many companies failed to see immediate results despite heavy investment. Now, with the rise of AI Agents in 2025, those earlier investments in generative AI are beginning to show large-scale value. The core lies in reinventing processes. Take customer service as an example:
- Gen AI-enabled: Provides assistive tools only, with humans still performing most of the workflow. Average handling time improves by just 5–10%.
- Agent-enabled (optimized): AI Agents start automating parts of the process—such as ticket classification or root cause analysis. Average resolution time improves by 20–40%.
- Agent-enabled (reinvented): The process is redesigned entirely around the Agent’s autonomy. AI Agents proactively detect and diagnose issues, cutting average handling time by 60–90%, while up to 80% of Level-1 problems are resolved automatically.
Compared with Gen AI, AI Agents can save more than 10 times the work hours, dramatically accelerating how quickly businesses respond to problems.
Reference: McKinsey & Company. Seizing the Agentic AI Advantage. Report. June 13, 2025.
3. Multi-Agent Collaboration and Governance
Fewer than 10 percent of use cases deployed ever make it past the pilot stage.
—— McKinsey’s ecosystem of strategic alliances brings the power of generative AI to clients,” April 2, 2024.
During the generative AI boom, many companies focused on broad applications such as chatbots or virtual assistants. These projects were easy to launch but often failed to deliver lasting value. In contrast, the use cases with the greatest impact—those that directly improve quality and efficiency—proved much harder to scale. Many stalled at the pilot stage due to challenges in organization, technology, data, or company culture.
To overcome this, forward-looking businesses are now prioritizing the basics. By strengthening data quality, improving system integration, and establishing clear governance, they ensure that AI Agents remain aligned with business goals and deliver results that can be measured and sustained.
Organizations will also need to set up the foundation to effectively operate in the agentic era. They will need to upskill the workforce, adapt the technology infrastructure, accelerate data productization, and deploy agent-specific governance mechanisms.
— McKinsey & Company, Seizing the Agentic AI Advantage, June 13, 2025
How do AI agents in manufacturing work?
Cross-Line Data and Information Sharing
In traditional factories, information/data silos between production lines and sites are a major challenge for efficient management. AI Agents can integrate data from MES, ERP, IoT devices, and vision systems, analyze it in context, and provide actionable recommendations—or even take direct action.
This unlocks two major advantages:
First, management models can be replicated quickly. If production processes are similar within the same plant, AI Agents can absorb the necessary process knowledge and decision logic, then replicate or adapt it for other lines. This shortens deployment time and reduces learning costs.
Second, best practices from a “golden line” can be rapidly transferred to new factories. With AI Agents, managers can replicate management systems, reallocate resources, and distribute capacity across sites in a short period. Key functions like operations, logistics, scheduling, and warehousing can be integrated horizontally, boosting supply chain efficiency, reducing waste, and speeding up time-to-market.
Multimodal AI Integration
Compal x PowerArena: Applying AI Vision + LLM in Factory Management
Factory data comes from many sources—production reports, IoT machine data, signal lights, sound sensors, gas monitors, and more. The volume is enormous and the structure highly complex. Within a multimodal framework, AI Agents can collect and interpret signals across systems to provide a unified, holistic analysis of factory status.
For example, if an AI vision system detects abnormal manual operations, MES records process errors, and IoT data shows irregular fluctuations in temperature or pressure, AI Agent can cross-reference these signals. Before the issue escalates, it can trigger alerts, initiate corrective actions, or even stop a line or machine automatically to prevent defective products from moving forward and to safeguard safety.
Proactive Quality Management
AI Agents shift quality control from after-the-fact inspection to real-time prevention. By combining vision data, IoT signals, and MES information, they continuously monitor workflows. When SOP deviations, fastening errors, or welding issues occur, the system can immediately issue alerts and prevent defects from passing to the next stage.
More than just a warning system, an AI Agent acts like a virtual industrial engineer. It can quickly identify root causes, cross-check data, and automatically generate improvement reports such as CLCA or 8D. Leveraging past improvement records, it delivers practical recommendations to help managers make faster, better-informed decisions. In some cases, AI Agents can even take direct action, such as adjusting machine parameters or optimizing production schedules, to resolve issues quickly and prevent recurrence.
AI Agents software and application for factory
As global supply chains are being restructured, many manufacturers are shifting to lower-tariff, lower-labor-cost regions such as Vietnam and Thailand to reduce reliance on a single hub. But once production begins, they face new challenges: high worker turnover, a shortage of industrial engineers, and productivity levels that lag behind their parent plants in China. Replicating proven management models through integrated systems has therefore become one of the industry’s greatest hurdles.
Read more: Replicating Production Management Models in New Sites with AI
Future trends in AI agents for manufacturing: 3 practical suggestions for manufacturers
1. Data Quantity and Quality Remain the Foundation
Accelerate data productization and address quality gaps in unstructured data. Finally, agents depend on the quality and accessibility of enterprise data. Organizations must transition from use-case-specific data pipelines to reusable data products and extend data governance to unstructured data.
— McKinsey & Company, Seizing the Agentic AI Advantage, June 13, 2025
While AI Agents hold the promise of autonomous problem-solving, their effectiveness depends on complete, high-quality data.
In manufacturing, success requires sufficient data support—production data, environmental parameters, workflow records, and even past management experience. Manufacturers should first examine whether their infrastructure can deliver consistent, high-quality data, and second, convert unstructured data such as images into structured data to build a solid foundation.
Only when digitization progresses smoothly and data governance ensures both quality and accessibility can AI Agent investments achieve maximum impact.
2. Security Risks and Data Privacy
In industries with high confidentiality and strict technical requirements, such as semiconductors and defense manufacturing, preventing data leaks is critical when deploying AI Agents.
All sensitive data and computation should remain on-premises, with strict rules and boundaries governing AI Agent operations. Enterprises also need clear accountability and compliance frameworks. This includes codes of conduct, trackable workflows, ongoing performance reviews, and robust risk management mechanisms.
3. Integration of IT and OT
Every AI application starts with a need. The key is understanding what problem it solves.
— Managing Director, AWS Taiwan & Hong Kong, Robert Wang
Integrating IT and OT is often the biggest barrier to AI adoption in smart factory transformations.
Before introducing AI, companies should plan architectures that support multi-agent collaboration and integrate deeply with existing systems. This enables cross-department and cross-process cooperation, overcoming the limits of isolated applications and improving overall efficiency.
Multi-agent systems must be capable of dynamic management and reuse, while also integrating asset governance. This means unified control of model settings, prompts, and tools, with strict version and permission management. Continuous feedback and standardized monitoring help optimize performance, while risk and compliance frameworks ensure safe, stable operations. Together, these elements allow AI to become a true driver of smart manufacturing that supports innovation and sustainable growth.
Learn more about how AI Agents can optimize smart factories and manufacturing processes.
Table of Contents
