Agentic AI in manufacturing: A step towards autonomous production
AI is a game-changer, and it is an inevitable element when it comes to driving transformative change. The emergence of ChatGPT at the end of 2022 has resulted in an AI hype that is still prevalent and manifests itself in an increasing number of AI-related use cases and implementations.
Now, AI encompasses multiple technologies — from machine learning (ML) to visual inspection — and the development of new types with ever-increasing capabilities is constant. One of these new technologies that is promising to enlarge the capabilities of AI is agentic AI.
Agentic AI refers to artificial intelligence systems that can act autonomously, make decisions and take actions in complex environments without any human intervention.
Unlike generative AI (Gen AI), which requires human interaction, agentic AI works more freely. Also, the impact of Gen AI depends on how well users engage with the AI software. Instead, agentic AI does not require the user to constantly drive the system, but ensures the system works for the user.
Deciphering agentic AI
Agentic AI relies on large action models (LAM). These models are designed to handle complex decision-making tasks across various domains, particularly in environments where multiple actions can be taken based on a variety of inputs and conditions. Leveraging extensive datasets and sophisticated algorithms, the models generate intelligent and contextually appropriate actions.
Furthermore, agentic AI capabilities are driven by reinforcement learning with human feedback (RLHF). This allows a model to learn from outcomes and to continually improve itself based on actual performance. This is a sharp contrast to traditional AI models that learn from static data sets.
AI agents are built on several components:
- User input or human instructions
- The physical or digital environment in which they operate
- Sensors through which they perceive their environment
- A control center involving complex algorithms and models
- The data inputs they receive about their environment
- Effectors i.e. the tools they use to act
- Actions, which represent the alterations made by the effectors
The control center is essential as it manages the flow of information between the user inputs, decision making and planning, access to tools and the system’s effectors, enabling the action in the environments.
A step towards the autonomous factory
Based on its characteristics and features, agentic AI looks promising in industry sectors such as manufacturing. Deloitte predicts that by 2027, 50% of the companies using Gen AI will have launched agentic AI pilots or proofs of concepts being capable of acting as smart assistants performing complex tasks with minimal human intervention.
Let’s distinguish between two different types of AI agents with varying impacts on manufacturing processes:
- Virtual AI agents are software programs that use AI to perform tasks, make decisions, and interact with users or systems in a virtual environment. Unlike physical robots, virtual AI agents exist in software form and can operate within various digital platforms and applications. They can mimic human behavior, understand natural language and execute tasks autonomously or semi-autonomously. Examples of their use on the shop floor include inventory optimization, where they can monitor inventory levels, predict stock shortages, and recommend replenishment strategies based on historical consumption patterns. Another example is the optimization of production schedules, where they can analyze demand forecasts, machine availability, and workforce capacity, ensuring efficient resource allocation.
- Embodied AI agents are artificial intelligence systems that possess a physical presence or form, such as robots or autonomous machines. Unlike virtual AI agents that operate solely in software environments, embodied AI agents can interact with the physical world, performing tasks and making decisions based on sensory input and environmental context. They often use a combination of robotics, computer vision and AI algorithms to function effectively. An example of their application on a shop floor is assembly line robotics where they can operate on assembly lines, performing tasks such as welding, painting, and assembling components with high precision and speed.
AI agents will enable manufacturing companies to get closer to their target, i.e., to realize autonomous factories with limited (or absolutely no) human interaction in operations. And this may result in a significant increase in productivity, too! Shop floor workers will transition from manual operators to supervisors, decision-makers and strategists.
Nevertheless, the companies need to balance their immediate operational needs with their long-term strategy and planning while successfully managing the transition.
The practical approach when embracing agentic AI
While agentic AI-based applications hold significant promise for transforming manufacturing processes, organizations must navigate a variety of challenges during implementation.
By proactively addressing issues related to integration, security, governance, data quality, employee engagement and compliance, manufacturers can enhance the likelihood of successful adoption and fully realize the benefits of agentic AI technologies.
Engaging stakeholders across the organization and investing in training and secure infrastructure will be critical to overcoming these challenges and achieving operational excellence.
>> Want to learn more about agentic AI and how it is changing the manufacturing industry? Let’s connect.
>> Find out how we're changing our clients’ businesses with intelligent automation and agentic AI.
Posted: 29/07/25
