AI in Manufacturing (AIM) # 6 - AI in Engineering: From Agentic Systems to the Future of Product Development

The sixth AIM event was held last week, 19th of March, where Agentic AI Systems across different engineering fields were the main topics. Speakers from Microsoft, NVIDIA, Accenture, PTC and Consat shared their insights.

But what are Agentic AIs really? Agentic AI refers to artificial intelligence systems that possess a certain level of autonomy, enabling them to initiate actions, make decisions, and pursue goals independently, without direct human intervention. These systems are engineered to operate proactively, adapting to new information and evolving circumstances. By breaking down complex objectives into smaller, manageable tasks, Agentic AI determines how to approach each task and learns from the outcomes to improve future performance. Unlike traditional AI, which typically operates by following pre-defined instructions, Agentic AI can dynamically adjust its strategies, collaborate with other agents, and optimize processes across various domains.

In comparison to traditional AI, which requires explicit instructions from humans, Agentic AI can plan, making decisions, and acting autonomously. Rather than simply compiling solutions, Agentic AI systems break high-level goals into sub-tasks, select the appropriate tools or APIs to achieve those tasks, and iterate on their own decisions based on feedback from the environment. This autonomy allows Agentic AI to respond proactively and efficiently to changing conditions.

The presenters in the event stressed how Agentic AI is reshaping the way we work. In manufacturing, predictive maintenance agents monitor sensors to forecast failures, order parts, and adjust schedules. Agentic AI boosts productivity by assigning tasks, tracking progress, and resolving team bottlenecks.

Speakers shared real-world applications of Agentic AI. PTC now deploys multiple AI agents in their CAD systems for faster, improved design, including agents that explain software errors and offer solutions. Accenture discussed the benefits, risks, and security measures of Agentic AI, highlighting their Secure AI service, which helps organizations adopt these technologies safely by integrating security and trust from the start. Next was a workshop titled “How to augment human operators using Agentic AI.” Attendees were divided into teams to pinpoint operator tasks suitable for automation or improvement with Agentic AI, as well as to discuss the technologies needed for deployment in industrial environments.

The day ended with our master’s students presenting their thesis projects. The first team introduced a gesture-based control system for Industry 5.0, using cameras to recognise specific arm movements and operate robots. Their goal is to design a flexible blueprint that can be easily adapted for any factory setup. They primarily demonstrated controlling a heavy goods vehicle (HGV), showing how arm gestures could start and stop a conveyor belt in real time. The second group demonstrated their human proxy application in Emulate3D, which creates a digital twin of an actual person. In their live demo, they managed to trigger a safety stop when a digital human arm entered a restricted area near the conveyor belt.

Overall, AIM 6 was a fun and insightful event.