Why AI agents aren’t quite living up to the hype in manufacturing

Key Takeaways

• Agentic AI, despite its promise, is facing challenges in delivering accurate results in industrial environments
• The main issue lies in the structural approach of AI systems, which are designed to generalize across domains, but require specificity in industrial settings
• Current AI tools achieve only 40-60% accuracy on complex technical content, which is insufficient for critical tasks
• The architectural problem of AI systems processing information is a deeper failure point, requiring a more nuanced approach

Introduction to Agentic AI

Agentic AI, a type of artificial intelligence designed to perform tasks autonomously, has been gaining traction in recent years. However, its application in industrial environments, such as manufacturing, has been met with limitations. The promise of agentic AI is to free up skilled technicians from mundane tasks, allowing them to focus on complex repairs and maintenance.

The Gap Between Promise and Performance

Manufacturers, OEMs, and service organizations are eager to leverage agentic AI to automate tasks such as opening and managing tickets, documenting field service work, and identifying parts. However, the reality is that current AI systems are struggling to deliver accurate results. According to Cooper Schorr, senior director of North American sales at Octonomy AI, "Most AI systems are built to generalize across domains... and in industrial work, the opposite needs to be true."

Comparison of Agentic AI Approaches

The Architectural Problem

The deeper failure point of agentic AI lies in its architectural approach. Most platforms convert complex technical content into a format that can be processed by AI systems, but this approach is flawed. The conversion process leads to a loss of context and specificity, resulting in inaccurate results. Schorr notes, "These tools typically cap at about 40 to 60 percent accuracy at best on complex technical content... That's not fine for dispatching the right part to a field site or walking a technician through a repair on a six-figure asset."

Conclusion

The application of agentic AI in industrial environments is hindered by its structural approach and architectural limitations. While the promise of autonomous workflows and self-solving systems is enticing, the reality is that current AI tools are not yet capable of delivering accurate results. To overcome these limitations, a more nuanced approach is required, one that specializes in specific domains and prioritizes accuracy over generalization.

Bottom Line

The hype surrounding agentic AI in manufacturing has not yet been met with tangible results. While the potential for autonomous workflows and self-solving systems is significant, the current state of AI technology is not yet capable of delivering accurate results in complex industrial environments. As the industry continues to evolve, it is essential to address the structural and architectural limitations of agentic AI to unlock its true potential.