What makes a manufacturing process “AI ready”?

[programmablemanufacturing]

Open discussion

One recurring observation in manufacturing AI is that many projects fail before the model itself becomes the primary bottleneck.

In practice, challenges often include:

• unclear operational objectives
• unstable processes
• inconsistent measurement pipelines
• weak feedback loops
• poorly defined decision points
• limited controllability
• sparse or expensive data

This raises a broader question:

What actually makes a manufacturing process suitable for AI-assisted prediction, optimization, or control?

Potential dimensions may include:

• process stability
• measurability
• controllability
• cost of experimentation
• feedback frequency
• operational decision relevance
• uncertainty tolerance
• availability of physics priors

Interested in perspectives from:

• manufacturing engineers
• robotics researchers
• industrial AI practitioners
• materials scientists
• process engineers
• automation teams

Especially curious about:

• examples of successful industrial AI deployment
• common failure modes
• where physics-informed approaches help most
• what “AI readiness” should mean in real manufacturing environments