GN 70: AI - The Enemy that we Love

AI is no longer something the foundry industry can laugh off as a distant software trend. Large language models are already being used by engineers, buyers, designers, and decision-makers to explore manufacturing options, compare processes, prepare for meetings, and form first impressions of suppliers.

In this episode, Fabian Niklas and Staffan Zetterström look at AI from the perspective of foundry guys, not AI experts. The central concern is not that LLMs (Large Language Models) exist. The real problem is that LLMs often provide poor or outdated answers about casting because the foundry industry has not produced enough accessible, up-to-date, publicly available information for these systems to learn from.

That has direct commercial consequences. If an engineer, product designer, or purchasing team asks an LLM whether a part should be cast, the answer may be shaped by outdated textbook assumptions rather than by modern high-pressure die-casting practices. In the worst case, AI does not just misunderstand casting, it quietly sends potential customers somewhere else.

The episode is a grounded conversation about what AI can do, where it fails, and why foundries should start testing it from the perspective of their target customers.

Key Topics

* Why AI drives customers away from casting

When an engineer asks an LLM for manufacturing advice, the answer depends on the information available online. If current casting knowledge and marketing material are missing, the model may recommend against casting based on outdated assumptions.

* The foundry industry has a visibility problem

AI often fails to distinguish between older textbook descriptions and modern die-casting processes. The issue is not only technical accuracy but also the lack of publicly searchable material showing what today’s foundries can actually do.

* Try AI Seach from your customer’s perspectiveAsk the same questions your target customers would ask. Search for your process, your applications, your company, and your alternatives and then see whether AI leads people toward your capabilities or away from them.

* AI is useful, but only with the right inputLLMs depend heavily on the quality of the question, the available data, and the user’s ability to judge the answer. It can produce convincing text even when the technical conclusion is weak or incomplete.

* Marketing cannot be fully outsourced to AI