Every second article about AI mentions ChatGPT these days. For manufacturing, the chatbot applications are real but limited. The more interesting question is: what else can generative AI do in industrial settings?

Turns out, quite a bit. Let me walk through applications that are already working in Australian manufacturing operations—nothing theoretical, just things I’ve seen in production.

Documentation and knowledge management

Manufacturing generates enormous amounts of documentation: maintenance procedures, work instructions, quality records, training materials, engineering specifications. Generative AI is surprisingly useful here.

Generating procedure drafts

When a maintenance tech figures out a fix for an unusual problem, that knowledge typically stays in their head or gets scribbled in a notebook. Generative AI can help turn verbal descriptions into structured procedures.

One client has technicians describe fixes verbally (voice recording). The recording gets transcribed and fed to an AI that generates a draft maintenance procedure in the company’s standard format. A supervisor reviews and approves. The procedure goes into the knowledge base.

This captures institutional knowledge that would otherwise be lost when people retire or move on.

Translating technical documentation

Australian manufacturers often work with equipment from Germany, Japan, or China. Technical documentation arrives in translation that ranges from awkward to incomprehensible.

Modern AI translation, especially for technical content, has improved dramatically. Several clients use it to make foreign-language manuals actually usable. Not perfect, but much better than paying for professional translation of every document.

Answering questions from documentation

Rather than searching through thousands of pages of manuals, operators can ask questions in natural language. “What’s the procedure for clearing a jam on the Model X conveyor?” The AI retrieves relevant documentation and summarises.

This isn’t about replacing human expertise—it’s about making documented knowledge accessible when the expert isn’t available at 2am.

Design and engineering applications

Generative design for parts

This predates ChatGPT but is worth mentioning. AI can generate structural designs optimised for specific load conditions and manufacturing constraints. Give it the forces a bracket needs to handle and the fabrication methods available; it produces designs humans wouldn’t think of.

An aerospace manufacturer in Victoria uses this for lightweighting components. The AI-generated designs look organic—almost biological—but they’re stronger and lighter than human-designed alternatives.

Automating CAD-to-manufacture preparation

Converting engineering designs to manufacturing programs (CAM) typically requires skilled programmers. AI tools are emerging that can do initial CAM programming automatically, with humans reviewing and refining.

This doesn’t eliminate the need for expertise—complex parts still need human judgment—but it speeds up routine work and reduces the skill barrier for simpler jobs.

Accelerating failure analysis

When something goes wrong, understanding why often requires digging through logs, sensor data, and historical records. AI can accelerate this by identifying relevant information and suggesting possible causes based on patterns.

A process engineer told me: “It used to take me half a day to pull together the data for a failure investigation. Now the AI does the initial gathering and correlation. I spend my time actually thinking about causes, not wrestling with spreadsheets.”

Quality and inspection

Defect description and classification

Computer vision systems can detect defects. Generative AI can describe what they found and classify them meaningfully.

Instead of just “defect detected at coordinates X,Y,” the system can generate “Scratch approximately 3mm long on the polished surface, likely caused by handling. Similar to defect type SC-3 in the quality database.”

This makes quality data more useful and searchable.

Automatic report generation

Quality reports are tedious to write but essential for compliance. AI can draft reports based on inspection data, statistical analysis, and standard templates. Human review catches errors; AI saves hours of assembly work.

One client generating dozens of certificate-of-conformance documents daily cut their documentation time by 60% with AI-assisted drafting.

Planning and scheduling

Natural language interfaces for planning systems

Production planning and scheduling systems are powerful but often have steep learning curves. Generative AI can provide natural language interfaces.

“Schedule an extra shift on Line 2 next Thursday” translates into the actual system commands. “What happens to our delivery dates if the Smith order doubles?” triggers a simulation and returns a plain-language summary.

This makes planning tools accessible to more people and speeds up scenario analysis.

Generating maintenance schedules

Given equipment condition data, production schedules, and maintenance requirements, AI can generate proposed maintenance schedules that balance all constraints. Humans review and approve, but the initial schedule generation happens automatically.

Communication and training

Multi-language communication

Many Australian manufacturing facilities have workforces that speak diverse languages. AI translation in real-time (or near real-time) enables communication that would otherwise require translators.

Safety briefings, work instructions, and team communications can be translated on the fly. Not perfect for legal documents, but practical for day-to-day operations.

Generating training content

Creating training materials is time-consuming. AI can help generate:

• Quiz questions from procedure documents
• Simplified explanations of complex processes
• Scenario-based exercises
• Refresher summaries

A training coordinator told me: “I used to spend weeks developing new training modules. Now I develop them in days. The AI drafts, I refine, and we get better materials faster.”

What doesn’t work (yet)

Let me be honest about the limitations.

Autonomous decision-making: Generative AI isn’t reliable enough for autonomous control of manufacturing processes. It can suggest and assist; it shouldn’t decide without human oversight.

Real-time process control: The latency and reliability requirements for process control don’t suit current generative AI. This is the domain of purpose-built control systems and specialised ML models.

Anything requiring guaranteed correctness: Generative AI makes mistakes. For applications where errors have serious consequences (safety-critical procedures, regulatory compliance), human review is essential.

Novel engineering problems: AI is excellent at synthesising existing knowledge. It’s less useful for genuinely novel engineering challenges that require creative problem-solving.

Getting started

If you want to experiment with generative AI in manufacturing:

1. Start with low-risk applications: Documentation, training materials, communication. The stakes are low if the AI makes mistakes.

2. Always include human review: Don’t deploy AI-generated content without human verification, especially for anything affecting safety or quality.

3. Use existing tools: You don’t need custom AI development for most of these applications. Commercial tools (ChatGPT Enterprise, Claude, Microsoft Copilot) work for many use cases with appropriate prompting.

4. Mind the confidentiality: Some AI tools send your data to external servers. For sensitive information, use enterprise versions with appropriate data handling agreements, or on-premises alternatives.

5. Measure value: Track time saved and quality improvements. Generative AI is cheap enough that modest improvements justify the cost, but you should still know whether it’s helping.

Generative AI won’t transform manufacturing overnight, but it’s a useful tool for the growing list of applications where it works well. The key is matching the technology to appropriate problems—and keeping humans in the loop.