Design, Data and Disruption: AI in Additive Manufacturing
- Dec 16, 2025
- 4 min read
Artificial intelligence is reshaping every sector it touches, and additive manufacturing (AM) is no exception. A recent expert panel explored how AI is accelerating innovation across the AM value chain, from design and optimisation to production and workforce readiness.
As with many industries, the AM community is attempting to make sense of a future where artificial intelligence-driven change seems a given.
The impact of the current set of generative AI tools is still being figured out. They could be understood as representing incremental, transformational or perhaps even existential innovations.
Hervé Harvard, Executive Director and the founder of UTS Rapido, described the December 2022 release of ChatGPT as a watershed moment.
“I think for the world, this was the first time they saw AI,” said Harvard, whose career includes digital innovation leadership roles at Canon, Siemens and Philips.
“And I would argue it’s the first time they could interact with a machine in plain English.”
For Harvard, generative AI and agentic AI represent a step change far beyond traditional algorithms and Machine Learning techniques.
Breaking the mould
H3D, which was established in 2018 based on Swinburne University PhD research, is concerned with the bottleneck of computer-aided design (CAD) for mass-customised products.
The company completed a $5.8 million Series A raise in September. It has applied AI to vastly speed up workflows for 3D printed custom audio products (including hearing aids and in-ear monitors) and has turned its attention to doing the same for dental clinics.
Executive Chair Anthony Shilton said the company’s cloud-based service processes over 100,000 designs each month, turning raw scans into processed files for customers, who turn these into 3D printed parts.
It improves on the speed and quality offered by the previous, manual design workflows which are frequently a bottleneck in production. With the capability to design hundreds of pieces in parallel the service allows manufacturers to keep their 3D printers fed with print jobs throughout the shift.
Another project involves a product to scan ears using a phone, with the data then sent to a cloud service to improve it using AI. The method is currently being trialled by “a small number” of companies.
“It can interpolate – we’ve got a lot of ears now, so we have been able to train up models that are somewhat predictive of at least the cartilaginous part of the ear based on the external structure,” explained Shilton.
“We can predict at least to a certain depth what the ear’s going to look like. And it’s sufficient to produce a lot of consumer-type devices. And we believe self-scanning will enable the market for custom consumer devices to grow.”
Maybe five years away
Professor of Additive Manufacturing at the University of Auckland, Olaf Diegel, likes to separate AI from computational design when he talks about CAD and its future.
A lot of his team’s work on automated product development is “more computational design rather than AI, but it uses AI quite heavily,” he explained.
The difference is that in computational design, an engineer determines all the rules and the software builds the CAD.
“Nothing is left to chance. Whereas with pure AI, the AI makes the decisions for you, and often gets them wrong,” he explained, adding that he expects this to change over time.
At the moment, he believes text prompts are converted quite well into 2D models by generative AI, though this is not so for 3D models, despite an “endless number” of programs that can attempt this now.
“They’re sort of toy quality, they’re not engineering-quality parts,” he added of the results.
The best text-to-3D model programs can generate “a pretty-looking model” if you ask for, say, a helical gear with 36 teeth, though it’s not relevant in an engineering sense, lacking parameters like thickness, diameter and gear ratio. A part also needs to be considered in the context of a product it’s going into and how that part fits into a system.
“So trying to give the ChatGPT of the future enough information to do an engineering quality-controlled part that does the job first time out is actually a serious, difficult problem,” said Diegel.
“The language model, I guess, is part of what we need. So as I said, in my opinion: two, three, four, five years [away.]”
“We can do this”
Further than design work, AI in its many forms has been applied to different steps involved in the AM supply chain. One local example is in process monitoring for Laser Powder Bed Fusion machines by Melbourne company Additive Assurance.
Others are working on functions including parameter optimisation for print jobs (a real issue given the complexity of the task) or in material development. According to a 2024 MIT study, AI-assisted scientists discovered 44 per cent more materials versus those working without AI assistance.
Michelle Circelli, Research + Insights Lead, Future Skills Organisation (FSO), believes that companies’ successful adoption of AI hinges as much on leadership and culture as it does on technology. FSO is one of ten Jobs and Skills Councils (JSC) funded by the Australian Government, with a focus on the finance, technology and business sectors.
Their research found 35 per cent of such companies in Australia are not using any AI tools whatsoever, for reasons including a lack of trust.
Leaders will play a pivotal role “in ensuring AI becomes a positive transformation rather than a disruption”, said Circelli.
They must be explicit about the problem and why AI is being used, as well as invest in training and change management, foster a supportive environment with the ability “to completely change the way we’re doing work” rather than just a willingness to shoehorn new technology in.
“Essentially, training for broad AI literacy is key. And it’s very much like basic computer literacy was the focus in the past. And I’d like to hazard a guess: many are a similar generation to myself and this is sort of our second major digital disruption to our workplace,” said Circelli.
“We’ve lived through that. We can do this, okay? And this is even more exciting. I don’t know if it’s existential, but it is pretty exciting.”
