Leaps and bounds in AI capabilities in recent months are making their mark up and down supply chains, from manufacturing to warehousing.

When it comes to product design, AI is bringing streamlined processes and even advancements in sustainability. But standing between product design and an AI automation takeover is a data problem.

Dirk Hartmann, head of technology innovation at Siemens Digital Industries Software, explained that product design requires an immense number of inputs, including materials, shapes, and the exact internal mechanisms. That means that algorithms need more data to learn from than if they were just asked to create, say, an image. And a lot of that data is protected intellectual property.

To illustrate the IP problem, Hartmann used the example of an image generator trained on pictures of cars made by a single company. It would produce limited results and likely fail to drive innovation, he told Tech Brew.

“Obviously, the big vision is there, that you can use the AI to automate what the designers do,” Hartmann said. “But under the constraint that, on the one hand…[we’re working with] very limited data. And on the other hand, if you look at ChatGPT as an example, it starts to hallucinate at some point. That is not something you can afford in engineering.”

That’s not to say that Siemens isn’t building for AI-powered design.

Computer-aided design and simulations go back decades, Hartmann said. AI is making those processes more accessible by facilitating design loops, helping designers find optimal iterations, and providing guidance, he said.

“A few years ago, you really needed very, very deep experts…to figure out algorithms that tell you, OK, you need to tweak the diameter of that tube that way to increase the performance,” Hartmann explained. “With the tools we’ve created or started to create in the last few years, you can actually do that [with AI].”

Such tools can run virtual simulations and rapidly come up with design changes, so a designer only reviews the results instead of a long series of “manual, intensive steps,” Hartmann added.

Technology isn’t the only sector with eyes on AI for product design. Consumer goods giant Unilever is using AI as a tool to add speed, agility, and efficiency to different parts of the business, as well as provide insights that can lead to industry differentiators, Unilever’s head of R&D digital and partnerships, Alberto Prado, told Tech Brew.

Unilever—which is structured around five business groups: Beauty, personal care, home care, nutrition, and ice cream—is increasingly focused on hiring talent with modeling and analytics skills, in addition to an understanding of biology and chemistry, Prado explained.

While machine learning and AI are key priorities across Unilever’s business groups, investment and implementation vary between sectors, Prado said. Ice cream, for example, is a manufacturing-heavy product, and scaling production is a key focus for AI investment in that field. Unilever’s beauty group, he noted, is more focused on the science behind products like skin creams, and AI’s role there is primarily on the research and development side.

But even in the world of skin care, data constraints can pose a challenge to would-be AI-enabled chemists.

“AI is only useful if you have the right data,” Prado said. Unilever’s ability to extract value from AI comes down to “not very sexy” investments over the last 10 to 15 years in things like product lifecycle management systems and experimental data-capture systems, he added.

“We’ve been telling people that data is important…they’re more disciplined about capturing and labeling data, and now we have enough data assets that you can apply artificial intelligence and do great things with it.”

AI for the planet

Both Hartmann and Prado pointed to sustainability efforts as a key arena in which AI is impacting the design of new products.

Molecular models and AI have enabled Unilever to find alternative ingredients for those that have been impacted by things like global warming or geopolitical tensions, Prado said. That agility leads to a more resilient supply chain, and has helped the company find substitute ingredients (like a vegan alternative to a beetle-fueled pigment for lipsticks).

Simulations using AI have also let Unilever test biological reactions (like those caused by creams or beauty products) virtually, without animal testing, Prado added.

AI in product design isn’t just about creating new products, Hartmann said.

“We will not solve the sustainability challenge by just replacing all existing cars, all existing products, with new products; that would probably make the thing worse,” he said. Instead, AI can be put to use figuring out how to operate existing products with greater energy efficiency.

“If we look ahead at the challenges we are facing in terms of sustainability…it really requires us to innovate at a speed and also a magnitude we haven’t seen before.”