Artificial Intelligence in product development: hype, tool or lever?

Last December I attended a meeting of the professional advisory committee for Industrial Product Design in Groningen. We discussed current developments in our field, with one clear main theme: Artificial Intelligence (AI). Not as an abstract future vision, but as a concrete question: what does AI really mean for productontwikkeling?

By 2025 generative AI has definitively entered the daily work of many professionals. Chatbots, image generation and code assistance have in a short time shifted from curiosities to everyday tools. In sectors such as education, software development, marketing and consultancy the impact is obvious. But when the question was raised about what this means for physical product development, nobody had an immediate concrete answer.

At PEZY we were grappling with the same questions at that time. Which AI tools genuinely add value to our work? How do we safeguard quality and confidentiality of information? And perhaps the most important question: where does the human hand remain essential?

I have been following AI development for years and have experimented with it at various stages myself. From early tools like DeepArt and Artbreeder to Stable Diffusion on my own laptop. At the same time I have remained critical. Where does AI truly add value? Where is it mainly hype? And where does the risk arise that speed is mistaken for quality?

That question became extra tangible for me when one of my interlocutors quoted a LinkedIn post by Guido Stompff, lector in Design Thinking at Inholland University of Applied Sciences. He described how his annual “coffee filter challenge” failed for the first time this year. Where students had produced a wide variety of ideas for years, almost every group now ended up with the same idea. The reason? Chatbots. What was once a creative exercise with fifteen minutes of exuberant interaction became a three-minute exchange between students and their phones. In silence.

It’s a striking example of AI’s paradox: more speed, but less variety, less experience and sometimes less true innovation.

Productivity, with caveats

Recent research shows a similar picture. AI can deliver productivity gains, certainly in software development. Whereas earlier studies pointed to slowdowns caused by AI tools, we now also see clear accelerations. Tools like GitHub Copilot, Claude Code and Codex are already indispensable for many developers.

But those gains do not come without questions. Quality, maintainability, scalability and security are all aspects AI still struggles with. At organisational level the picture is also uneven: many AI initiatives do not yet deliver measurable impact while costs rise.

AI has also become established in marketing and design. Major brands are experimenting extensively with AI-generated campaigns. At the same time, resistance is growing. Debates about copyright, authenticity and intentionality are more relevant than ever. Particularly in more artistic domains the pushback is significant.

Why product development is different

You might expect product development to follow a similar trajectory. After all, we sit at the intersection of engineering, design and business. Yet there is an important difference.

AI is only as good as the data it is trained on. That is exactly the crux in product development.

Where software code and imagery are abundantly available, that is not the case for physical products. Companies carefully protect their designs. CAD data is seldom public. And even if you had that data, it remains questionable whether it would be readily usable.

Products are complex. They combine multiple functions, integrated solutions and countless trade-offs between cost, manufacturability, user experience, reliability and aesthetics. Genuine inventiveness is particularly hard to train for, because by definition there is no dataset of solutions that haven’t yet been conceived.

There are now impressive tools that can generate 3D models. But the gap between a visually convincing model and a manufacturable, reliable assembly is large.

Within CAD software we also see AI-driven features emerging, such as generative design. These can be valuable, but for the time being they are mainly focused on creating individual components. Not on designing complete, integrated products that must function in practice.

Where AI does work

Does that mean AI has no role in product development? Absolutely not. I now use it almost daily.

But mainly for the “low-hanging fruit”. Translations, text editing, summaries. And perhaps the most valuable: searching. With chatbots you can search far more purposefully than with traditional search engines. They help, for example, in finding suppliers, researching physical principles or data analysis methods, or mapping competing products.

The gain here is in speed, without necessarily sacrificing quality — provided you remain critical, probe further and verify sources.

We also see applications in the design area. Not by letting AI do the work, but as support in preparatory tasks. For moodboards, reference material, initial directions for materialisation, colour and finish, and visualisations. The role of the designer thus partially shifts from executor to director.

Perhaps the most interesting application is exploring product propositions. AI can analyse large volumes of reviews and recognise patterns. What are users satisfied with? Where are frustrations? Which expectations recur? It can help in formulating USPs relative to the existing market, structuring ideas and visualising concepts at a level suitable to excite potential customers or stakeholders.

Over the past year three clients approached us who refined and documented their product idea thanks to AI tools. Not as an endpoint, but as a stepping stone. It gave them enough confidence to take the next step.

But that is also where a risk lies: AI can create the illusion that you are further along than you actually are.

The indispensable human factor

Because between a good idea and a successful product lies a world of work.

Translating a concept into a functional, manufacturable, reliable and appealing product requires more than data and algorithms. It requires judgments. Context. Conversations with stakeholders. Resolving conflicting requirements. Experience with materials, manufacturing processes, tolerances, use situations and behaviour in practice.

And ultimately it requires testing in the real world. Does the product actually work? Is it user-friendly? Does it feel right? Does it hold up? Those are questions you don’t answer with a prompt, but with physical prototypes, user feedback and iteration.

The human hand therefore remains essential.

Conclusion

AI is changing our profession, but for now less radically than in software or graphic design. It does not take over our work, but it does cause shifts. Just as CAD software, Photoshop and 3D printers did before.

We see a partial shift from creating to curating, and from making to judging. The real value at the moment lies mainly in supporting tasks: gaining insight faster, exploring possibilities, conducting source research, sharpening propositions and improving communication.

But the core of product development remains, for the time being, human work: making things that not only sound logical or look good, but reliably function in the real world.