AI Can Accelerate Hardware, But It Cannot Skip Reality
Article
Featuring
Junus Khan
The current AI conversation often treats every industry as if it will be transformed in the same way. That misses an important distinction. Software and physical products do not operate on the same clock.
In software, a product can be released, measured, patched, and updated almost continuously. Mistakes can often be corrected after deployment. Physical products are different. A shoe, a material, a component, or a piece of equipment has to be built. It has to survive use. It has to perform under real conditions. The final proof still lives in the physical world.
That does not make AI less important for hardware, materials, or footwear. It means AI’s role is different.
In physical product development, AI is not a shortcut around reality. It is a way to get to a better version of reality faster.
At Carbitex, our work is fundamentally about creating desired physical effects. In footwear, that means understanding how a product should bend, resist, return, stabilize, or move with the body. The challenge is not simply drawing a part or choosing a material. The challenge is translating a complicated set of variables into a specification that can become a real product.
Those variables can include biomechanics, sport-specific movement, individual physiology, material behavior, manufacturing constraints, and the intended performance outcome. Traditional computational tools can help, but they do not always solve the harder problem: defining the right specification in the first place.
That is where AI has become valuable. It helps organize complexity. It helps generate better starting points. It helps reduce the distance between intent and specification.
The product still has to be made. It still has to be tested. It still has to be validated. AI does not remove those steps. It compresses the path to them.
That compression matters. In physical product development, speed is not just a convenience. It changes how many ideas can be explored, how quickly assumptions can be tested, and how much learning can happen within the same development window. When the starting point is better, the number of physical iterations can decrease. When the process moves faster, the team can do more with the same resources. When the output improves, the final product has a better chance of performing as intended.
"Ai helps organize complexity. It helps generate better starting points. It helps reduce the distance between intent and specification. The product still has to be made. It still has to be tested. It still has to be validated. AI does not remove those steps. It compresses the path to them."
This is also why AI works best when used by people who already understand the work without AI. A plausible answer is not the same as a useful answer. In physical products, an output eventually has to become something real, and that reality is unforgiving. Materials behave a certain way. Manufacturing has constraints. Bodies move in ways that are not always intuitive. Testing reveals what theory misses.
AI can help generate options, but expertise is still required to judge them.
The analogy I use is the nail gun. The invention of the nail gun did not make everyone a carpenter. It made skilled carpenters faster and more effective. AI is similar. It gives capable people more leverage. It increases bandwidth. It allows engineering talent to spend less time on repetitive translation work and more time on judgment, development, and future applications.
One of the most useful changes has been our ability to create focused AI-enabled tools around our own workflow. Many advanced engineering platforms already contain powerful capabilities, but they can be expensive, complex, and far broader than what a specific task requires. AI makes it possible to build lighter, more targeted tools that fit the way a company actually works.
For example, if we want to extract properties from an existing product, we can upload a 2D drawing, trace the relevant lines, select our material combinations, and complete in minutes a task that previously required a much more complex workflow in software like SOLIDWORKS. We still use complex. CAD software for the functions where it is the right tool. But AI allows us to build purpose-specific interfaces that remove unnecessary friction from the process.
That is a meaningful shift. AI is not only changing individual tasks. It is changing how work can be structured.
Over time, I think this will affect organizational design. Companies should not think of AI only as software that sits on top of existing workflows. The larger opportunity is to rethink where work should happen, who should do it, which decisions should be automated, and where human judgment matters most. AI can sit inside the workflow, not just beside it.
Fifteen years ago, I had a vision of a process flow where a customer come to our website, input the specifications we need, generate a product recipe automatically, and send that recipe into a highly automated manufacturing environment. Fifteen years ago, that kind of end-to-end system felt was a guiding principle – are our decisions getting closer to that outcome or further away. Now it feels within reach.
But even that future does not eliminate the realities of physical execution. Hardware companies still need infrastructure. They need manufacturing knowledge, supplier relationships, materials expertise, quality control, and the ability to produce at scale. AI will not erase those requirements.
"AI is not only changing individual tasks. It is changing how work can be structured. Over time, I think this will affect organizational design. Companies should not think of AI only as software that sits on top of existing workflows. The larger opportunity is to rethink where work should happen, who should do it, which decisions should be automated, and where human judgment matters most. AI can sit inside the workflow, not just beside it."
That is why I do not think AI makes it easy for entirely new hardware companies to appear overnight. Physical industries still have barriers that software does not. But among companies already competing in physical categories, the difference between using AI well and using it poorly will become significant.
The advantage will not always be immediate or dramatic. It will compound.
A company using AI well can learn faster. It can test more ideas. It can reduce wasted cycles. It can preserve expert judgment while removing unnecessary friction. A company that does not use AI may still be able to operate for a while, especially in industries with long timelines and deep inertia. But it will be operating at an older clock speed.
That is the real strategic risk. Not that AI instantly replaces physical product companies. Not that every competitor suddenly becomes dangerous because they have access to the same broad tools. The risk is that some companies will increase their rate of learning while others continue to move at the pace the industry has accepted for decades.
Footwear is a good example. The industry has tremendous inertia. Many products are still developed through long, familiar, sequential processes. That does not mean those processes are irrational. Physical product development is hard, and companies are cautious for a reason. But caution can become structural drag. Over time, the companies that learn faster will have an advantage.
For hardware, materials, footwear, and other physical industries, AI should not be understood as a replacement for building, testing, or proving. It should be understood as an accelerator for the parts of the process that happen before proof.
The physical world still gets the final vote.
AI just helps companies arrive there with better questions, better specifications, and better starting points.
