The Prototype Problem: Rethinking Product Development in the Age of AI

80–85% of testing is now done in software, not hardware. What does that mean for the economics of product development? Why this matters to manufacturing CXOs today Product complexity is moving rapid

Why this matters to manufacturing CXOs today

Product complexity is moving rapidly into software, electronics and system integration. Yet, for many organisations, critical validation still happens only after physical hardware is available, when failures are more expensive, timelines are tighter and design changes are harder to make.

The traditional question was: How fast can we build a prototype?

That question is changing.

The shift

Digital twins, simulation and software-based testing are enabling engineering teams to learn, validate and optimise much earlier, before physical hardware exists.

On Stellantis's newest platforms, 80–85% of testing runs in software, allowing engineers to validate designs up to a year before hardware becomes available.

The prototype is no longer the main teacher. It is becoming the final check.

What changes on the economics

This is more than a reduction in prototype costs. It changes the economics of learning. Physical prototypes make every iteration expensive. Simulation shifts more iterations into a computing environment where the cost of each additional learning cycle can be dramatically lower.

Where deployed effectively, programmes report 50% faster time-to-market and up to 25% better product quality, while compressing testing and validation costs significantly.

But adoption remains uneven: 92% of engineering teams have surrogate models, yet only 8% rely on them extensively.

The question for every manufacturing CXO

How much of your next product could you prove before you build one and what is actually stopping you?

What the full article covers

The Prototype Problem examines how complexity, tighter product timelines and falling computing economics are reshaping product development. It explores examples from Stellantis, Marelli, Hindustan Unilever and Dr Reddy's, and identifies five conditions required to move from simulation pilots to scaled engineering transformation:

Governed data → Engineering-grade models → Scalable computing → Connected factory thread → Trusted rules

The bigger question is no longer simply whether an organisation should adopt simulation. It is how quickly the organisation can learn and what that means for R&D capital allocation.

Download