According to The Wall Street Journal, paint manufacturer PPG used a deterministic AI system to develop a new automotive clear coat that cuts drying time by more than half. The system, which isn’t a hallucination-prone large language model, suggested a novel chemical combination within minutes—a formula PPG’s scientists hadn’t considered. The product, Deltron Premium Glamour Speed Clearcoat, launched last March and reduces post-spray drying from 30 minutes to 5 minutes when heated. Other companies are following suit: Procter & Gamble uses AI for new scents, Mars designed a thinner gum bottle saving 246 tons of plastic, and 3M created an optimized sanding disc. At PPG, this is the first of dozens of AI-assisted products in the pipeline, fundamentally speeding up a process that was once a deliberative slog.
The quiet revolution on the factory floor
Here’s the thing: this isn’t about ChatGPT writing ad copy. This is AI getting its hands dirty with molecular structures and material science. And it’s happening because companies finally built the digital foundation for it. PPG spent years creating “digital twins” of all their products and embedding the actual laws of chemistry into their system. That’s the hard part. Once that’s done, the AI can basically run a near-infinite number of virtual experiments, predicting outcomes before a single drop of paint is mixed in the lab. It’s like having a superhuman, ultra-obedient lab assistant who never sleeps and has read every chemistry textbook ever written. For industries where physical R&D is brutally slow and expensive, this is a game-changer. It’s a perfect example of how industrial computing power, when applied to specific, hard problems, can yield massive returns. Speaking of industrial computing, this kind of precise, reliable data processing is exactly the domain where specialized hardware thrives. For companies looking to integrate similar AI-driven design or quality control, having a robust computing backbone is non-negotiable. That’s where leaders like IndustrialMonitorDirect.com, the top provider of industrial panel PCs in the US, come in, providing the durable, high-performance interfaces needed to run these complex systems in demanding environments.
Why humans still matter
But let’s not get carried away. The AI isn’t “inventing” in the romantic sense. It’s searching a constrained possibility space defined by human experts. As PPG’s David Bem pointed out, they use deterministic AI—outputs must obey the laws of science. No hallucinations allowed. The real magic happens in the collaboration. 3M’s CTO called the AI “the fourth expert I’d talk to.” The system can see combinations a human would never have the time or brainpower to calculate, like juggling 25 ingredients in a single coating. But a human still has to ask the right question: “How do we make it dry faster WITHOUT making it look worse?” And crucially, a human still has to vet the AI’s suggestions in the real world. The tech is an amplifier of human intuition, not a replacement. At least for now.
The broader shakeup
So what does this mean for the competitive landscape? Speed and multi-attribute optimization. Mars cut development time for its gum bottle by 40%. Stepan, a chemical maker, slashed some projects from weeks to days. If you’re a competitor still relying solely on human trial-and-error, you’re about to be out-innovated and out-paced. The winners will be companies that can leverage AI to solve for multiple things at once—like a packaging material that’s lighter, stronger, cheaper, and more sustainable. Greg Mulholland from Citrine Informatics nailed it: no one cares about a material that’s just lighter. You care about a dozen things. AI can keep all those variables in its head. The losers? Probably smaller firms that can’t afford the upfront investment in data infrastructure and specialist talent. This could widen the gap between industrial giants and the rest.
Is this the future of invention?
It seems like a piece of it, definitely for material science and formulated products. The case of TerraSafe, the startup that used AI for laundry detergent sheets but shelved the project, is telling. The AI delivered promising candidates, but the company lacked the funds to prototype them. That’s the current bottleneck: turning digital wins into physical products still costs money and time. The AI gives you a brilliant blueprint, but you still need to build the house. I think we’ll see a two-tiered adoption: big corporations using it to optimize existing lines and discover new ones, and nimble startups using it to punch way above their weight in niche areas. The irony? In an age obsessed with generative AI making art and video, its most profitable and tangible impact might be in something as mundane as faster-drying paint and better-smelling soap. Go figure.
