News People & Careers
Why AI makes human judgement more valuable in coatings
AI is transforming formulation, regulatory work and technical training in the coatings industry. Dr. Evripidis Tsaousoglou explains why faster access to answers does not replace expertise—and why practical judgement, scientific understanding and real-world experience will matter more than ever. An interview by editor Yeray López Arauco with Evripidis Tsaousoglou, Managing Director of the Institute of Coating Technologies.
Will AI replace coatings professionals—or change what expertise is worth?
Evripidis: I actually ask myself this a lot.
Because I manage the Institute of Coating Technologies but also remain highly active on the factory floor, I can’t afford to look at AI as just some abstract trend. I have to be completely realistic: AI can already do parts of my job faster, and sometimes better, than I can.
Tasks like digging through technical literature, comparing document revisions, screening regulatory updates, or throwing together the first draft of a report? AI accelerates all of that dramatically. You can feel the pressure this creates in the industry. A recent University of Melbourne and KPMG study looked at over 48,000 people globally. They found that two in five employees think AI will replace jobs in their field, and nearly half are essentially concerned of being left behind if they don’t adapt. But here is the catch: 56% admitted to making mistakes because of AI, and two-thirds are just blindly trusting AI outputs without properly evaluating them at least some of the time.
That combination tells you everything you need to know. We’re terrified of not using AI, but we aren’t really prepared to use it right.
For me, the real question is this: If information is no longer a scarce resource, what is?
In the coatings industry, the answer is judgement. A formulator isn’t valuable because they have a better memory for raw material data than a computer. They’re valuable because they know how to define the actual problem. They understand the chemistry, they can spot a weird lab result, they know what can actually be manufactured and applied in the real world, and—crucially—they take responsibility for the final decision. AI is going to change our jobs. Some tasks will vanish, others will speed up, and certain skills we valued twenty years ago just won’t matter as much. But honestly? It raises the bar for the human professional. My worry isn’t that AI will make expertise irrelevant. My worry is that professionals will stop developing their expertise because the AI is doing the heavy lifting, and that is what will make them irrelevant.
Is there a risk that AI creates a generation of professionals who can find answers quickly but understand the science less deeply?
Evripidis: Yes, absolutely. But not just because young professionals are using AI. The danger starts when people confuse having immediate access to an answer with actually understanding the topic.
A simple incident I recently witnessed in a laboratory made me think about this. We had an intern who had read up on the relevant quality-control standards but kept forgetting one specific detail: the exact temperature required for a viscosity measurement. After answering the same question a few times, the lab supervisor finally told him, “You need to find a way to remember this. Write it down, put it in your notes—whatever works for you.”
The student’s response was instant: “Why? I’ll just ask ChatGPT.”
My first instinct was exactly what you’d expect from someone of my generation: No, you need to know this stuff. But later, I thought about it more critically. Does a young chemist really need to memorise an isolated number that they can pull up in three seconds? Maybe not.
But here is what that student does absolutely need to understand: why temperature affects viscosity in the first place, which test method applies, why that specific standard exists, whether the AI pulled the correct and most recent revision, if the sample was conditioned properly, and whether the final number actually makes physical sense. And the biggest question of all: If ChatGPT hallucinates and gives the wrong temperature, does the student have the baseline knowledge to realize it’s wrong?
The number itself is just information. Knowing how to verify and interpret it—that’s competence.
The education system has to adapt to this. That same KPMG study noted that 83% of students are already using AI, but only about half get any real guidance on how to use it responsibly [1]. The challenge in education isn’t giving people access to information anymore. It’s stopping them from mistaking fluency for actual understanding.
The rarest skill in the next generation won’t be knowing the right answer. It will be having the intuition to spot the wrong one.
If professionals can learn directly from AI, what additional value can an organisation like the Institute of Coating Technologies provide?
Evripidis: Our value begins exactly where access to basic information ends.
Don’t get me wrong, AI is incredible at delivering information. It can break down a complex concept, translate industry jargon, generate test questions, and act as a 24/7 tutor. Educational organisations should be leveraging this, not trying to fight it.
But throwing a bunch of isolated prompts at an AI is not the same thing as going through a structured, coherent learning process.
A specialised institute has to figure out the roadmap: what foundational principles need to come first, how to bridge theory with practical application, which standards actually hold weight, and where the common industry pitfalls are.
At the Institute of Coating Technologies, our biggest edge is that we aren’t sitting in an ivory tower. The people running our programmes are out there every day dealing with real formulations, lab testing, production scale-ups, certifications, and consulting.
We connect the dots between what the scientific theory says, what the ISO standard demands, what the lab actually measures, what the factory can reliably produce batch-after-batch, and what actually happens when a contractor applies the paint in the field.
Think about it this way: You can ask ChatGPT why a coating blistered, and it will give you a highly convincing bulleted list of causes in two seconds. But a competent human trainer teaches you how to systematically investigate the failure. They show you what physical evidence to look for, which theories to eliminate first, and why the most “obvious” AI-generated answer is probably wrong in this specific context. We provide what an open AI interaction cannot provide on its own: expert-vetted context, structured progression, real-world case studies, feedback, assessment and accountability for what is being taught. This is a massive issue right now. The ChemSkills survey recently highlighted the availability of appropriately skilled people as a major challenge in the European chemical sector, while access to suitable training programmes remains an important barrier to closing the skills gap.
Authority doesn’t just come from an institution’s name, just like it doesn’t come from the confident tone of an AI output. It has to be earned through practical relevance and the willingness to constantly update what you’re teaching based on what’s actually happening in the market.
Where do you see the most immediate practical application of AI in the coatings and chemical industry?
Evripidis: If you want immediate, practical ROI right now? Regulatory intelligence.
Not because regulations are simple, but exactly because they are a nightmare of complexity.
A coatings manufacturer has to juggle REACH, CLP updates, changing hazard classifications, specific customer restrictions, and a dozen different national sustainability criteria. One tiny reclassification of a single substance can force you to overhaul hundreds of formulations, rewrite safety data sheets, and pull products from certain markets.
This is the perfect playground for AI. It can monitor massive volumes of text, flag changes between document revisions, instantly connect a flagged raw material to every formula that uses it, and help regulatory teams prioritize the real fires.
The wider chemical sector is catching on. A 2025 study looking at German chemical and pharma companies showed active AI use jumping from 34% to 76% in just five years [3]. And in that ChemSkills survey, dealing with regulatory changes was ranked as the second biggest driver impacting our industry, right behind sustainability [2].
But regulatory work also perfectly highlights why AI shouldn’t make the final call. Regulations aren’t just “yes or no” questions. They are full of thresholds, exemptions, transitional grace periods, and overlapping jurisdictions. An AI can give you a beautifully written, highly confident answer while missing the one condition that completely changes whether the product can legally be placed on the market.
The workflow needs to be clear: AI does the grunt work—searching, monitoring, and flagging. The human expert interprets, verifies, and makes the call. Deciding if a product goes to market isn’t just data retrieval; it carries heavy legal and commercial weight. I’d much rather pay my regulatory specialist to analyse a tough grey area than pay them to spend three hours manually comparing two PDFs.
AI can now propose promising coating formulations. How far can it take us before industrial reality takes over?
Evripidis: As a formulator, this is the stuff that really excites me.
There was a fascinating 2026 paper in npj Materials Degradation. Researchers took historical salt-spray test data covering 148 different ingredients and fed it into a machine-learning model. The model proposed two formulations using non-intuitive combinations of ingredients that a human would rarely think to mix. They actually made the paint, ran the cyclic salt-spray tests, and the proposed formulations showed corrosion-protection performance comparable to or better than the reference formulation [4].
When you see that, you have to pay attention.
The formulation universe is just too big. No chemist has the time to physically test every possible combination of resin, additive, pigment, and PVC level. AI can point us toward combinations we’d normally ignore because our experience whispers, “Don’t mix those.”
Usually, experience is right. But sometimes, experience just traps us inside the tiny corner of formulation space we already feel comfortable in. That’s where AI is truly disruptive.
However, getting a good salt-spray result in the lab isn’t the finish line—it’s just the starting gun for the real problems.
Can we actually manufacture it consistently? Does it separate after three months in a warehouse? How does it spray? Is it a nightmare for the applicator? Are the raw materials commercially available at an acceptable cost, and is their regulatory profile compatible with the intended market?
The authors of that study were completely upfront about this. They admitted it was a proof of principle, that the formulas weren’t commercially viable as-is, and that the AI couldn’t explain why the paint prevented corrosion.
That doesn’t make the AI useless; it just defines its role. AI isn’t going to hand me a finished, commercialized recipe ready for the factory. But it will highlight the three experiments I should run that I never would have thought of on my own.
I don’t expect AI to replace the lab bench. I expect it to eliminate a massive amount of wasted time at the lab bench.
A commercial paint formulation isn’t just a recipe. It’s a brutal compromise between chemistry, factory physics, environmental regulations, cost, and what the customer is actually willing to pay. AI can help us navigate that compromise faster than ever—but a human still has to understand it.
