News People & Careers
AI Transforming Resin Development in Coatings
AI is driving measurable efficiencies in the coatings industry, yet adoption remains exploratory, says Ioanna Tzortzi, associate specialist at Perstorp. In this interview, she shares insights into how AI optimises resin development, addresses data quality, and the potential for broader adoption.
How is AI being used to accelerate the development and optimisation of resin and additive formulations?Ioanna Tzortzi: AI can accelerate resin and additive development by learning relationships between formulation choices, process conditions, and performance, then using that knowledge to predict outcomes and guide decisions. It enables faster screening of candidates, identifies influential variables, and recommends robust operating windows beyond single-point recipes. Active/sequential learning has proven effective, where models update after each experiment and propose optimal trials to improve performance or reduce uncertainty. This approach is applicable across the coatings value chain — from resin and additive design to formulation tuning, application performance, and scale-up — by continuously learning from structured lab and process data.
Where do you see the greatest impact of digital tools today in the coatings value chain, and where is AI still underutilised?
Tzortzi: The coatings industry is not yet at a point where digital tools are embedded across manufacturing and application development workflows, so assessing “greatest impact” areas is premature. AI has reached a stage where companies are intrigued by its potential, but adoption remains uneven and exploratory. The market is actively seeking credible success stories that demonstrate where AI adds value, how it can be operationalised in day-to-day workflows, and the tangible benefits it delivers — such as improved speed, quality, and robustness — compared to traditional methods.
How do you ensure data quality and model robustness when deploying machine-learning models for R&D or process optimisation?
Tzortzi: Data quality, integrity, and structure are critical prerequisites for applying machine learning to R&D or process optimisation. We use a project-specific approach. For instance, in our AI-driven alkyd emulsification work, we defined all relevant qualitative and quantitative target variables and ensured experiments were consistently documented with no missing values. Where necessary, we engineered descriptors to reliably represent product metrics for model training. Model robustness is maintained through regular data updates, human-in-the-loop validation, benchmarking predictions against lab results, and monitoring performance metrics over time to track improvements or detect deterioration.
Do you see increased demand from coatings manufacturers for digital or AI-enabled services?
Tzortzi: Rather than clear demand for AI-enabled services, we see growing curiosity from coatings manufacturers. Customers want success stories and practical explanations of AI’s capabilities, how it can be integrated into daily workflows, and its advantages compared to the traditional trial-and-error approach. Our AI-driven alkyd emulsification work using Neptem exemplifies this. We benchmarked the AI model against a human-led approach, demonstrating significant gains in resource efficiency and material discovery.
