Key takeaways

• A fully autonomous, self‑driving laboratory for chemical reactions has now been demonstrated.
• By integrating automation with AI, the system can iteratively move toward a defined experimental goal with minimal human intervention. 
• A comparable self‑driving lab for coatings is well within reach, and may already exist. 
• Chemists who combine deep domain expertise with the ability to work effectively with AI and automation will become exceptionally valuable in the near future.

Artificial intelligence holds tremendous promise for R&D, particularly in laboratory environments. The AI used for experimental data analysis is analytical AI, which is fundamentally different from the generative AI behind ChatGPT and other chat interfaces powered by large language models (LLMs).

Although ChatGPT brought AI into the public spotlight, those of us in the lab know that AI in scientific research predates ChatGPT by many years. Laboratory informatics platforms with built‑in machine learning capabilities have been available for at least seven years. These systems are designed to store all of a research organization’s experimental data in a modeling‑ready format, making it straightforward for scientists to build predictive models and generate data‑driven suggestions.

Consider a coatings chemist who has prepared 30 formulations without hitting the performance targets. Using the informatics platform, the chemist can build a model from those 30 formulations and their respective test results, then ask the model to suggest the next formulation most likely to meet the targets. After preparing that formulation in the lab and adding its measured properties to the dataset, which now totals 31 formulations, the chemist can retrain the algorithm on the 31 formulations to generate an updated model. This iterative process can continue with formulations 32, 33, and so on. Eventually, the chemist reaches the targets far faster than by trial‑and‑error alone.

Of course, the quality of any model depends on the quality of the data. Variability in how a single chemist performs a test, combined with variability from one chemist to another, introduces noise that ultimately reduces the reliability of any model built on that data.  Automation is the most effective way to minimize this noise. When formulation preparation and property testing are automated, data becomes far more reliable, and the resulting models become far more accurate and precise.

In the ideal scenario free from human error, a scientist could instruct an automated system to prepare an initial set of say 30 formulations with measured performance properties. The system would prepare each formulation, measure its performance properties, and send the data to modeling software. The software would build a model, suggest the next experiment to run in order to hit the targets, and send that suggestion back to the automated system. The system would then prepare formulation 31, measure its properties, and feed the new data back into the model. The loop continues; experiment, measure, model, suggest until the targets are met. The scientist’s role becomes one of defining the initial experiments, selecting raw materials and designing the workflows; after that, the system runs autonomously.

This autonomous vision is no longer theoretical. In November 2025, Atinary and Chemspeed unveiled a Self‑Driving Lab for chemical reactions, which is a fully closed‑loop AI and robotics system that embodies this autonomous experimentation paradigm.(1) For a short video on how the system works, click on the following link:

Post | LinkedIn

This is one of the first real examples of a fully autonomous, self‑driving laboratory. Chemists are still essential for designing the initial experiments and workflows and getting the system started, but once it’s running, the platform learns from each experiment and continues the cycle on its own. A chemist with strong domain knowledge (knowledge of their field) and the ability to work with AI and automation will be extremely valuable in the near future  The particular installation here is focused on chemical reactions and it naturally raises the question: how far are we from a comparable self‑driving system for coatings?

On the Chemspeed website, there’s a March 2025 announcement about Covestro opening an automated laboratory for developing coating and adhesive formulations.(2)  Chemspeed appears to be the vendor behind that system. Whether it is already fully autonomous isn’t stated, but given the technology currently available, turning an automated coatings lab into a truly self‑driving one doesn’t seem like a stretch.

(1) Grand Opening of Chemspeed X Atinary Self Driving Lab

(2)  Covestro to Open Automated Laboratory for Developing Coating and Adhesive Formulations

 #ai #artificialintelligence #automation #autonomous #atinary #chemspeed #innovation #coatings #chemicals #consulting

(photo credit: unsplash.com, hermeus)