End trial and error
in coatings R&D.

For most of my career I watched the same thing happen over and over. Brilliant people, decades of hands-on knowledge, and a way of running experiments that worked for them and almost nobody else. Hold one variable fixed and change another. Then hold that one fixed and change the first. It is intuitive and it seems efficient, but it costs a lot of time, and you still miss most of what the experiments could have told you, because you never cover the design space in a structured way.

Design of Experiments and Bayesian optimization solve this, and they have existed for decades. But most teams never picked them up. The methods were hard to grasp without statistical training, and the software was complicated, with more features than most users need.

Generative AI changes that. You describe a problem to your computer in plain language and get an experimental plan back. No waiting for software support, no one person in the lab who has to know DoE. That's what EDH is for. Modern experimental design methods, available to everyone, not only the labs that can afford a data scientist.