September 17, 2026
AI Is Becoming a Better Lab Partner, Not a Replacement for Expertise
The useful role for AI in laboratory work is iterative: organize evidence, accelerate the experimental loop, and bring testable options back to the people accountable for the science and safety.
The useful version of AI in the lab is not a machine that declares the answer. It is an iterative partner: one that gathers and organizes experimental data, searches relevant literature, spots patterns across more conditions than a person can hold in their head, and brings new, testable options back to the experts responsible for the work.
That distinction mattered at ACS. Conversations with both researchers and EHS staff made it clear that better lab intelligence has to serve more than discovery. Researchers need help turning scattered information into defensible next experiments. EHS teams need workflows that are easier to inspect, document, and discuss before risks become incidents. GreenChemistry.AI can support both, because safer and more sustainable chemistry depends on the same foundation: visible assumptions, usable data, and informed human decisions.
AI in the experimental loop
The current trend is toward AI systems that participate in the experimental loop. Self-driving-lab platforms can help identify conditions, run structured experiments, analyze results, and propose the next iteration, while leaving researchers in supervision and decision-making roles. As Chemical & Engineering News reports, these facilities are becoming more autonomous, but experts building them still say no facility can operate without human intervention.
This isn't a limitation to work around. It is the point.
An experimental recommendation only becomes useful when someone with chemical knowledge can judge whether the evidence applies, whether the tradeoff is acceptable, and whether the protocol fits the operational reality of their laboratory. AI can make the loop faster and more complete. It cannot substitute for chemical intuition, operational experience, or accountability.
Safety has to be part of the loop
The same logic applies to EHS. A recent open-access paper, Chemist Eye, explores using vision-language models to identify possible PPE non-compliance, accidents, and fire hazards in self-driving laboratories, then notify staff and influence robot behavior. Its results are promising, but the authors are clear that the system is not ready to operate as a full-scale safety system.
That is the right standard. AI can expand situational awareness and make risks easier to surface. Safety professionals still determine what requires action, how a response should work, and where the system's authority must stop.
That is the conversation we are looking forward to continuing at Future Labs: with startups and research-science innovators building tools that make experts more capable, not less necessary.