Experimental optimization is a cornerstone of chemical research, driving the perpetual search for novel methods to accelerate the process. Statistical design of experiments (DoE) methodologies are tools designed to accelerate experimental optimization but traditionally struggle to incorporate the qualitative observational data that are often paramount to optimization. The rapid advancement of artificial intelligence (AI), particularly general‐use large language models (LLMs), presents a new avenue to address this limitation. We present an iterative AI‐enhanced DoE workflow that utilizes accessible AI tools available for minimal cost. In this workflow, an LLM digitizes handwritten lab notes into machine‐readable Markdown format. NotebookLM then consolidates observational data and integrates them into a prompt used by an LLM to perform a retrieval‐augmented generation (RAG) analysis. The result is a detailed research report grounded in the user‐provided information to provide insight into the chemical space, relevant literature to investigate, and experimental suggestions toward the goal of optimization. Finally, a human‐in‐the‐loop (HITL) review validates the AI‐generated report to ensure accuracy, relevance, and chemical safety. This approach aims to effectively bridge the gap between statistical DoE and qualitative insights, democratizing access to AI‐assisted experimental optimization.
Dahlen et al. (Tue,) studied this question.