Formulation teams in home care, industrial cleaning, and specialty detergents are under growing pressure. Development cycles are shorter, supply chains less predictable, and regulatory requirements around surfactant biodegradability and REACH compliance are tightening. At the same time, many R&D organisations are being asked to evaluate artificial intelligence (AI) while still relying on fragmented data systems that were never designed for iterative formulation work. 

This article argues that the data infrastructure problem needs to be resolved before AI can deliver any value for R&D teams and formulators. It examines why common laboratory tools, including LIMS, electronic lab notebooks (ELN), and spreadsheets, fall short of what modern formulation development requires. It then outlines three principles for building functional data infrastructure within a material intelligence framework: data integration across instruments and IT systems, structured workflows, and AI readiness as a consequence of the first two. 

The article distinguishes between two practical AI applications, an AI assistant for data retrieval and predictive modelling using machine learning and Bayesian optimisation, and describes what each requires to function reliably. It concludes with a practical approach to implementation, starting with a scoped assessment of current workflows rather than a full system overhaul.