CMC-in-the-Loop

AI is accelerating drug discovery faster than the work downstream can absorb it. The constraint that now decides how quickly a molecule reaches patients has moved into Chemistry, Manufacturing, and Controls (CMC), the process knowledge, methods, controls, and quality evidence that turn a candidate into a reproducible product. CMC is where timelines and cost concentrate, and it no longer happens inside one company’s walls. A modern development program runs across dozens of external partners connected largely by hand-keyed batch records, PDFs, and email, and data and context leak at every handoff.

This paper makes the case for a shift the industry can no longer defer: from accelerating individual experiments to orchestrating the entire scientific outcome. Lab-in-the-Loop speeds learning within a defined problem, yet local automation reaches its limit the moment a molecule crosses quality gates, manufacturing sites, and external partners. CMC-in-the-Loop changes the unit being coordinated, so process knowledge and quality data follow the molecule across R&D, MSAT, manufacturing, and external partners on one shared, governed data model, from research through clinical manufacturing and technology transfer.

Written by L7 Informatics Founder and CEO, Vasu Rangadass, Ph.D., the paper lays out what a coordinating layer requires: a common language, drawn from the ISA-88 and ISA-95 standards, that reduces translation across partners, and one governance framework that holds for people and AI agents with controls suited to each. It explains why adding more point tools compounds integration debt, and how a federated operating model and the FDA’s shift toward structured quality data through KASA and PQ/CMC are converging on the same direction. The organizations that treat the coordinating layer as strategic infrastructure will add molecules and partners without rebuilding their foundation each time, an approach the paper grounds in L7|ESP, the agentic operating system for precision science.