Augury open-sources profile to capture plant-floor know-how for AI

Augury has released an open-source manufacturing profile for turning experienced workers’ unwritten plant-floor know-how into auditable files for AI agents, according to its Sept. 10 company announcement. The OKF Industrial Profile sets conventions for recording operational advice, the evidence behind it, contributor credentials and human review status.

Built on Google Cloud’s Open Knowledge Format, the profile is intended to capture details such as how a production line behaves during cold starts, how an experienced lead handles shift handovers and when a sensor reading should not be trusted. Operator knowledge can be recorded with verbatim quotes and source records. Separate rules tell AI agents how to handle stale or stuck sensor feeds before making diagnostic recommendations and distinguish cause-and-effect relationships from redundant signals.

“The true bottleneck in plant-floor AI isn't model intelligence; it's capturing the unwritten wisdom of expert workers before they retire,” Anoop Mohan, Augury’s chief product and technology officer, said in the announcement.

Human judgment also figures in how recommendations are used. In a company-published Q&A, Ronen Salomon, principal in Augury’s CTO office, described traceability as a way for plant teams to weigh recommendations against their own knowledge and decide for themselves. He suggested starting knowledge capture through writing, recording or conversations with the people who hold it.

The profile is designed to work alongside existing plant systems, referencing operating limits and live data from their source systems rather than copying values, Augury said. The company is making it available on GitHub under the Apache 2.0 license, with documentation and an example knowledge bundle covering common plant assets and their relationships. It also plans to submit that bundle to the upstream OKF project through its contribution process.