AI pricing workflows highlight the need for data lineage

A sales representative asks an AI agent to find customer pricing. The agent combines information from restricted pricing systems, internal communications, and archived quotes. Although the result appears accurate, another agent uses it to prepare a proposal. Neither transfer receives human review. The finished document contains unapproved, confidential, and sensitive information, but the organization cannot determine where the problematic data entered the process or how it moved through systems, files, and agents.

The scenario is presented as evidence that AI governance must address data that is continuously ingested, copied, summarized, transformed, reused, and redistributed. The article says that conventional controls focused only on applications, systems, networks, or endpoints are insufficient for agentic AI workflows. It argues that organizations need data lineage—the ability to understand how information travels—to support governance, accountability, and trust across human, agentic, and hybrid processes.

The material was produced by Insights, MIT Technology Review’s custom-content division, rather than its editorial staff. It was researched and written by humans; any AI tools used were limited to production under human oversight.