Agents Cannot Be the Directly Responsible Individual

Simon Willison · rss · 2026-07-13

The article discusses the concept of the DRI (Directly Responsible Individual). The author looks up GitLab's definition and notes its alleged Apple origins: a DRI is the person ultimately accountable for the success or failure of a project, plan, or activity.

Applying this concept to LLM-driven agents, the author concludes that agents should not be viewed as DRIs. The reason isn't that they can't execute tasks, but because accountability inherently requires human ownership; machines lack the capacity to truly be responsible for their actions.

The article also references IBM's classic 1979 training page:

> A computer cannot be held accountable, so a computer must never make a management decision.

The overarching view is that as agents enter organizational workflows, humans must retain ultimate accountability rather than outsourcing it to the system.

Original post →

More from AGI Musings

AGI Musings channel →