Persistent delegation
Handing an AI agent a goal it keeps carrying over time — remembering context, noticing relevant events and acting across applications — rather than completing a single prompted task.
Most AI use today is a tool we operate: we write a prompt, the system returns an output, and the conversation ends. Persistent delegation is a different contract. The user hands over a goal, and the agent keeps working on it between prompts, across applications, for as long as the goal matters.
To make that work, the system has to remember what it is trying to achieve, understand which new events are relevant, decide which actions fit the goal, distinguish routine work from consequential decisions, and keep a usable record of what happened while the user was elsewhere. The product is continuity.
What makes it valuable
The real value is not more automation. It is less coordination in the user's head: the goal, the work already done and the systems where changes appear stay connected without the user holding them together.
What makes it risky
Persistence is the strength and the liability. It compounds useful context and bad assumptions alike, creating context debt. It forces clear lines of authority, and it must be judged by supervision load, not by how much the agent does while you are away. The real test is whether we can stop managing the AI.
Read more in OpenAI Dots is a test of whether AI can carry a goal, not just complete a task.
Related terms
Context debt
The growing difficulty of knowing what a persistent AI agent currently believes, where that belief came from, and how to correct it, as the agent accumulates memory over time.
Capability is not authority
A design principle for AI agents: being technically able to perform an action does not mean the agent should be permitted to perform it. Delegation needs gradients of authority.
Agentic AI
AI systems that pursue a goal by planning, using tools and taking multi-step actions with some autonomy, rather than only answering a single prompt.
Supervision load
The human attention an AI system still consumes — context supply, review, approval, correction and exception handling — measured per dependable, completed unit of work.