Agents & Agentic AI
Working memory
The actively maintained information an agent uses while solving the current task.
Example
While solving a task, the agent tracks its plan, intermediate facts, and unfinished steps.
Why it matters
It helps an agent keep track of details needed for the next few steps.
You'll hear it
“Which assumptions are we still using for this calculation?”
Can you control it?
Sometimes
Sometimes. Some products or workflows let you influence working memory, while others keep it fixed or manage it automatically. Check the settings and documentation for the tool you are using.
Common questions
- Can working memory fill up?
- A system can have limits on how much active information it carries, requiring some details to be summarized or dropped.
- Does keeping something active make it true?
- No. An incorrect assumption can remain in use until it is corrected.
- Can interruptions disrupt the task?
- Yes. A resumed process may need its active information restored before continuing reliably.