Short-term memory stores recent inputs, decisions, and conversational context that an agent uses during an active session. It keeps the agent coherent and relevant within a task without mixing in long-term data or history from unrelated past interactions.
A customer service agent that remembers the user mentioned a specific order number earlier in the conversation and references it when answering a follow-up question — without asking again — is using short-term memory effectively. Without it, users experience the frustrating phenomenon of an AI that seems to have forgotten what they just said.
