Context drift in conversation AI is the gradual loss or corruption of the conversation state—what the caller is trying to do, what has already been confirmed, and which constraints apply—so the system starts interpreting new utterances against the wrong context.
Operationally, it matters because it increases misroutes, repeated questions, incorrect confirmations, and longer handle times. It can also create compliance and customer-experience risk when the AI applies the wrong account, policy, or prior answer after a hold, escalation, or topic change.
Tracking context drift as a signal helps teams identify where the AI needs better state management, clearer handoff rules, or stronger recovery behaviors (for example, re-confirming intent after interruptions) to reduce rework and avoid avoidable transfers.
