Resolution rate

Resolution rate is the share of support conversations that end with the customer's issue solved. In AI customer support it is usually split in two: the automated resolution rate (solved by the AI agent with no human involved) and the overall rate, which includes conversations a human finished after escalation.

How it is measured

Resolution rate = resolved conversations ÷ total conversations, for a given period and channel. "Resolved" needs a definition: the customer confirmed, the customer did not write back within a window, or a human marked it closed. Pick one, write it down and keep it fixed so the trend is comparable.

Resolution vs deflection

Ticket deflection counts conversations that never reached a human. Resolution rate asks whether they were actually solved. A deflected conversation that left the customer without an answer is not a resolution, which is why both numbers belong on the same dashboard.

What raises automated resolution

• Knowledge coverage. The agent can only resolve what its knowledge base covers. Review unresolved conversations weekly and add the missing articles.
• Actions. Questions that depend on live data (order status, booking availability, account details) are only resolvable if the agent can query the system.
• Honest escalation. An agent that admits it does not know and escalates will have a lower automated rate but a higher overall rate and better CSAT than one that guesses.

In SolveoAI

SolveoAI's analytics report conversation outcomes per agent and channel, and unresolved conversations are the usual source of the next knowledge base additions. See the analytics documentation.