80% of your queries do not need a person
The problem in a support operation is not volume: it is that the 80% shares a queue with the 20% that does need judgement, and delays exactly what matters.
Ask at your institution how many queries come in each month and you will get the number right away. Ask how many of those actually needed a person and the conversation gets uncomfortable, because that figure almost never exists.
It is the figure that decides everything else.
Two different queues in the same line
In most operations there are two kinds of query mixed in the same place.
The first are repetitive and have a correct answer: the balance, the due date, the status of an order, how to update a detail. They do not require judgement. They require access to a system.
The second do require judgement: a complaint, an exception, an upset customer, a case that does not fit the policy. There a person contributes something no system does.
The problem is not that the first kind exists. It is that they sit in front of the second kind in the same line. The customer with a complaint waits behind thirty people asking for their balance.
| Queries resolved without human involvement the agent closes the case, it does not hand it off | over 80% |
|---|---|
| Time to first response from when the query arrives | −61% |
Why a chatbot does not shrink the queue
| Criterio | Chatbot | AI agent |
|---|---|---|
| Where the answer comes from | Text written in advance for a recognised intent. | The institution's systems, queried at that moment. |
| What it does with the case | Hands it off. It returns to the queue. | Completes it. It does not return to the queue. |
| Effect on the workload | Moves it, and adds a customer who already tried and failed. | Removes it, and frees the line for what needs judgement. |
An AI agent checks the institution’s systems, decides within the policies it was given and completes the task. The query does not come back to the queue because it finished.
The cost that does not appear in the report
Support reports usually measure volume, response time and satisfaction. None of the three captures what the mix really costs.
What it costs is that the case that could have retained a customer was handled late, that the person with the best judgement on your team spent the morning confirming balances, and that growing in customers meant growing in headcount at the same rate. That last point is the one a CFO recognises instantly: a cost that rises with every new customer instead of falling.
Questions on this topic
- What is the difference between a chatbot that answers questions and an agent that resolves?
- A chatbot recognises an intent and returns text written in advance; if the customer needs something that requires checking a system, it hands off to a person. An AI agent checks the institution's systems, decides within the policies defined and completes the task, leaving a trace of every decision.
- How is it decided which cases a person should handle?
- The institution defines it, not the vendor. You declare what the agent can do, how far it can go and under what conditions it hands the case over. Anything outside that policy escalates to a person with the full context of the conversation.
