Handoff rate and escalation rate
In a team without AI, the escalation rate measures the cases that move from the first line to a higher tier.
With an AI agent, many tools use the same name for the step from the AI to a person. Before comparing two figures, it is worth knowing which one each is.
Why it matters
A low handoff rate is not good in itself.
It can come from an agent that answers what it does not know or that makes it hard to reach a person. A high one can be right in a month with an incident. That is why the number matters less than the split by reason.
The formula, with an example
The handoff rate is the conversations handled by the AI that passed to a person, divided by all the conversations the AI handled, times one hundred.
The figures in the example are illustrative.
- Example. If the AI answered in 1,000 finished conversations and 380 passed to a person, the handoff rate is 38%.
- With resolution. Counted over the same conversations, the handoff rate and the AI resolution rate add up to 100%.
The reasons, and what each one fixes
Each handoff reason points to a different fix.
| Reason | What it usually means | Where it gets fixed |
|---|---|---|
| The customer asked for a person | An answer that did not convince or a sensitive topic | Reading those conversations before changing anything |
| It found no answer | Missing documentation | The missing article |
| A rule flagged it | The topic goes to the team by company decision | Nothing, if the rule still makes sense |
| The customer was upset | A product problem or an earlier answer | The underlying issue, with product |
| Too many turns | The AI does not understand the request | The instructions or a guided response |
How Intake counts it
Intake calls it “Passed to a person”.
Of the finished conversations where the AI gave an answer of its own, it counts those a team member stepped into, because it was handed over, assigned or someone wrote. Added to the AI resolution rate it makes 100%, and every handoff keeps its reason.
