What is a handoff in customer support?
A handoff is the transfer of a conversation from an automated system to a human, along with enough context that the person does not have to start over.
Every automated support system needs a way out, and the quality of that exit matters as much as the answers. A good handoff carries the transcript, so the customer is not asked to repeat what they already typed — which is the single most common complaint about support automation, and one that is entirely avoidable.
Why it matters
The useful question is not whether to have handoffs but when to trigger them, and three conditions are concrete enough to implement. First, weak retrieval: nothing sufficiently close to the question was found, so any answer would be invention. Second, an explicit request — someone who asks for a person should get one, immediately, with no interrogation about why. Third, subject matter the business has decided not to automate at all, such as anything involving a complaint, a legal question, or someone's health. A system that hands off on all three will contain fewer conversations and lose fewer customers.
Examples
- A question about a specific order, which no published page can answer, routed to an agent with the transcript attached.
- A visitor typing 'talk to a human' and being given the contact form immediately.
- A message that reads as a complaint, routed to a person regardless of whether an answer existed.
Not the same as
Escalation
Escalation usually means moving up a support tier between people. A handoff is the first move from machine to person.
Related
term
What is containment rate?
Containment measures where a conversation ended up, not how it went. A conversation is contained if it never reached an agent — including one where the customer gave up. This makes it easy to report and easy to game: removing the handoff button raises containment immediately.
term
What is deflection rate in customer support?
If a thousand people ask something in a month and seven hundred get what they needed without a person, deflection is seventy percent. It is the headline number for whether support automation is doing anything, and it is the number most often reported wrong.
term
What is a confidence score in AI answers?
Two very different things get called confidence. Retrieval confidence measures how close the best-matching passages were to the question — a property of the search step, computed before any answer is written. Model confidence is the model's own assessment of its answer, which it will happily report as high while inventing a price.
use case
Answer customer questions from your own content
Answering questions well means answering them from something true. RubyRep retrieves the relevant passages from your site before it writes anything, and cites the page it used, so an answer can always be checked.
tool
First contact resolution calculator
Works out FCR as a percentage, then goes further: how many conversations needed a second touch, and what that repeat contact cost in hours and money.
industry
AI customer support for home services
A trades enquiry is worth a great deal and expires quickly — someone with a leak calls three numbers and books whoever picks up. You cannot pick up while under a sink. RubyRep answers the qualifying questions on your website so the enquiries that reach you are the ones you can actually serve.
RubyRep is this, applied to your website.
Give it your URL and it reads your pages, then answers your customers from them.
Train My RubyRep