Stepthrough Start free

What is ticket deflection?

Quick answer

Ticket deflection is when a customer's question gets answered by self-service, such as an FAQ page or an AI chat agent, so nobody on your team has to open a ticket or write a reply. The deflection rate is the share of questions handled that way.

Last updated

Take the free module →Module 33 of the free AI Apps Worth Knowing course, about 14 minutes. No account needed.

Updated · By Robert Breen

Why it matters for a small business

Every repeat question costs staff time, and the cost is easy to miss because each one only takes a few minutes. Forty emails a week asking about hours, prices and refunds add up to hours a week of skilled people typing answers that already exist in the FAQ. Deflection is the plain business measure of how much of that work a self-service tool takes off the team.

It is also a number that can lie. A customer who gives up and closes the chat counts as deflected, yet their problem is not solved, and they may call later in a worse mood. So pair the deflection rate with a quality check: read a sample of the chats, look at feedback, and count how many people came back with the same question.

Working out a deflection rate (example numbers)

ItemFigureHow you get it
Questions a week40Count the repeat emails before the agent goes live
Answered by the agent30Chats closed with no ticket and no follow-up email
Handed to a person10Tickets created or chats a person joined
Deflection rate75%30 divided by 40
Time saved a week2 hours30 questions at 4 minutes each

In a real lesson: Chatbase: Answer Customer FAQs with an AI Agent

The Chatbase FAQ agent lesson opens with the deflection case without using the word. Dana Ruiz at Juniper Lane Bookkeeping (made up) gets about 40 emails a week with the same questions. At 4 minutes each and $45 an hour, that is about $520 a month, before counting the customers who wait overnight.

You train the agent on a 22-question FAQ file, tell it in Guardrails to answer only from that, and test real questions in the Playground. Anything it cannot answer goes to a person, through a fallback answer or the Escalations action that creates a support ticket. The lesson's habit is 15 minutes a week in Activity, reading the Chat logs and fixing weak answers with Revise answer. Those logs are where you would count what was answered and what was handed off.

Chatbase's Playground for the chat bubble: Model, Data sources marked Trained, Actions including Escalate to human, Instructions, and a chat preview of the demo agent TaskFlo (screenshot from Chatbase's docs)
Chatbase's Playground for the chat bubble: Model, Data sources marked Trained, Actions including Escalate to human, Instructions, and a chat preview of the demo agent TaskFlo (screenshot from Chatbase's docs)

Try this lesson free or read the step-by-step guide.

Common confusions

Ticket deflection vs ticket resolution

A resolved ticket was opened and then solved by a person. A deflected question never became a ticket. Deflection only counts as a win if the customer really got their answer.

Ticket deflection vs a knowledge base

A knowledge base is the set of articles or facts an agent answers from. Deflection is the result you measure once customers start using it.

Tips

  • Count the repeat questions for two weeks before you launch, so you have a baseline.
  • Read a sample of "deflected" chats each week to check the answers were right.
  • Add every new question you find in the logs to the FAQ.

More Business terms

Where you use it: free lessons

AI tools that use it

More terms by tool

Frequently asked questions

What is a good ticket deflection rate?
There is no single target. Start by measuring your own baseline, then aim to raise it while the answers stay correct and customers stop sending the same question twice.
How do I measure deflection?
Compare how many questions came in with how many needed a person. Chat logs, ticket counts and follow-up emails give you both sides.
Does deflection mean customers never reach a person?
No. Good deflection answers the routine questions and hands everything else to a person quickly, with the details already collected.

All AI glossary terms, A to Z · Free prompt templates