Deflection vs resolution.

Deflection rate is the share of support contacts handled without a person. Resolution rate is the share of contacts where the customer’s problem was actually solved. Deflection counts everyone who left a self-service or AI conversation, including the people who gave up, so a team can report a high deflection rate while many of those customers are still stuck.

By Colin Moroney, co-founder of Manic Edge · Published 15 October 2026

The two definitions

Deflection rate is the contacts handled entirely through self-service, a chatbot or an AI agent, divided by all contacts in the same period. It measures how much volume stays out of the human queue.

Resolution rate is the contacts where the problem was solved, divided by the contacts handled. For an AI agent, it measures how many of the conversations it took on ended with an answer that worked.

Vendors define both terms differently, and some use resolution to mean what others call deflection. Before either number goes into a report, write down what counts as handled and what counts as solved. Those two definitions decide whether an abandoned conversation is a success.

Side by side

Why deflection misleads on its own

Deflection rises for good reasons and bad ones, and the number looks the same either way. A customer who reads an article and gets their answer counts as deflected. So does a customer who gives up on the chatbot, a customer who closes the window and phones instead, and a customer who accepts a wrong answer and finds out a week later.

The last case is the expensive one. An AI agent drafting from out-of-date content gives confident answers that look like successes in the deflection report. The cost shows up later as repeat contacts, escalations and complaints, often on a different channel from the one being measured.

How to measure resolution

No single signal tells you whether a problem was solved. Three together come close.

Re-contact rate

The share of automated conversations followed by another contact about the same issue within a fixed window, such as seven days, on any channel. This is the most direct check on a deflection number.

Confirmation

An explicit step at the end of the conversation that asks whether the problem is solved. It undercounts, because many customers leave without answering, so read it alongside re-contact.

Scored conversations

A sample of automated conversations scored against the same quality rubric the team uses for human replies. This is the only signal that catches a confident wrong answer the customer accepted.

Where this sits in the maturity model

In the Support Maturity Model, deflection belongs to stage four, where AI deflection and agent assist are deployed. Resolution depends on stage three of the Metrics & Quality lens: quality scored against a fixed rubric, applied the same way every time. A team that adds deflection before it can score quality has a volume number and no way to check it.

That is why the model puts AI at stage four. The quality score built at stage three is what an AI deployment gets checked against, and it is what turns a deflection rate into a resolution rate.

What to report

Report the two together, with re-contact beside them. Deflection shows how much volume automation keeps out of the queue. Resolution shows whether that volume was helped. Re-contact shows whether the help held. If deflection rises while re-contact rises with it, the automation is moving work around, and the pair is what makes that visible to leadership before customers make it visible in other ways.

Take the Support Maturity Check Read the full model