Customer support · Production

Breathe boosts Fin AI resolution rate from 56% to 82% in under a year

The problem

Breathe's existing custom bots and knowledge base failed to deflect most customer queries, with the majority still reaching live agents. The support team was flooded with repetitive how-to questions, preventing agents from focusing on complex issues, while the business needed to scale support without growing headcount.

First attempt

Custom bots linked to the knowledge base did not prevent most queries from reaching the live support team.

Workflow diagram · grounded in source

Customer query arrives

trigger

“Breathe receives around 7,000 conversations per month, and Fin is involved in 62% of them”

Knowledge base lookup

ai_action

“I put lots of work into our knowledge base initially to make sure that Fin knew how to answer a range of questions about the functionality of our platform”

Fin Guidance rules applied

ai_action

“the team turned their focus to Fin Guidance and implemented a set of clear rules to ensure Fin handled each interaction with care, empathy, and the right tone”

Sensitive topic routing

routing

“For sensitive topics like cancellations or pricing, Fin was guided to instantly route the conversation to a human”

Autonomous query resolution

output

“Fin is involved in 62% of them – resolving the majority. This has taken a significant load off the support team”

Snippet creation from misses

feedback_loop

“Every time Fin missed the mark, they reviewed the conversation and created a custom Snippet – a reusable, accurate answer Fin could use next time”

Reported outcome

In under a year, Breathe increased Fin's resolution rate from 56% to 82%, with Fin involved in 62% of approximately 7,000 monthly conversations.
CSAT scores consistently hit 85–90%, equal to agent CSAT ratings, and Fin now resolves up to 88% of queries.

Reported metrics

Fin resolution rate improvement56% to 82%

Fin current resolution rateup to 88%

Fin involvement rate62%

Monthly conversationsaround 7,000

Show all 8 reported metrics

Fin resolution rate improvement56% to 82%

Fin current resolution rateup to 88%

Fin involvement rate62%

Monthly conversationsaround 7,000

CSAT score85 – 90%

Resolution rate milestone October 202468%

Time to reach 82% resolution rate9 months

Support team load reductionsignificant load off the support team

Reported stack

Fin AI AgentFin Guidance

Source

https://www.intercom.com/customers/breathe

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

In under a year, Breathe increased Fin's resolution rate from 56% to 82%, with Fin involved in 62% of approximately 7,000 monthly conversations.

What tools did this team use?

Fin AI Agent, Fin Guidance.

What results were reported?

Fin resolution rate improvement: 56% to 82%; Fin current resolution rate: up to 88%; Fin involvement rate: 62%; Monthly conversations: around 7,000 (source-reported, not independently verified).

What failed first in this deployment?

Custom bots linked to the knowledge base did not prevent most queries from reaching the live support team.

How is this customer support AI workflow structured?

Customer query arrives → Knowledge base lookup → Fin Guidance rules applied → Sensitive topic routing → Autonomous query resolution → Snippet creation from misses.

In under a year, Breathe increased Fin's resolution rate from 56% to 82%, with Fin involved in 62% of approximately 7,000 monthly conversations.CSAT scores consistently hit 85–90%, equal to agent CSAT ratings, and Fin now resolves up to 88% of queries.