C
Cloptim
Book intro call
← All case studies

B2B SaaS · Customer Operations

Composite engagement

Cutting tier-1 ticket volume 71% at a Series-B SaaS

Replaced first-touch human triage with an agent that resolves billing, account, and integration questions, escalating only what genuinely needs a human.

71%
tickets auto-resolved
62s
median first response
$340K
annual cost recovered
4.6/5
post-resolution CSAT

Client type

Series-B SaaS

Size

~140 employees, $25M ARR

Region

North America

The problem

Customer Ops was drowning in tickets after a successful product expansion. 60% of inbound was billing, account, and integration questions that didn't need a human, but the team was spending 80% of their hours on them. Hiring couldn't scale fast enough, and CSAT was sliding from 4.7 to 4.2.

Approach

  1. 01

    Pulled 6 months of ticket transcripts; clustered by intent and identified the top 22 resolvable patterns

  2. 02

    Built a retrieval-augmented agent with ground truth from the help center, billing system, and product docs

  3. 03

    Wired tools for refund processing, plan changes, integration token rotation, and seat management. All with allowlists and human-approval gates on destructive actions

  4. 04

    Stood up a Langfuse-based eval suite running against 500 historical tickets on every code change

  5. 05

    Rolled out behind feature flag to 10% of traffic, then 50%, then 100% over 3 weeks based on eval data

Architecture

EmailZendeskIn-app chatCS Agentretrieval · tools · evalsHelp CenterBilling APICRMRefundsAnswerActionEscalate

Outcomes

71%
tier-1 auto-resolution rate
from 0% baseline
62s
median first response time
from 4h13m
$340K/yr
team capacity recovered
redeployed to enablement
4.6/5
post-resolution CSAT
above pre-launch baseline
<0.4%
incorrect-action rate
all caught by approval flow
0
production incidents in 90 days
tracked via eval suite

"We had three vendors quote this. Cloptim was the only one who started with our eval strategy and ticket data, not the model. That's why we picked them, and that's why it shipped."

Director of Customer Operations

Timeline

  1. Week 1

    Discovery: ticket clustering, ROI model

  2. Week 2

    Design sign-off, eval suite spec

  3. Weeks 3-4

    Agent v1 + tool integrations

  4. Week 5

    Eval suite live, internal beta on 10% traffic

  5. Week 6

    Rolled to 50%, then 100% production

Tech notes

Want one of these for your team?

Most engagements look like one of these archetypes plus a tailored discovery sprint. Tell us your workflow on a 20-min call.