Vertex SaaS: AI support ticket triage
Incoming tickets get classified, prioritized, and answered by AI in seconds, only the tricky ones reach a human agent, with full context attached.
68%
Tickets fully auto-resolved
38 sec
Median first response (was 4 min)
+19 pts
CSAT improvement in one quarter
2 agents
Redeployed to customer success roles
Overview
Vertex SaaS runs a B2B project-management platform with support volume growing threefold year over year. Their six-person support team was drowning: tickets sat unclassified, urgent outages waited behind password resets, and agents spent their days copy-pasting the same answers.
We built an AI triage layer in front of their helpdesk. Every ticket is now classified, prioritized, and, where possible, resolved by AI in seconds. Human agents only see the tickets that genuinely need judgment, arriving with full context and a suggested response already drafted.
The challenge
- 01
Ticket volume had tripled in a year while headcount stayed flat, backlog was measured in days, not hours.
- 02
Urgent issues (outages, billing failures) queued behind routine questions with no prioritization.
- 03
Agents answered the same 20 questions on repeat, burning out the team on work a machine could do.
- 04
First-response times were slipping and CSAT was trending down quarter over quarter.
Approach
How we got there.
No magic, just a disciplined process, applied with care at every step.
Step 1
Classify everything
Trained a classifier on 18 months of ticket history to tag topic, sentiment, urgency, and account value, reaching 94% agreement with human taggers.
Step 2
Resolve the routine
Built an answer engine on their docs, changelogs, and resolved tickets that fully resolves common requests, password resets, plan changes, how-tos, without human touch.
Step 3
Route the rest
Complex tickets route to the right specialist with priority scores, full thread context, and an AI-drafted suggested reply the agent can send, edit, or escalate.
Step 4
Keep humans in charge
Confidence thresholds decide what auto-resolves versus what gets reviewed. Every AI action is logged and reversible, and the team retrains the model monthly on edge cases.
The outcome
68% tickets auto-resolved
“Support went from our biggest cost center to a quiet machine. Our agents finally do the interesting work, the AI handles the repetitive 70%.”
- 38 sec median first response (was 4 min)
- +19 pts cSAT improvement in one quarter
- 2 agents redeployed to customer success roles
Have an idea? Let’s ship it.
One conversation is all it takes to turn “someday” into a launch date. Free consultation, zero pressure, fixed quote.