68% of tickets, automated.
A support agent trained on their product catalog and policies now resolves most tickets without a human — and customers don't notice the difference.
Industry : E-commerce
Service : Custom AI Agent
Tickets resolved without humans
sec
Average first response time
3
3
X
Saved in support costs
Weeks
From kickoff to live
Havelio is a DTC skincare brand doing 40,000+ orders a month. Support volume had outgrown their 4-person team, and hiring wasn't keeping pace with growth.
The Problem
Support couldn't keep up.
Order status, returns, and product questions made up 80% of tickets — all answerable from existing docs, but still routed to humans one by one.
Response times stretched past a day during peak seasons. CSAT dropped as fast as order volume grew.
The Approach
A 5-week build.
Phase 01 — Weeks 1–2: Audit & Design
Categorized 12 months of ticket history by type and resolution path
Mapped product catalog, return policy, and shipping rules into a knowledge base
Designed escalation logic for anything outside the agent's confidence threshold
Phase 02 — Weeks 3–4: Build
Trained a support agent on the full product catalog, policies, and brand voice
Integrated with their helpdesk and order management system
Built automatic refund and exchange handling within policy limits
Phase 03 — Week 5: Deploy & Refine
Launched alongside the human team, agent-first with human fallback
Tuned confidence thresholds based on real ticket outcomes
Trained the team on monitoring and edge-case handling
The Solution
One agent, every channel.
The agent now handles the full support flow across email and chat:
Customer submits a question via chat or email
Agent pulls order data and matches it against policy
Agent resolves directly — refund, exchange, tracking, or product info
Anything uncertain routes to a human with full context attached
Every resolution is logged and used to retrain the agent monthly
The Results
90 days of results.
68% of tickets resolved without human involvement
$180K saved in support costs annually
2 minutes average first response time (down from 14 hours)
CSAT up 12 points despite the drop in human touch
Support team refocused on VIP customers and retention
"
"We expected the agent to handle the easy stuff. It's handling almost everything — and customers are happier than when humans were doing it manually."
Justin
Head of Customer Experience, Havelio
Releated Case Studies
+$340K pipeline in Q1
B2B SaaS • Lead Generation
4-hour response time, cut to 90 seconds.
2x client capacity, zero new hires
Marketing Agency • Workflow Automation
35 hours per week, returned to every account manager.
WANT RESULTS LIKE THIS?
Be our next case study.
We can usually predict your outcome on the first call.