AskFlow — AI Customer Support Agent
Two support agents, 400+ tickets a week, 14-hour response times, both flagging burnout. Off-the-shelf AI tools couldn't handle 1,200 SKUs and 6 languages. The custom agent now handles 70% of tickets in under 30 seconds — and one team member said it's the first job where she doesn't dread Mondays.
Scope: Support agent + Zendesk + order lookup; trained on catalog, policies, and six months of tickets.
The Challenge
400+ tickets/week, 14h average first response, CSAT dropped from 4.6★ to 3.9★. If one agent quit, replacing them was a €60k, 3-month problem. Zendesk AI, Intercom Fin, and Freshdesk all tried and failed.
The Solution
GPT-4o agent with RAG over the full product catalog and policy docs, real-time order lookup via REST API, confidence-threshold escalation with pre-drafted replies, Zendesk integration, and session memory via Redis across 6 languages.
Client
Priya N.
“My team was on the verge of quitting. 400 tickets a week, 14-hour response times, weekends ruined. We'd looked at every off-the-shelf AI support tool — none of them could handle our product catalog or our return policy without six months of configuration. Kiril had it running in four weeks. It handles 70% of tickets now, correctly, on its own. My team went from firefighting to actually improving things. One of them told me last month it's the first job where she doesn't dread Mondays. That means more to me than the metrics.”
Priya N.
Head of Customer Success, Norvik Home
norvikhome.eu
Key Results
Tickets fully automated
First response (was 14 hours)
CSAT restored (was 3.9★)
Before → After
First Response Time (hours)
CSAT Rating
Tickets Automated
Timeline
4 weeks · 2025
Completed 2025
Scope: Support agent + Zendesk + order lookup; trained on catalog, policies, and six months of tickets.
More case studies
Want results like these for your business?
I'll map out a clear plan, timeline, and realistic results for your project — no obligation.