AI SaaSFor OmniPulse
OmniPulse — AI Support Automation
Duration
3 Months
Team Size
4 Engineers
The Challenge
The client's support team was drowning in 12,000 monthly tickets with a 19-hour average response time — directly hurting NPS and causing enterprise churn.
The Solution
We built a Next.js 14 dashboard with a fine-tuned GPT-4o model trained on 3 years of resolved tickets. A confidence-threshold router escalates edge cases to human agents, while a live analytics layer tracks deflection rate and CSAT in real time.
Tech Stack
Next.js 14OpenAI GPT-4oPineconePostgreSQLIntercom APIRedis
The Impact
0Tickets resolved autonomously
0Average first response (was 19 hrs)
0NPS improvement in 60 days