SupportIQ

home-hero
features
how-it-works
customers

Inspiration

Support teams lose time when every ticket arrives as plain text and someone has to manually decide: Is this billing? Is it urgent? Who owns it? Is the SLA at risk?

SupportIQ was built to make that first step faster and clearer. The goal is to help small teams turn messy customer requests into organized, prioritized support work.

What it does

SupportIQ lets a user create support tickets, then automatically triages each ticket by:

Category
Priority
Team
SLA risk
Confidence level
Matched keywords
Secondary signals
Needs-review flag for uncertain tickets

The app also includes:

Authentication
Protected dashboard
Ticket inbox with search, filters, and sorting
Ticket detail pages
Status updates
Notifications
Profile page
Settings backed by DynamoDB
Analytics from real ticket data

How we built it

The app is built with Next.js and deployed on Vercel. DynamoDB stores users, sessions, tickets, notifications, and workspace settings.

The triage engine is a rule-based MVP. It scores matched keywords, detects urgency and low-urgency language, routes tickets to the right team, and marks uncertain tickets for review.

We also added regression tests inspired by public customer-support datasets so triage quality can be measured instead of guessed.

Challenges

The hardest part was making triage trustworthy. A simple keyword match can be too blunt. For example, “car is broken” originally became high priority just because of the word “broken,” even though the message said “no urgency.”

We improved this by adding confidence scoring, low-urgency detection, secondary category signals, and a regression benchmark.

Another challenge was removing fake/static data. We moved toward real authentication, real DynamoDB-backed records, and dynamic dashboard data.

What we learned

We learned that a useful support system does not need to start with a giant AI model. A clear rule-based MVP with good tests, explainable reasoning, and human review paths can already create a strong foundation.

We also learned that triage should not just answer “what is this?” It should answer “how confident are we, what matched, and what should a human review?”