AI-Assisted Support System Prototype

The problem

A 3-person Service Desk supporting ~1,500 users was drowning in repetitive, common-pattern tickets — password resets, access requests, known-issue troubleshooting — while running at roughly 75% of designed headcount. Every hour spent on a question that had already been answered a hundred times was an hour not spent on the tickets that actually needed a human.

The solution

I built a system prototype trained on historical ticket data, existing knowledge base articles, and recurring-issue patterns mined from the ITSM system. It handles common requests automatically and around the clock, and surfaces emerging issue clusters back to the team before they escalate.

Impact

• 24/7 automated coverage for the most common request types
• Meaningfully reduced administrative load on a 3-person team
• Faster identification of recurring incidents before they became trends

How it works

Lessons learned

• The highest-leverage automation isn't the flashiest ticket — it's the one that repeats the most.
• A system is only as good as the knowledge base underneath it; cleaning up KB content was half the project.
• Transparency with the team mattered — this was framed as taking work off their plate, not replacing them, which is exactly what happened.