How Muni-Link answers 80% of municipal support chatsfrom verified knowledge

The Challenge

Municipal billing staff do not have time to wait in a queue. When a billing run is due or a batch will not post, the clerk needs the answer now, and the same handful of questions arrive from hundreds of municipalities.

• Deadline-driven volume. Questions spike around billing runs and due dates across every municipality at once.
• Specialized billing workflows. Batches, readings, adjustments, and bill corrections follow Muni-Link-specific procedures that generic AI cannot guess.
• Incidents look like how-to questions. When something is down, customers ask why their task is failing. Without knowing about the incident, any bot sends them through troubleshooting steps that cannot work.
• Knowledge scattered across systems. Years of answers lived in chat transcripts and an older ticketing system, while the public knowledge base covered only part of what customers asked.

The Solution: Grounded Chat With Live Incident Awareness

IrisAgent runs on Muni-Link's Zendesk desk and moved with them through their migration from Jira to Zendesk, with no break in coverage.

What was implemented:

• Grounded chat for municipal staff. IrisGPT answers questions from Muni-Link's approved knowledge only, and hands off to the support team when the answer is not there.
• Status page awareness. IrisAgent reads Muni-Link's live status page. When an incident is open, the chat tells customers about it directly instead of walking them through troubleshooting.
• Knowledge gap analysis from four years of chats. IrisAgent analyzed four years of chat history to show which questions recur, which are already covered, and which are action requests that no article can deflect, so the team knows where product changes beat more documentation.

The Results

Across June to September 2026, measured in production:

• 80% of chat questions answered from verified knowledge. 80% of chat questions were answered from Muni-Link content. The rest went to people, not to a guess.
• Municipal staff served during billing runs. Clerks and billing staff got answers during their billing runs without waiting for an agent.
• Incident-aware answers. During outages, customers get the incident status first, which cuts repetitive tickets about an issue the team is already fixing.
• A clear map of what to fix next. The chat history analysis separated true knowledge gaps from product requests, giving Muni-Link a prioritized list rather than a guess.

Why Municipal Software Teams Choose IrisAgent

• Grounded, not guessed. Answers trace to your approved knowledge. Wrong billing answers are a public problem for your customers, so the AI declines rather than improvises.
• Survives your platform changes. IrisAgent moved with Muni-Link from Jira to Zendesk. Your AI does not have to be rebuilt when your helpdesk changes.
• Small-team friendly. No engineering project and no data minimum, which matters for vendors serving municipalities.