Tennessee's Department of Commerce & Insurance deployed Iris, an AI chatbot designed to handle professional licensing inquiries across 26 regulatory programmes serving 357,000+ licensees. The system integrates with the existing CORE licensing database and operates 24/7 in English and Spanish, addressing a genuine operational need: the department processed 220,960 customer inquiries in 2025 alone. Iris can access applicant files and licensee records whilst maintaining data privacy guardrails, and critically, it recognises its own limitations by escalating to human agents when necessary. The soft launch in August preceded the formal announcement, suggesting the department validated the approach before full rollout.
The deployment reveals a pragmatic government adoption pattern that CX teams should monitor closely. Rather than replacing existing customer-service infrastructure, Iris functions as a triage layer—extending availability beyond the 8 a.m.–4:30 p.m. weekday window whilst the department maintained a 96% satisfaction rating and reduced average wait times by four minutes in 2025. This raises a pertinent question for teams evaluating conversational AI: how much of your current ticket volume represents answerable, database-backed questions that could be deflected without degrading satisfaction? For organisations already running enterprise platforms like Salesforce Service Cloud or Zendesk, Iris demonstrates that government agencies are now comfortable embedding AI into regulated, data-sensitive workflows—suggesting compliance and audit concerns that previously blocked adoption are becoming surmountable.
The Tennessee model also exposes a gap in how many CX operations approach AI implementation. Rather than treating the chatbot as a replacement metric, TDCI framed Iris as "a seamless transition" between channels, implying the department understands that AI success depends on handoff quality and human oversight. For support teams currently piloting or scaling AI agents, this distinction matters: the technology's value lies not in deflection rates alone, but in whether it reduces friction for both customers and agents handling escalations. Whether your organisation operates in regulated sectors or not, Iris suggests that transparent failure modes and clear escalation paths are now table stakes for AI adoption.
Meet Iris: Tennessee launches AI chatbot for professional licensing help WZTV