City of Kingsport: Grace Goes Live With a Twenty-Year System Replacement
By Dial AI · May 19, 2026
Kingsport replaced a customer system it had run for two decades and put an AI agent on the phones the same week. Grace now handles 30+ concurrent calls, resolves more than 60% end to end, and does it about 50% faster than a human agent.
The City of Kingsport, Tennessee serves more than 55,000 residents with water, storm and wastewater utilities. For roughly twenty years it ran those services on a customer system built in-house — a platform that had long since stopped being able to flex the way a modern utility needs.
In April 2026 the city replaced it with Hansen CIS. Grace went live at the same time.
The decision that mattered
Most utilities would separate these. Replace the system of record first, let it settle for a quarter or two, then look at the contact centre.
Kingsport did both at once, and the sequencing is the interesting part of this deployment.
A CIS cutover is precisely when a utility’s call volume spikes. Every bill looks unfamiliar, every portal login is new, autopay enrolments have to be recreated, and the contact centre absorbs all of it in the same fortnight its own staff are learning the new system. Deferring the AI agent means asking customers to adjust twice, and riding out the worst weeks with the least help.
Going live together meant the phone line changed once — and Grace was reading from the new system of record from her first call, rather than being retrofitted to it later.
What went live
Cutover began the evening of Friday 23 April 2026. The city announced the new system publicly on 27 April, and go-live was confirmed complete in the days following.
Alongside the CIS, customers got a phone agent that answers questions about balances, due dates and water usage and takes payments securely, a new mobile app and web portal, and a much wider set of payment options — pay-by-text, digital wallets, and in-person payment at local retailers.
Results
Figures published by Hansen in May 2026:
| Measure | Result |
|---|---|
| 30+ | Concurrent calls handled |
| 60%+ | Of interactions resolved end to end, without escalation |
| ~50% | Faster to resolve than a human agent |
| 2 | Languages supported live, English and Spanish |
The concurrency number is the one utilities tend to underestimate. A contact centre’s worst day is not defined by call volume in aggregate — it is defined by how many people call at the same moment. Staffing for that peak is what makes coverage expensive, and it is the reason hold queues exist at all. An agent that takes thirty calls at once does not have a queue to manage.
Sentiment recognition runs underneath it, so a caller who is getting frustrated is identified as such rather than processed identically to everyone else.
In their words
Floyd Bailey, Chief Information Officer at the City of Kingsport, described the deployment as a “powerful force multiplier” integrating directly with the city’s customer information and relationship management system.
Bobby Slaton, EVP Energy & Utilities Americas at Hansen, made the point that the call centre is the most human-facing part of a utility, and that AI with sentiment capability should enhance that rather than disrupt it.
What we took from it
Two things worth carrying into the next one.
The migration is the moment, not the risk. The instinct is to treat a CIS cutover as too fragile to add anything to. In practice it is the point of maximum customer contact, which makes it the point where an agent earns its keep fastest.
Self-service adoption moves with it. Autopay enrolment through the new portal ran at roughly double the previous system’s within the first week. A customer already being asked to re-enrol is a customer willing to change how they pay — and the phone agent is what absorbs the questions that decision generates.
Work has continued since on outage management integration, so an outage call gets the same treatment a billing call already does: a real answer from the real system, rather than a promise that somebody will ring back.