Alex SaagiEvery finance office keeps a list of analyses it would run if someone had a free week. At Hilton Head Public Service District, CFO Amy Graybill is working through hers by asking an AI assistant questions about the district's ADG Vantage data.
"A lot of this is stuff that we've always wanted to do, but nobody ever had the time, so we just never quite got to it."
Hilton Head Public Service District, SC
The district and its setup
Hilton Head PSD is a water and sewer special district on Hilton Head Island, South Carolina. It runs utility billing and finance in Vantage.
Vantage gives AI assistants read-only access to its data through the Model Context Protocol (MCP). The district's IT team recommended Claude, and Amy uses it with a standard Claude Pro subscription connected to Vantage. The assistant reads the data its key has been granted and answers in plain language. It can't change anything in Vantage. IT is now extending access to the purchasing manager and other staff, covering everything except payroll.
Three questions from the list
"How many accounts are in this circle?"
An engineer printed a map from the district's GIS and circled an area by hand. Counting the accounts inside it used to mean exporting addresses, stripping out the street names, sorting them, and picking out the ones inside the line by hand.
Amy gave Claude a picture of the map and asked, "Can you tell me how many accounts are in this circle?" Claude matched the drawing against the latitude and longitude stored on the district's service addresses in Vantage and returned a count that stopped cleanly at the line. "Sure enough, it just did it," Amy said. Working from coordinates also caught "tiny little streets with only a couple things on them. I never would have picked those up."
[Map photo attached] Can you tell me how many accounts are in this circle?
59 accounts fall inside the circle. I matched the latitude and longitude on each service address against the area you drew. Four of them are on a short side street just inside the north-east edge, which is easy to miss on the printed map.
"What would a vac truck have saved us?"
The district's collections department has questioned whether it would make sense to purchase a sewer vacuum truck. Making the business case meant combing through years of AP invoices for work a district-owned truck could have done, and the analysis had been on the radar "for a long time. No one's ever had the time to do it."
Amy asked Claude which invoices would have been avoided if the district had its own vac truck. It built a four-year table in two tiers: invoices the truck would have eliminated entirely, and invoices worth a closer look because part of the cost might have been reduced. "It totaled up to more than I was thinking." Whether that covers the price of a truck and its annual maintenance is still an open question, and the district now has the numbers to answer it.
"Which assets are in the wrong department?"
Location fields on the district's fixed asset records had been filled in inconsistently over the years. Amy gave Claude the district's departments, which match how its general ledger is organized, and asked it to allocate the assets across them. Along with the allocation, Claude flagged assets that looked miscoded in Vantage, such as "a sewer line marked as our distribution system for water." The finance team will correct those records.
The same approach updated the district's allocation of its insurance premium across departments, which Amy described as "an allocation we've been hoping to update for a while."
Only as good as the data
Amy is clear about where the answers come from. The vac truck analysis rests on the invoice descriptions her AP staff typed when they entered each invoice: "that's all it has to use."
That fits the read-only model. The assistant reads, counts, and drafts. Staff check the results and make any changes in Vantage under their own permissions. The fixed-asset review shows how it works in practice: the assistant found the problems, and people fix them.
Amy sees data cleanup as a use in its own right. In a utility billing system with thousands of accounts, a status changed by mistake can stop an account from billing, and "you don't notice that that one account now isn't billing anymore." An assistant that can read every account is well placed to find those.
What's next at Hilton Head PSD
Amy plans to use the assistant to streamline her board report and to support the annual cost-of-service study for the budget, looking at full trend data of how and when actual demand peaks and whether rates are designed around it instead of relying on a limited sampling of data. The purchasing manager, who is "excited to try it out," is next in line for access.
Starting at your organization
MCP access is included with Vantage at no additional cost. An administrator creates an AI-enabled key, grants it the modules and tables it should read, and staff connect the assistant your organization already uses, whether that's Claude, ChatGPT, or Microsoft Copilot. Every call is logged against its key. The MCP launch post covers setup, and AI & MCP in ADG Vantage covers governance and common questions.
Running a utility district? See how Vantage supports water and sewer districts and special districts.