Student demonstrates AI chatbot for Kensington website; board asks about cost, data and accessibility

A University of New Hampshire student presented a prototype municipal chatbot to the Kensington Select Board, saying it can help residents find information buried in PDFs and webpages and can be trained to answer local questions.

The student identified himself during public comment and described the product as a low‑cost, add‑on widget that integrates with common municipal website platforms. "I'm offering every time two months, completely no cost or commitment," he said, adding that after an introductory trial the service could be priced at roughly $100 per month if adopted.

Board members pressed for technical details. The student said the system produces monthly reports showing key questions and chat logs and that he had piloted the tool for other New Hampshire municipal sites. He acknowledged accuracy risks and said he includes a thumbs‑up/thumbs‑down user feedback option and an explicit disclaimer because "artificial intelligence ... can sometimes make mistakes." The student said he pre‑trains the model by manually extracting structured information from PDFs (setbacks, zoning information) before handoff to town staff.

Select Board members pointed to an upcoming 2027 requirement that website PDFs be accessible and asked whether the chatbot approach would meet accessibility and records‑management requirements. Staff noted that compliance with Web Content Accessibility Guidelines is a forthcoming legal requirement and that any vendor solution would need to demonstrate how it reads and cites source documents.

What’s next: The board discussed a potential trial later in the year and asked the presenter and staff to provide a clear onboarding timeline, sample reports and references from other municipalities that have used the tool.