SAILing into the future: chatbot helps small municipalities navigate resident requests

Chatbots are rapidly reshaping the way we communicate with businesses, blending AI and instant messaging to deliver customer service, tech support and even therapy – without a human on the other end. Two entrepreneurs from Tiffin are hoping to bring that same technology to smaller, local municipalities. Friends of eight years Braden Smith, a data and analytics professional, and Brandon Laidig, a software designer, have created a chatbot designed specifically to help small cities and organizations manage resident requests and streamline services. Their creation – SAIL Assistant – aims to give under-resourced municipalities a digital assistant of their own, capable of answering questions about everything from trash pickup schedules to permit applications, and in several different languages. “Initially, it was some of my own frustrations with local websites, trying to find some information,” Mr. Smith said, of the inspiration behind SAIL. “I noticed that on social media, some people were asking similar questions repeatedly,” he added, realizing that smaller cities and organizations were fielding repetitive questions to issues that should be uncomplicated to find independently. SAIL, a mashup of Mr. Smith and Mr. Laidig’s surnames, SAIL uses OpenAI's large language model (LLM) via its application programming interface (API), and tailors the system for local government and small organizations. “I wanted the smaller towns and smaller organizations to have access to an MVP (minimum viable product) solution for this,” Mr. Smith said. “I was hoping that rather than maybe hiring an additional person or being able to redirect some efforts to more high-touch community or organization engagement efforts, that this could handle maybe some of low-level things and make people's days a little better,” adding that it’s tailored to be an affordable and tech-easy solution to understaffing.

How it works

The process begins with SAIL scanning an organization’s public website for important pages. “When someone asks a question, we figure out which pages are most relevant, pull in that information in real time and send it - along with the question – to OpenAI's model,” Mr. Smith said. “This ensures the answer is based on the most up-to-date public information. We also include guide rails in our prompts to keep responses accurate, consistent and grounded in facts.”

SAIL’s system indexes website links found on a city’s website and generates embeddings for them. When a user submits a query, it creates an embedding for the query, compares it to the indexed embeddings, and retrieves the top five matching links. It then fetches the content from those pages and uses it, along with a prompt, to generate an answer to the user's question. “Importantly, OpenAI doesn't use any of the data sent via their API to train their own models, and we never send non-public information,” Mr. Smith said. SAIL does save the conversation history, as it can act as a valuable tool for gauging the most frequently requested information. “We could also do things like sentiment analysis (with conversation history), which would score people's responses,” Mr. Smith added. SAIL Assistant can do more than answer simple questions, however. If a city indexes its council minutes, the system can give users a quick summary of the meeting on any given date. “It can also answer questions like, 'who's on the council? What is the council's contact information?'” Mr. Smith said, adding that it can also help users draft emails to city leaders.

A partner to pilot

Currently, Mr. Smith and Mr. Laidig are searching for partner organizations to pilot the program. Most municipalities and organizations the entrepreneurs have approached have expressed trepidation surrounding AI, Mr. Smith said. Potential partners would benefit from a reduced charge since it would be a trial period, and there would be no setup fees, since “(it) would be kind of a learning experience for us as well,” Mr. Smith said. Part of the learning experience has involved learning to think from a “product mindset,” he said. “Lots of times with my job in data and analytics, we get requirements (and) we fulfill those requirements, but this is more of trying to solve a pain point and then making sure that we're considering marketing,” he added. To learn more or see a demo of SAIL Assistant, visit sailassistant.com.