Introduction
Cork County Council (CCC) is the second largest local government body in Ireland, servicing a population of over 360,000 people across 8 municipal districts comprising hundreds of towns and villages. They support local communities and businesses in Cork through a wide variety of services, with expertise in Roads, Housing, Community Development, Commerce, and Tourism.
One of CCC’s most engaged support functions is housing, where they work to facilitate the provision of suitable, quality and cost-effective accommodation services across the county. Through updating their housing support services, CCC recognised an opportunity to further digitalise their online engagement process, simplifying access to information for residents with specific housing queries.
Challenges
CCC approached CeADAR through the EDIH for AI programme, with a goal to prototype a ChatBot for their housing department. The challenge they faced at the time revolved around managing large volumes of queries relating to new affordable housing schemes. They hoped to develop a chatbot that could assist citizens find the information they needed in a more efficient way.
The initial end goal for CCC was to provide users referencing their website with a simple-to-use chatbot interface that could answer queries related to affordable housing schemes. It would provide citizens with a straightforward User Experience (UX) to retrieve and summarize materials through commonly asked questions such as:
● “What schemes are currently open?”,
● “How do I apply for x development?”,
● “What documents do I need to provide?”
Solution Developed
After an initial consultancy phase, CeADAR and CCC agreed to develop a Retrieval Augmented Generation (RAG) ChatBot prototype, which would operate as a free-to-access tool on the CCC website. It would be built using a basic LLM architecture, with the source documents used to direct the prototype’s answers consisting of a variety of housing scheme and policy documents and pages situated across the CCC website.
Over the course of the project the initial idea of a chatbot for affordable housing schemes evolved into a tool designed to help citizens access information on a wider range of services, assisting them to navigate the CCC website to find the information they needed.
Listed below are a number of key system features that form the basis of the prototype:
● A front-end chat window embedded into CCC’s digital environment.
● Service selection buttons (e.g. Housing)
● Optional quick-reply buttons for common topics (e.g. “Current schemes”, “Civic amenity sites”, “Renewing motor tax”).
● Free-text questions and follow-ups in natural language.
● Voice interaction to enable hands-free use and improve accessibility for a wider range of users.
● An answer panel that always links to the source for transparency.
Cork County Council RAG Bot Prototype UX Design
This hybrid design would allow both novice and confident users to find what they need quickly, whether they prefer clicking buttons, typing, or speaking. The CeADAR team have also worked on enhancing the existing RAG bot since its initial development, with the latest improvement incorporating a fully automated data ingestion pipeline and expanding the service options to Motor Tax and Waste services.
This new workflow seamlessly connects a number of cloud-based apps with backend APIs to ensure all source content changes underpinning the RAG ChatBot are instantly reflected in its knowledge base, eliminating manual intervention and enabling dynamic, real time updates in the user interface.
Results and Benefits
The goal of this project was to demonstrate how a well developed RAG prototype could benefit CCC’s citizens. It had a dual benefit of showcasing how AI integrations can also benefit CCC staff.
By the end of development, the CCC prototype had evolved into an intelligent, multi-service RAG platform, with a detailed understanding of how to:
- Answer Natural Language Queries.
- Route queries accurately across a range of service functions
- Return concise, cited answers from both official PDFs and live website content.
If this prototype is transitioned to production, it will help to streamline the engagement process for citizens looking for information, . It should also help to reduce the volume of queries received by the CCC support teams. This prototype has the ability to enhance the engagement process across CCC’s entire community support remit, putting people in touch with essential service information.