Generative AI: feedback from the city of Saint-Genis-Laval

As local governments undergo modernization and digital transformation, the city of Saint-Genis-Laval is leading the way. As part of a complete overhaul of its website, the city has chosen to incorporate two innovative features: a citizen chatbot powered by generative AI and a “next-generation” search engine capable of understanding queries in natural language.

This project, led by the city of Saint-Genis-Laval in collaboration with Wikit, demonstrates a practical application of AI in local communities, providing residents with faster, more accurate, and more accessible responses.

In this interview, Jacky Béjean, deputy mayor in charge of digital affairs, offers a behind-the-scenes look at this ambitious project.

Could you introduce yourself and explain the background of the project?

My name is Jacky Béjean, Deputy Mayor of Saint-Genis-Laval in charge of digital services, gender equality, accessibility, and building safety. From the very beginning of our term, our goal has been clear: to simplify and modernize access to municipal services. The city had an outdated website and a fragmented ecosystem, with multiple accounts for each service (media library, cafeteria, after-school programs, etc.).

We have therefore launched a comprehensive redesign project, with the aim of centralizing procedures, information, and services within a unified citizen portal.

Why did you add a chatbot powered by generative AI to your website?

We were seeing more and more chatbots being developed in local communities, and we felt it made sense to offer one to our users. The goal was to make searches easier, provide clear, quick, and context-specific answers, and modernize the user experience with an intelligent conversational assistant.

How did the collaboration with Wikit go?

Serval Agency, the company we hired to develop the city’s new website, put us in touch with Wikit. From our very first discussions, I appreciated their technical expertise, their ability to tailor the solution to our needs, and the opportunity to test different AI models. We compared the results from Mistral and OpenAI and chose the latter for the quality of its responses. What reassured us was that the data is hosted in France, ensuring our digital sovereignty—a very important point for the city of Saint-Genis-Laval.

What issues did you hope to address with this project?

First and foremost, we need to modernize the interface and improve access to information. AI allows us to provide more relevant and concise answers, moving away from traditional lists of links. It also saves time for citizens, who can find what they’re looking for more quickly.

You’ve also integrated AI into the site’s search engine. Why did you make that choice?

This is a true innovation that emerged as the project progressed and through our discussions with Wikit. While AI chatbots are becoming increasingly common, few cities have taken the leap to implement an intelligent search engine, and we felt this was just as essential to improving the user experience.

A traditional search engine works in a very basic way: it simply identifies exact keywords within a website’s content and displays a raw list of results, without any real ranking or understanding of the context. Users then have to open several pages—often of little relevance—in the hope of finding the answer they’re looking for. It’s a very linear process that can quickly become frustrating.

In contrast, a search engine powered by generative AI uses natural language processing to understand the query as a whole, even when it is phrased in everyday language. It analyzes the user’s intent, selects the most relevant information, and presents it as a concise, clear, and contextualized response, often accompanied by useful links.

In practical terms, what is the added value of this AI integrated into the search engine?

For users, this changes everything: fewer clicks, less effort, and greater precision. They get a clear answer right from their first query, without having to navigate between multiple pages. This saves time, enhances convenience, and boosts efficiency, making access to information more intuitive and seamless—on par with the standards of major platforms—but with the added guarantees of data sovereignty and control that we consider essential for a local government.

Did you encounter any technical challenges on this project?

Yes, particularly the integration of administrative documents (decrees, resolutions, minutes). We had to adapt the AI-powered search engine so that it could read and extract information from PDFs, identify the correct documents, and display them clearly. The Wikit team successfully rose to this technical challenge!

What advice would you give to other cities considering launching a similar project?

Don’t hesitate! AI doesn’t replace humans—it assists them. It’s a complementary tool that enhances the user experience and takes the pressure off support teams when it comes to recurring questions. My advice: put serious effort into your data. An effective chatbot relies above all on clear, structured, and up-to-date content. And of course, you need to test, iterate, and adapt continuously.

Any final thoughts on Wikit?

Wikit has supported us with precision, patience, and flexibility. The team is attentive, has a thorough understanding of the challenges generative AI poses for local governments, and has successfully adapted the technology to our business and technical requirements. It has been a seamless and effective local collaboration. This project is innovative, ambitious, and scalable: it is a first step toward a more connected city that is closer to its residents.