Testimonial: a look back at the deployment of INSA Lyon's AI chatbot

INSA Lyon, a leading engineering school, called on Wikit to deploy a chatbot integrating generative AI on its website dedicated to future students: www.rejoindreinsalyon.com. Ophélie Tambuzzo, Director of Communications, talks about the implementation of this chatbot and its benefits.

Pouvez-vous présenter l’INSA Lyon en quelques lignes ?

For over 60 years, the Institut National des Sciences Appliquées de Lyon has been training humanist engineers to meet the socio-ecological and digital challenges of a changing world. Each year, INSA Lyon welcomes over 6,000 students and 600 PhD students, and graduates over 1,000 engineers and 150 PhDs. As a reference center, the school is also developing excellent, responsible and supportive research, based on 22 laboratories. INSA Lyon is part of the INSA Group, the leading network of French public engineering schools, which currently comprises 7 establishments and 6 partner schools in France.

Quelle(s) problématique(s) souhaitiez-vous adresser en mettant en place un chatbot ?

As part of the expansion of our website, www.rejoindreinsalyon.com, we were looking for solutions to improve the experience of high school and university students wishing to join us, and to enable interactivity with this target group.

The chatbot was one of them, facilitating access to the wealth of information available on the site and offering a new response channel to the many requests we receive, particularly during the opening period of the Parcoursup platform. Another advantage is that the chatbot offers a fun, differentiating experience for our main target of high school students.

Pourquoi votre choix s’est porté sur Wikit ?

Our tight schedule for implementing the tool meant that we had to quickly rule out solutions based on conversational trees, which would have required a heavy investment on the part of INSA. Generative AI seemed a good compromise.

Après avoir étudié plusieurs propositions, notre choix s’est porté sur Wikit pour plusieurs raisons : la pertinence de l’offre au regard de notre cahier des charges, le respect du calendrier, l’accompagnement pour la mise en œuvre, le coût et la proximité géographique.

Quelles ont été les étapes du déploiement de votre chatbot ?

Nous avons travaillé dans un premier temps à la définition du périmètre et des éléments de personnalisation du chatbot tant d’un point textuel que graphique. Puis ce fut la phase « jeu de tests » avec pour objectif d’avoir au moins 50 questions avec la source de la réponse et la réponse attendue. Nous avons fonctionné par itération, afin d’affiner progressivement les réponses du chatbot. Cela a pris un peu de temps et des ajustements ont été nécessaires. L’INSA Lyon faisant partie d’un groupe d’écoles ; certaines réponses ne s’appliquaient pas à notre établissement. Cela nous a permis d’identifier que certaines sources de données devaient être retirées du périmètre.

Finally came the pre-production acceptance phase before going live.

We met weekly to check on the progress of the project, discuss any difficulties encountered and find solutions. The project ran from mid-November to the end of January.

Pour vous, quel est l’apport principal de l’IA générative pour le chatbot ?

The main benefit of using generative AI for chatbots lies in their ease of implementation. AI is capable of processing very large amounts of data automatically, enabling rapid deployment with 24/7 availability.

Be careful, however, to remain vigilant and regularly monitor conversations to identify unanswered questions or incorrect answers.

The introduction of a "reflex sheet" provides a quick and effective solution to recurring questions and/or questions that are not satisfactorily answered.

Quel(s) bénéfice(s) constatez-vous suite à la mise en place du chatbot ?

Our team was able to see for themselves just how useful the tool can be at an event. When we organized the school's Discovery Day on Saturday, February 3, it was important to ensure that everything was clear about the day's program, especially as we had both face-to-face and distance learning registrants. The chatbot was thus able to answer practical questions about registration, access, conferences and replays , and significantly reduced the number of requests we would have received by e-mail or telephone.

Don't miss our next resources