Call centers, technical support in casinos, AI assistants for all employees—the Barrière Group is deploying artificial intelligence on three fronts simultaneously.
The century-old luxury-hospitality operator, which operates 33 gaming properties and 20 hotels, is testing the approach with methodical rigor. Salomon Bentolila, Director of Data & Acquisition, leads the project with Artefact as the consulting partner.
A roadmap built around three pillars
His approach rests on three clearly defined goals. First pillar: boosting productivity and operations — AI is accessible to every employee, regardless of their level of technological proficiency, and applies across all business lines. Second pillar: transforming the customer and employee experience — AI-powered chatbots and assistance solutions enhance operational efficiency without compromising service excellence. Finally, the Group is experimenting with new business models to differentiate itself through generative AI.
He has thus developed a generative AI platform organized around three core agents, rolled out progressively in an iterative fashion.
Barrière GPT: democratizing access to AI
Based on Gemini, Barrière GPT provides secure access to an advanced language model. The platform includes libraries of prompts that teams can share and reuse. Deployed to about a hundred employees, the tool has already generated more than 4,000 prompts.
“We learn and improve rapidly by questioning things. Google integrates many features into its Workspace. That raises the question: should we maintain Barrière GPT or lean on native tools? We stay agile,” explains Salomon Bentolila.
This pilot phase enables real-world adoption to be assessed and the strategy adjusted in real time, rather than committing to a large-scale rollout.
Call Center Agent: centralizing information
Advisors previously had to consult multiple sources — procedures, product catalogs — wasting valuable time during customer interactions. The team developed an agent based on the Retrieval-Augmented Generation (RAG) architecture that centralizes all commercial documentation.
The results are compelling: more than 1,000 interactions generated by over 60 active users, with a satisfaction score of 3.64/4. Deployment is planned across all of the Group’s call centers.
Barrière Play Support: 24/7 assistance
Barrière Play is an app that lets customers connect to slot machines via a rechargeable digital wallet. Faced with technical issues that field teams did not always master, the Group built an integrated FAQ linked to a RAG system offering AI assistance 24/7.
Tested in three pilot casinos, it features more than 350 questions, over 30 active users, and a satisfaction score of 3.77/4.
User adoption, the key to success
“Our call center staff already have access to about thirty tools. Adding another one had to be worth it; the investment had to pay off,” notes Salomon Bentolila.
Field feedback prompted adjustments. For the call centers, teams had to improve the quality of the knowledge bases within the RAG and demonstrate tangible added value. For the Barrière Play agent, since the initial interface via Google Chat was not suitable for mobile staff, the Group integrated the agent directly into the management tools via API.
LLM as judge: ensuring quality
The system’s longevity rests on rigorous control. Artefact and Barrière Group have implemented a judge-expert framework to continuously assess the relevance of the responses. The methodology unfolds in four steps: first, building reference data libraries with annotated question/answer pairs; second, generating answers via the RAG agent to be evaluated; third, a legal expert automatically assigns a score from 1 to 4 to the prediction, with explicit justification; finally, reviewing the justifications and updating the agent/RAG context.
This continuous improvement cycle strengthens user trust and ensures long-term reliability.
With over 5,000 requests generated, the Group foresees new use cases in 2026. Agility remains the priority.
“Rather than following a rigid roadmap, we prioritize seizing opportunities likely to interest our teams. We set a budget and deploy it according to needs,” concludes Salomon Bentolila.
