Building AI for Travel at Scale: Strategy, Systems & Agents

Complementary lessons from internal productivity to customer-facing AI

AI is transforming the travel industry, but turning ideas into reliable, measurable systems in production remains hard—especially at scale. In this meetup, engineers and leaders from BlaBlaCar and AMEX GBT Egencia will share practical lessons on how they structure AI efforts, ship real systems, and operationalize agentic workflows—from internal platforms to customer-facing travel products.
While both companies tackle similar challenges, they’ll share perspectives from different starting points:

• BlaBlaCar will discuss its focus on internal productivity and developer enablement.
• AMEX GBT Egencia will share learnings from building customer-facing, knowledge-grounded GenAI agents.

Together, we’ll cover:

• how AI initiatives are structured, scaled, and measured for impact,
• how agentic AI is used for both internal productivity and external user experiences,
• what it takes to evaluate, orchestrate, and operate agents reliably in production.

Expect concrete examples, honest trade-offs, and shared lessons from teams approaching similar problems from different angles.

Format (Hybrid)
This will be a hybrid meetup. If you’re around the Paris area, we strongly encourage you to join in person for discussions, networking, and pizza. Remote attendance is also possible—RSVP to receive the Zoom link.

Practical Information

Date: March 31st, 2026

🕕 Schedule:

• 6:00 pm – Welcome & networking
• 6:30 pm – Talks start
• ± 9:00 pm – Pizza & drinks

📍 Location: AMEX GBT Egencia, 11 Chemin de Bretagne, 92130 Issy-les-Moulineaux (Agora conference room)
🎯 Audience: (ML) engineers, data scientists, AI enthusiasts, AI practitioners, leaders
🗣️ Language: English (questions welcome in English, French, and a few other languages)
👉 RSVP early — in-person spots are limited.

Agenda & Talks

Part 1 — AI Strategy: Organizing for Impact at Scale
BlaBlaCar — Yannick HOFFSESS
In the race to “AI-everything”, a lot of companies are burning cash on demos that never leave the lab. At BlaBlaCar, we took a different route: we are treating AI integration like our carpooling model — focused on efficiency and shared resources. In this talk I will go under the hood of BlaBlaCar AI strategy. This isn't about theoretical models. It is a blueprint of how we integrated AI in our practices. I’ll share our reality: the mindset that actually matters, the projects that delivered good ROI and the hurdles that are slowing down our journey.

AMEX GBT Egencia — Bernard CHEA, Véronique BLIN-ROYER
Great AI products don’t start with a model — they start with the fundamentals that make them reliable at scale. We’ll share how we structured and evolved our AI initiatives to move from experimentation to sustained impact: building strong data and AI platform foundations, adapting our approach for GenAI without reinventing everything, and partnering closely with Product to define success criteria upfront. We’ll also cover the governance and prioritization choices that helped us scale—and the lessons from what slowed us down.

Part 2 — Agentic AI: Internal and Customer-Facing Use Cases
BlaBlaCar — Mike BENOIT, Valentin FRIES
In this talk, we’ll first break down our journey in enabling engineering teams with a tailor-made multi-agent ecosystem. How to create an orchestrator agent? Are they good at introspection? How to get the most out of your trusty IDE plugin? We’ll share how we moved beyond standard Copilot usage to significantly increase developer experience.

We’ll then demonstrate how agentic tools can be leveraged internally to lift operational bottlenecks, like solving the flakiness of external test systems, assisting customer support in investigating escalations, or optimizing the cost of profile picture moderation. We’ll showcase real-world implementations that directly improve our teams’ quality of life and share how we track the tools’ efficiency to provide measurable business value.

AMEX GBT Egencia — Mohamed HAMHOUM, Riham MANSRI
We’ll share a production-grounded journey of building customer-facing GenAI support, starting with a multilingual RAG-based FAQ system deployed in 10 languages. We’ll do a candid retro on what it really takes to keep a RAG experience healthy once it’s live: defining success criteria, putting the right guardrails in place, building an evaluation approach that scales across languages, and maintaining quality over time through monitoring and iterative improvements—not just one-off demos.

We’ll then show how those learnings informed our move from Q&A to our first Flight Exchange AI agent in production. Using insights from real chat and live-agent sessions, we’ll explain why flight changes were the right next step, what changed when we introduced an agentic workflow, and how we scoped an initial v1 to deliver value safely while creating a feedback loop to expand capabilities.

Speakers:

Bernard CHEA is initially a full-stack software developer since 2005 and joined Egencia back in 2012, building the booking platform, then led the data platform and lastly the AI engineering and AI platform teams. His objective has always been to have a full end-to-end view on all aspects of the business to better serve our customers, asking the same to the teams he’s managing. Outside of work, he likes spending time with friends, a former bachata dancer – connecting to humans and not only computers.

Mika BENOIT works as a Software Engineer at BlaBlaCar since 2021. He joined the Booking squad, creating since then the backend serving the booking flow for all the buses and trains on the platform. During his free time, Mika likes to learn new maker skill, maintain a too complicated home automation setup and play Dungeon & Dragon.

Mohamed HAMHOUM is a Machine Learning Engineer at Amex GBT since 2024 passioned by GenAI. He has worked on turning structured search systems into intelligent, agent-driven assistants using RAG and Agentic architectures. He has also led efforts in evaluating and strengthening these systems, ensuring performance, stability, and reliability in high-stakes enterprise settings. Outside of AI, he's a big fan of board games and traveling, though he ironically forgets the names of most of the games he plays, despite being able to design agentic ai systems at scale.

Riham MANSRI is a Machine Learning Engineer at Amex GBT, where she develops AI solutions and machine learning models across a range of use cases since her final year R&D internship back in 2022 at Egencia. She enjoys working with advanced ML technologies and applying them to real-world problems. Most recently, her work has centred on Generative AI, spanning the design and evaluation of RAG systems, agentic AI, and the use of A2A to improve agent-to-agent communication and coordination. Outside of work, Riham enjoys movies, especially horror and thrillers, has grown into an F1 fan over the past few years, and is an enthusiastic bootcamp workout devotee.

Valentin FRIES joined BlaBlaCar as a Senior Software Engineer in 2022 and worked since then on bootstrapping and evolving the Global Distribution System enabling train sales on the platform. Prior to that, he worked for multiple years as Software Engineer in the travel industry, in education, and infrastructure. During his free time, Valentin practices Taï Chi Chuan and likes to start new hobbies — woodworking, bass guitar — that he won't be able to invest in since he has a newborn child.

Véronique BLIN-ROYER has been working in product management for over 15 years. She started her journey when agile practices were beginning to take root — around 2012, when agile was becoming mainstream. Since then, she had the opportunity to work across different product development teams, from front-end to back-end, always refining her craft as a product leader. Product management has always been a passion for her: “not only do I practice it daily, but I also teach it in a business school and create internal training programs to help others grow in this discipline”. This background has shaped how I approach AI Product Management today — blending rigor, agility, and a deep curiosity for how AI can transform products at scale.

Yannick HOFFSESS is the Head of Bus and Train Engineering at BlaBlaCar, a company he joined 4 years ago, bringing over 15 years of tech leadership experience from roles in the travel industry at Amadeus. Yannick balances his professional life with a passion for sports, van-life adventures with his family and needles.