Lufthansa Group
multilingual customer service conversational agent
The Lufthansa group handles 16 million conversations per year on the Cognigy platform, with peaks of up to 375,000 interactions per day, for rebooking and multilingual customer service.
Key points
- Multilingual customer service conversational agents across the Lufthansa hub airlines.
- Cognigy.AI platform, NLU, real-time translation, chat, SMS and voice channels.
- 16 million conversations per year, peaks of up to 375,000 interactions per day.
- Evidence B, confirmed status: Cognigy case study and interview with the product manager.
Objective
Replace a rigid internal chatbot, overwhelmed by demand spikes during and after covid, with a platform able to absorb millions of conversations in several languages and free up agents for complex cases.
The deployment
The Lufthansa group runs its digital assistants on the Cognigy platform for the hub airlines (Lufthansa, Austrian Airlines, SWISS). A customer can ask a question on the site, in live chat, by SMS or by voice, in their language, and the agent handles common tasks: rebooking, travel information, alternative flight search, refund request. Real-time translation makes it possible to serve an international audience on a single foundation. When the request goes beyond what can be automated, it is handed off to a human advisor.
Results Proof B
Quantified Cognigy platform case study (annual volume and daily peak), complemented by a public interview with the Lufthansa product manager (VUX World) detailing the chat and voice scale. Hard figures on the vendor side, corroborated by the executive's statements.
How it works
Documented architectureThe stack in detail
- plateforme Cognigy.AI enterprise conversational AI platform (NLU, dialog orchestration) that powers the hub airlines' assistants, 16 million conversations per year
- outil Canaux chat, SMS et voix Cognigy the same conversational foundation exposed in live chat (about 10M interactions per year), SMS and voice (over 1M per year)
- outil Traduction temps reel makes it possible to serve a multilingual international audience with a single dialog foundation
- infra Systemes de reservation et statut vol Lufthansa real-time access to booking records, flight status and inventory for rebooking; the critical signal during a crisis
How it runs, concretely
For ops teams-
1Multichannel entry customer
The customer writes or speaks on the site, in live chat, by SMS or by voice, in their language.
-
2Understanding and translation AI
The agent identifies the intent and applies real-time translation for a single multilingual foundation.
-
3Request handling AI
Rebooking, travel information, alternative flight search or refund are executed or prepared.
-
4Human escalation AI
Complex or sensitive cases are passed to an advisor with the conversation context.
The customer's intent crossed with the state of their booking and the availability of alternative flights. If the inventory and flight-status feed degrades during a crisis, the rebooking proposals become wrong just as the volume explodes.
How your customers perceive this type of use
Sourced studiesLes consommateurs n'acceptent pas les chatbots par defaut : 64% prefereraient que les entreprises n'utilisent pas d'IA dans leur service client (Gartner, 2024) et pres d'un utilisateur sur cinq du service client par IA n'en retire aucun benefice (Qualtrics, 2025). L'acceptation se construit sur trois conditions mesurees par Salesforce : savoir qu'on parle a une IA, pouvoir escalader vers un humain, comprendre la logique de l'agent.
Acceptance conditions
- Etre informe qu'on parle a une IA et non a un humain (pres de 75% le demandent, Salesforce 2024)
- Un chemin d'escalade clair vers un agent humain (45% plus enclins a utiliser l'agent IA, Salesforce 2024)
- Une logique de l'agent clairement expliquee (44% plus enclins, Salesforce 2024)
Red lines
- Rendre l'humain injoignable : c'est la premiere inquietude des consommateurs sur l'IA dans le service client (Gartner 2024) et 50% craignent que l'IA les coupe du contact humain (Qualtrics 2025)
- Remplacer le service client par l'IA sans alternative : 53% envisageraient de partir chez un concurrent (Gartner 2024)
Sources: Salesforce 2024 · Gartner 2024 · Qualtrics 2025
How to replicate
Inference, not sourcedData prerequisites
- real-time access to flight status and inventory
- access to booking records
- multilingual knowledge base
Org prerequisites
- product team dedicated to the assistants
- escalation process to advisors
- ability to hold the load during crisis peaks
Possible stack
- enterprise conversational AI platform
- real-time translation component
- connectors to booking systems
The plan, step by step
- Step 1Frame the priority use cases (rebooking, travel information, refund) on one channel and one languageDeliverable: v1 scope + targeted conversational journeys
- Step 2Connect the booking systems, flight status and inventoryDeliverable: Real-time connectors tested, including in degraded mode
- Step 3Build and launch the agent in live chat with escalation to an advisor, conversation context passed alongDeliverable: v1 agent in production on one channel, resolution rate measured
- Step 4Extend languages via real-time translation and add SMS then voiceDeliverable: Multilingual multichannel foundation in production
- Step 5Test load handling on peaks (weather, strikes) and industrialize measurementDeliverable: Capacity plan + volume, resolution, AHT dashboard
First step: Deploy an agent on one channel (live chat) and one language for rebooking, then extend to the other languages and channels.
The same pattern, proven in other industries
ADT
genAI customer agent
AT&T
Governed enterprise genAI assistant: an LLM connected to internal knowledge, open to tens of thousands of employees and called at scale by automated systems
Bank of America
conversational financial assistant at scale
Sources
- S1 Lufthansa - Cognigy Case Study Interested party archive pending
- S2 The AI chatbot serving 10m customers a year, with Nick Allgaier, Lufthansa Secondary archive pending
An error, newer info, a source?
This page lives on its accuracy. If a figure has moved, if the deployment has changed, or if you have a higher-quality source, tell us. Every sourced correction is verified before publication.