KLM Royal Dutch Airlines
conversational booking and service agent, hybrid AI/human
KLM has operated BB (BlueBot) since 2017, its conversational agent built with Google Dialogflow, which books flights on Messenger and Google Assistant and combines AI and 250 human agents in the same conversation.
Objective
Offer a conversational entry point to travelers who do not download the app, so they can book and get help where they already are (messaging apps and voice assistants), combining AI and human agents in a single conversation.
The deployment
BB (Blue Bot) is KLM's conversational agent, launched in September 2017 as a booking bot on Facebook Messenger. The traveler states a destination and dates, BB displays the options, and the purchase, personal details, and confirmation happen within the conversation. In December 2017, KLM added a luggage-packing service on Google Assistant. BB is built largely in-house with Google Dialogflow and connected to KLM's CRM, which lets a human agent take over when the bot can no longer keep up. KLM presents BB as a member of its service family, backed by its social teams that handle a high volume of messages each week.
Results Proof C
Official KLM press release (subject brand) plus a Google Cloud customer case study, two concordant official sources that name the project, its channels, and its actors. KLM did not publish a deflection or conversion rate in figures, so the results remain qualitative and about scale, which caps at C.
How it works
Documented architectureThe stack in detail
- plateforme Google Dialogflow natural language understanding (NLU) for BB, built largely in-house on this foundation
- infra Facebook Messenger et WhatsApp messaging channels where BB books and serves the traveler within the conversation thread
- infra Google Assistant voice channel added in December 2017 for the luggage-packing service
- infra CRM KLM connection that lets a human agent take over the same conversation when the bot can no longer keep up
How it runs, concretely
For ops teams-
1Conversational entry client
The traveler writes to BB on Messenger or WhatsApp, or queries it through Google Assistant.
-
2Understanding and search AI
BB recognizes the intent (book, pack luggage) and proposes the suitable options.
-
3Booking in the thread client
The traveler chooses, enters their details, and receives confirmation without leaving the conversation.
-
4Human takeover customer service
When BB can no longer respond, the CRM lets an agent take over the same conversation.
The quality of intent understanding (NLU) and the link to inventory and CRM. If the NLU does not recognize the request or the CRM does not keep up, human takeover becomes systematic and the channel loses its value.
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
- flight search engine and inventory via API
- connected CRM for human takeover
- corpus of intents to train the NLU
Org prerequisites
- social media team that runs the bot
- escalation procedure within the same conversation
- multilingual management
Possible stack
- NLU / conversational platform
- messaging integration (Messenger, WhatsApp)
- CRM connector
The plan, step by step
- Step 1Choose a simple purchase journey (direct booking) and build the corpus of intentsDeliverable: List of target intents and dialogues
- Step 2Build the NLU and the conversation flows on one messaging channelDeliverable: Working bot in sandbox
- Step 3Connect the inventory (search and booking API) and the CRM for human takeoverDeliverable: End-to-end booking within the conversation + tested escalation
- Step 4Open a beta on one market and one languageDeliverable: Automated conversation share and satisfaction measured
- Step 5Extend languages and channels with the social team as backupDeliverable: Multilingual bot in production with a well-run escalation procedure
First step: Connect a booking bot to a messaging app with human takeover through the CRM, on a simple purchase journey.
The same pattern, proven in other industries
Amazon
genAI shopping assistant
Burberry
genAI copilot for consultants (search, image recognition, complete-the-look)
Flipkart
conversational genAI shopping assistant
Sources
- S1 KLM welcomes BlueBot (BB) to its service family Primary archive pending
- S2 KLM builds booking and packing bot 'BB' with Dialogflow Interested party archive pending
- S3 Meet BB (that's short for BlueBot), KLM's smart assistant Primary 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.