Singapore Airlines doubles Kris chatbot satisfaction after moving to OpenAI models

OpenAI published a case study on November 10 detailing how Singapore Airlines uses its models across customer-facing services, an internal employee platform and software development. The headline result: after moving its Kris chatbot from a traditional dialog-flow system to an LLM-powered experience built on the OpenAI API, the airline saw a 2x increase in customer satisfaction.

George Wang, senior vice president of information technology at Singapore Airlines, said AI is central to delivering more personalized customer experiences and improving productivity across the business. Beyond the customer-facing API work, the airline runs OpenAI models inside Jarvis, its governed internal AI platform, and in its software development workflows.

Kris illustrates the shift. In the old design every topic had to be mapped manually and every flow written in advance, which made coverage hard to scale; if an intent was missing, the conversation was more likely to end up with a human agent. The LLM version interprets intent in natural language, grounds answers in a knowledge base, handles multi-topic conversations and hands over with context when needed.

Customers now use Kris for general queries, special requests, real-time flight disruption information and selected transactional workflows, with booking changes, baggage transactions and more complex agentic interactions in the works. Separately, AI-assisted tools such as call summarization cut average handling time for customer service representatives by roughly 10%.

On the travel discovery side, Flight Recommender lets customers describe a trip in natural language instead of adjusting filters, dates and destinations repeatedly; usage has grown nearly 3x over the past year. It uses an orchestration layer connected to fare and availability data, with guardrails for input and output validation, malicious queries, hallucination reduction and PII handling.

In software development, teams use the models for requirements refinement, stakeholder-session analysis, product specifications, edge-case identification, test-case creation and architecture review, with productivity gains of up to 20% across selected activities. For one commercial feature, the team used the API to challenge assumptions and surface ambiguous requirements before translating them into design options, implementation plans, tests and code.

Jarvis gives employees a governed way to use AI across PDFs, Word documents, Excel files and PowerPoint decks, and to build custom workflows and lightweight knowledge bases. More than 60% of Singapore Airlines employees now use Jarvis monthly, according to the company, and adoption is still increasing.

The significance is that this is not a pilot: models are embedded in customer service, sales, engineering and back-office workflows at production scale. What to watch next is whether Kris can take on more real transactions, whether these efficiency gains hold as usage grows, and how the airline balances autonomy with human review and data governance.