Singapore Airlines has adopted OpenAI API and models to enhance customer support, travel discovery and employee productivity. The carrier runs OpenAI models in customer‑facing features and inside Jarvis, its governed internal AI platform for staff, and also uses the models across parts of its software development lifecycle.
Key measured impacts
- Customer satisfaction score (CSAT) for the Kris chatbot doubled after migrating from a traditional dialog‑flow system to a large‑language‑model (LLM) based experience.
- Average handling time for customer service representatives fell by about 10% after introducing AI‑assisted tools such as OpenAI API‑powered call summarization.
- Monthly users of Flight Recommender, the natural‑language travel discovery tool, increased nearly threefold over the past year.
Kris chatbot: from predefined flows to LLM understanding
Previously, the chatbot relied on manually mapped topics and predefined conversation flows, which limited scale and coverage: missing intents often led to unnecessary handovers to live agents. Singapore Airlines rebuilt Kris as a fully LLM‑powered conversational agent using the OpenAI API. The new chatbot interprets customer intent expressed in natural language, grounds answers in a knowledge base, manages multi‑topic conversations, and hands customers over to agents with relevant context when needed.
Customers now use Kris for general inquiries, special requests, real‑time disruption information and selected transactional flows. The airline is expanding Kris to handle booking changes, baggage transactions and more complex agentic interactions to increase self‑service completion.
Flight Recommender: describe your trip instead of toggling filters
Flight Recommender lets customers describe the trip they want in natural language instead of repeatedly adjusting filters, dates and destinations. Built on the OpenAI API, the tool uses an orchestration layer connected to fare and availability data to interpret intent, invoke the appropriate tools and return personalized recommendations in natural language.
The team added guardrails for input and output validation, handling of malicious or inappropriate queries, reduction of hallucinations, PII handling and authenticated access to only the context required for each query. The airline is preparing to evolve Flight Recommender into a more conversational experience in which customers can iteratively refine travel ideas without restarting the search.
AI across the software development lifecycle
Singapore Airlines also applies OpenAI models to planning, refinement, testing and review activities in software development. Selected lifecycle activities delivered up to 20% productivity gains.
Generative AI supports requirements refinement, stakeholder session analysis, product specifications, edge‑case identification, test‑case generation, architecture review and decomposition of complex work into smaller, testable tasks. For one commercial feature, the team used the OpenAI API to challenge assumptions, surface ambiguous requirements and identify edge cases before development, then translated those into design options, implementation plans, test cases and code.
During deployment and production monitoring, AI helps interpret system behaviour and flag issues that might otherwise go unnoticed. With AI involved at each stage, teams shifted from producing every artifact themselves to directing, reviewing and validating AI outputs — freeing people to focus on prioritization, architecture, governance and stakeholder outcomes.
Jarvis: a governed internal AI platform
Jarvis provides employees with a governed way to use OpenAI models day‑to‑day, from summarizing documents to creating custom workflows. Staff can work across PDFs, Word, Excel and PowerPoint files in a secure environment while building lightweight knowledge bases and bespoke workflows.
More than 60% of Singapore Airlines employees use Jarvis monthly today, and adoption continues to grow.
Next steps and significance
Singapore Airlines plans to expand Kris’s self‑service and transactional capabilities, and to make Flight Recommender a more conversational, multi‑turn discovery tool. With Jarvis rolling out across the business, the airline aims to enable more employees to apply AI in daily work.
Taken together, these initiatives are intended to make travel smoother for customers, increase employee productivity, and help the business respond more quickly to changing needs.
Summary of measurable outcomes
- 2x increase in CSAT for Kris
- ~10% reduction in average handling time for customer service agents
- ~3x growth in monthly Flight Recommender users over the past year
Singapore Airlines’ deployment pairs OpenAI models with orchestration, data controls, validation and human review to bring AI into production workflows responsibly and deliver measurable improvements in service and internal efficiency.
