November 11, 2026, (Inside AI) — Singapore Airlines has disclosed a set of production AI deployments built on OpenAI technology, spanning a customer-facing chatbot, a natural-language flight discovery tool, internal employee workflows, and software engineering. The airline reports a 2x increase in customer satisfaction after rebuilding its Kris chatbot on large language models, an approximately 10% reduction in average handling time for service representatives, and nearly 3x growth in monthly users of its Flight Recommender tool over the past year. More than 60% of employees now use the company's internal AI platform, Jarvis, on a monthly basis.
The results matter beyond one carrier. Airlines have spent a decade layering automation onto service operations with mixed returns, and the industry has watched chatbot projects stall when scripted flows failed to cover real customer intent. Singapore Airlines is now claiming measurable gains from a different architecture, and it is pairing those claims with governance language that enterprise buyers increasingly demand before signing off on AI spend.
Why Scripted Bots Hit A Ceiling
Kris previously ran on a traditional dialog-flow system. Every topic had to be mapped manually and every conversation path written in advance. Coverage was difficult to scale, and any missing intent pushed the customer toward a human agent even when self-service could have resolved the request. The airline moved Kris to a fully LLM-powered experience using the OpenAI API. The rebuilt bot interprets intent in natural language, grounds answers in a knowledge base, manages multi-topic conversations, and hands off to a representative with relevant context attached.
Customers now use Kris for general queries, special requests, real-time flight disruption information, and selected transactional workflows. The airline says it is expanding into booking changes, baggage transactions, and more complex agentic interactions. The disclosed satisfaction gain is notable because chatbot satisfaction scores often fall after launch, when novelty fades and edge cases accumulate. A sustained doubling, if it holds, would suggest the LLM layer solved a coverage problem rather than a demo problem.
Read: Northern Trust, Microsoft Deploy AI Agent to Cut Client Onboarding Time by 83%
The company's second customer-facing tool, Flight Recommender, lets travelers describe a trip in natural language instead of adjusting filters repeatedly. It runs on the OpenAI API with an orchestration layer connected to fare and availability data. The airline says it has implemented guardrails for input and output validation, malicious query handling, hallucination reduction, PII handling, and authenticated access limited to the context each query requires. A conversational version that lets customers refine ideas across multiple turns is in preparation.
The pattern here is worth watching across the sector. Airlines sit on dense proprietary data, from fare rules to disruption logs, and the competitive advantage may shift from model access to orchestration quality. Any carrier can call the same API. Fewer can wire it to live inventory, enforce access controls per query, and keep hallucination rates low enough for regulated customer communication. Singapore Airlines is effectively arguing that its moat is the plumbing, not the model.
Jarvis Puts Guardrails Around Employee AI
Internally, the airline routes employee AI use through Jarvis, a governed platform powered by OpenAI models. Staff work across PDFs, Word documents, Excel files, PowerPoint decks, and other formats in a secure environment, and they can build custom workflows and lightweight knowledge bases. Monthly adoption has passed 60% of the workforce and continues to climb.
In software development, teams use OpenAI models across the delivery lifecycle, not just for code generation. That includes requirements refinement, stakeholder-session analysis, product specifications, edge-case identification, test-case creation, architecture review, and task breakdown. Across selected activities, the airline reports productivity gains of up to 20%. For one commercial feature, the team used the API to challenge assumptions and surface ambiguous requirements before development began, then moved through design options, implementation plans, test cases, and code generation.
The shift changes what engineers do. Team members now direct, review, and validate AI output rather than producing every artifact themselves, reserving human judgment for prioritization, architecture, governance, and stakeholder outcomes. The airline says the workflows improve requirements quality, test coverage, and conformance with internal standards.
Singapore Airlines frames all of these deployments around a common control stack: orchestration, internal data controls, validation, and human review. That framing is as much a procurement signal as an engineering one. Regulated industries face pressure to show auditors how AI outputs are checked, and a named internal platform with usage metrics gives leadership something concrete to point to.
Read: Gilbert + Tobin CEO Sam Nickless Deploys Agentic AI in Legal Workflows
The airline says the next phase for Kris is deeper self-service with more transactional capability, while Flight Recommender moves toward multi-turn conversational discovery. Jarvis expansion across the business remains a priority. None of the disclosed figures have been independently audited, and the carrier has not published methodology for how it measures satisfaction or handling time. Those details will matter as rivals attempt to replicate the results.
Shamil has a low tolerance for AI that only works in controlled demos. He covers agentic systems and generative AI with the skepticism of someone who has seen too many “revolutionary” tools quietly disappear six months after launch.
