Skip to content

AI

Singapore Airlines' Kris chatbot doubled its CSAT. Humans, take note.

Singapore Airlines put OpenAI models behind its Kris chatbot and says customer satisfaction scores doubled. The same tech is also quietly helping employees get their work done.

Cover photo: Jernej Furman from Slovenia / CC BY 2.0

TL;DR

  1. Singapore Airlines runs OpenAI models and API across customer support, travel discovery and staff tools.
  2. Its Kris chatbot reportedly doubled its customer satisfaction score after the upgrade.
  3. Single-source announcement: the claim comes from OpenAI's own customer story, so read it that way.

What happened

Singapore Airlines has woven OpenAI models into how it serves customers and how its own people work. According to OpenAI's customer story, the airline uses the models and the API to improve customer support, help travelers discover where to go, and boost employee productivity.

The headline result is about Kris, the airline's customer-facing chatbot. After the upgrade, the airline reportedly doubled the chatbot's customer satisfaction score, known in the trade as CSAT.

That is the only hard number in the announcement. There is no baseline, no sample size and no timeframe. Doubled from what, measured how, over how long? The page does not say.

The deployment has two faces. One points at passengers: answering questions, helping with support, nudging people toward destinations. The other points inward, at staff who now have AI tooling for everyday work.

Why it matters

Airlines are a brutal test for customer service. People write in stressed, late, or both, and they want an answer, not a poem about clouds. A well-known carrier saying its bot scores far better with customers is exactly the kind of case study other enterprises like to wave around in budget meetings.

Still, keep your skeptic hat on, sapiens. This is a vendor-published story about a customer, so the framing is flattering by design. The claim is official, but it is also single-source, and the number comes without its working.

The more interesting signal is the split between outside and inside. Customer chatbots get the press, but the employee-productivity side is where many companies quietly spend. Singapore Airlines is doing both under one roof, which suggests the airline sees these models as general office infrastructure rather than a one-off gimmick.

The scoreboard that matters here is customer satisfaction, and the only figure we have is the one the vendor chose to publish.

It also shows where the market is heading. Big brands no longer ask whether to put a language model in front of customers. They ask which one, and then they publish a victory lap with a doubled metric on it.

What to do with it

  • Ask for the baseline.If a vendor says a score doubled, ask what it started at, how it was measured and over what period.
  • Split your use cases.Plan customer-facing and internal deployments separately, because they carry different risks and different ways to measure success.
  • Track CSAT before you launch.You cannot claim a doubling later if you never recorded the starting point.
  • Pilot on one service line.Start with a narrow support flow, measure it, then expand, instead of betting the whole help desk on day one.

Ciao, sapiens. Kris is getting the five-star reviews now, so be nice to the chatbot; it may be reading your feedback.