How KLM uses artificial intelligence in customer service


For better or worse, airlines have treated social media as a critical customer service tool, catering to cranky travelers who tweet their gripes. That willingness has led to a deluge of issues for customer service to address — and created the need to automate it.

Take Dutch airline KLM. In a typical week, KLM has to respond to 15,000 social conversations in a dozen different languages. There is no programmatic solution here: KLM has a 235-person social media team. But KLM is exploring ways to combine artificial intelligence and humans in providing customer service that’s somewhat automated but still has a personal touch.

In the most basic form, KLM added in a Facebook chatbot in March. KLM uses Facebook for automated updates around checking in and flight delays and also gives users a copy of their boarding pass. When there’s a bigger ask, a (human) KLM agent will jump in. In its first month, KLM’s volume of Facebook messages jumped 40 percent. KLM said 1.7 million messages have been sent on Messenger by over 500,000 people.  

“Social media is about having a timely and correct answer,” said Tjalling Smit, the airline’s senior vice president of digital. According to Smit, the time users are willing to wait for a response is dwindling, often becoming impatient in 20 minutes or less.

KLM is trying to find new ways to combine machines with humans in tackling the deluge of social media inquiries. KLM’s customer service agents on social media are given advice from a machine learning system created by tech company DigitalGenuis that has mastered over 60,000 KLM customer questions. It will get smarter over time as it monitors the social media team’s work. After taking in suggestions from the platform, the agents can then edit this text and hit send. “Right now 10 percent of responses are sent without any alteration,” Smit said. “But in time we’ll see more pure AI conversations.” 

But Smit emphasized that while AI is good for “functional” answers, the human touch is still important. “People want a personalized answer, for example dealing with compliments is hard for an AI system,” he said. So far the system has increased the speed of replies by 20 percent.  

“People do say that over time our tone of voice has become more functional,” Smit said. “But equally, they do appreciate it because they come to us to get a problem solved.”

KLM is now exploring adding more functions to Facebook Messenger and expanding its chatbot to other platforms like WeChat. These are expected to roll out within the next year. Integration with “conversational” services like Amazon Alexa is also a possibility. While its social teams were the most immediate need, Smit says AI has a clear use case for personalizing customers’ web and call centre experience. 

“Robotics has a cold feel to it, people think it’s about efficiency, automation and having less people. But we don’t see it as a substitute for human interaction, just a way to improve it.”

More in Marketing

What TikTok’s e-commerce launch could mean for marketers and content creators

TikTok has officially launched its new e-commerce platform, TikTok Shop, earlier this month on August 1. Using the new e-commerce platform, brands and creators can sell products directly on the platform, potentially creating new revenue streams, and tap into the short-form video platform’s growing popularity.

‘The influencer industry can be really vile’: Confessions of an influencer marketer on the industry’s unfair hiring practices

While the influencer industry might sound exciting and like it’s full of opportunities, one marketer can vouch for the horrific scenarios that still take place behind the scenes.

Digiday+ Research: Marketers said revenue grew in the last year, with more growth expected ahead

After a tumultuous 12 months, marketers are getting a clear picture of how they really did during a time of true uncertainty. And, as it turns out, it wasn’t all that bad.