KLM Customer Service Reps Avoid Turbulence in Social Media with AI Tool

by Tony Kontzer

Delays can be excruciating when you’re trying to change a flight by chat, figure out a return policy online or get a retailer to answer your questions via Twitter. But that could soon change.

A leading indicator: KLM Royal Dutch Airlines has become the first major consumer brand to deploy an AI-infused deep learning application. It will help customer service agents contend with the overwhelming volume of messages coming at them via social media and other channels.

DigitalGenius is the company behind the technology. The London-based firm built some of the very first customer service chatbots for companies like BMW, Panasonic and Unilever. And it’s  been developing an AI-powered customer service solution, using NVIDIA technology, for the past two years.

The application, also called DigitalGenius, draws from deep learning algorithms that enable it to ingest vast volumes of historical customer service data. This history is then integrated with customer service, consoles such as Salesforce Service Cloud or Zendesk.

“We’re unlocking the intelligence value of historical data while helping customer service agents deliver a faster and more accurate experience for their consumers,” said Mikhail Naumov, president and chief strategy officer at DigitalGenius.

Smarter, Faster Customer Service Responses

When a new message comes in via a digital channel such as email, chat, social media or text, DigitalGenius’ deep learning model takes a couple of actions:

  1. It predicts and auto-fills metadata related to the incoming message.
  2. It predicts the best response to the incoming message and shows it to the contact center agent for approval or personalization before sending it to the customer.

DigitalGenius screenshot

The result: huge time savings for customer service agents, who can instead focus on customers with more pressing or complicated needs. And it allows companies to maintain a high-touch approach to customer service amid deafening digital noise, even while opening new channels like Facebook Messenger.

This is critical for a company like KLM. The airline boasts more than 22 million social-media followers, who mention it on various platforms more than 100,000 times a week. KLM’s team of 235 social media service agents engage in 15,000 conversations a week across all its social platforms, offering 24/7 service in 10 languages.

“A personal approach is extremely important to KLM as this is what defines our social media service,” wrote Tjalling Smit, vice president of digital for the carrier, in a post on the company’s website. “Applying AI, KLM can handle a greater volume of questions while still maintaining its personal approach and speed.”

The GPU Effect

DigitalGenius uses NVIDIA TITAN X GPUs to train its deep learning neural networks on a company’s historical data. Production workloads for enterprise customers run on NVIDIA GPUs in the AWS cloud, with the CUDA parallel computing platform providing acceleration. Naumov said the rapid increase in the collection of historical customer data and advances in NVIDIA hardware have combined to make AI-powered customer service practical for the first time.

“NVIDIA GPUs make it possible for us to train very large neural nets with millions of parameters in a matter of hours rather than days,” Naumov said.

KLM and the airline industry are well positioned to use AI products like DigitalGenius. But Naumov sees potential in any industry that has large or rapidly growing volumes of customer service conversations, including telecom, utilities, consumer products and services, and transportation.

“There is just no way to keep up with such rapid growth in message volume without reinforcing your contact center with a layer of AI,” Naumov said.

Subscribe to NVIDIA’s AI Podcast on iTunes or Google Play Music.