A call center.
Photo: Kuni Takahashi/Bloomberg via Getty Images

Contact centers have a language barrier problem — and AI could help

Protocol Enterprise

Hello and welcome to Protocol Enterprise! Today: What happens when your contact center employees don’t speak your customer’s language, data and more on Intel’s new acquisition.

Spin up

The chip shortage has pushed chip companies to their limit trying to satiate the world’s demand for computing power, so much so that sales rose to over a half-trillion dollars last year for the first time. According to the Semiconductor Industry Association, industry-wide sales climbed 26% to a record $555.9 billion.

How many languages does your contact center speak?

Large enterprises need to provide customer service around the world. But what if your contact center employees don’t speak your customer’s language?

Language barriers are a problem for contact centers. In an age of increasing globalization, businesses are being forced to reckon with the challenges of communicating across cultures, especially companies that sell software as a service.

  • It’s not merely a cultural problem, though. It’s an economic one. “If you're not able to support languages, you're actually limiting yourself,” because you can only serve a small fraction of the global market, said Vasco Pedro, co-founder and CEO of language platform Unbabel.
  • Contact centers — where conversations with customers happen in multiple other languages — feel these barriers most acutely.
  • Managing contact centers is already challenging due to major infrastructure requirements and large head counts. Throw in the complexities of multilingual support, and it’s clear why many companies fail to support multiple languages.
  • It’s also why customers are increasingly interested in software, AI and natural-language processing technology from Unbabel, Asapp, NICE and Google, among others, to bring multilingual support to global contact centers.

But executing a global-contact center strategy is far from easy. Companies need to balance the complexities of scaling infrastructure with customer demands for localization and native language support.

  • The solution thus far has been a complicated set of partnerships between companies, the cloud-contact software vendors that provide the technology and business-process outsourcers that handle physical operations.
  • Cloud contact leaders like Genesys have succeeded by growing these types of partnerships, said Steve Blood, VP analyst at Gartner. Because it's a multibillion-dollar company, "they have multiple offices in multiple countries, so their ability to recruit and retain partners is pretty strong in many countries around the world,” he said.
  • NICE — which has partnerships with Atos in Europe and TeleTech, ConvergeOne and RingCentral in North America — is another leader, though Blood said it’s not the level of Genesys.
  • Companies like Talkdesk and Five9, which Zoom attempted to buy last year for $14.7 billion, are also considered less competitive by Blood because they “haven’t matured their networks to the level that Genesys has.”
  • At the other end of the industry are outsourcing companies (BPOs), which can provide scale, infrastructure and expertise. But not all BPOs offer true multilanguage support.
  • Some BPOs in the U.S. claim to offer multilingual support. But when you talk to them, “98% of their workforce is English-speaking or in America, and they've got a couple of people in Sofia, Bulgaria,” said Edmund Ovington, head of customer experience for Unbabel.

Speaking to robots may be the answer. These problems are exactly what make global contact center strategy challenging, but also present enormous opportunities for software companies.

  • “There’s definitely going to be this next level of outsourcing also to bots that go on, but with the ability to escalate to a human,” said Barry Cooper, a president at NICE.
  • Companies like Google, Asapp and Unbabel are already using a mix of AI and natural-language processing to deploy bots, virtual agents and auto-translation services.
  • Google allows “companies to create a virtual agent once and deploy it in multiple languages,” said the company’s director of Product Management Yariv Adan.
  • Asapp is using NLP to offer real-time transcription of customer calls and Unbabel is using AI-powered translation to help agents reply to customers in their native language.
  • But despite advancements in NLP and AI, industry experts say many of these technologies are still a ways off.

But if contact centers can solve the challenges of language, they may define the next wave of globalization — for better or worse.

— Aisha Counts (email | twitter)


Dataiku is the only AI platform that connects data and doers, enabling anyone to transform data into real business results — from the mundane to the moonshot. Because AI can do so much, but there's no soul in the machine, only in front of it. Without you, it's just data.

Learn more

Banking on the lakehouse

Some say a data lakehouse is really nothing more than a data warehouse with a fresh coat of data lake capabilities sprucing it up. But as more so-called lakehouses hit the market, companies like Databricks want to stand out and speak directly to specific types of customers.

The data and AI company’s latest move to verticalize comes on Tuesday with the launch of a lakehouse geared toward the financial services industry, which includes several built-in tools for compliance, risk management and fraud detection purposes. The company said partners, including Avanade and Deloitte, would help sell the platform.

Databricks competitor Snowflake also offers a financial-services-aimed data cloud with tools for building fintech platforms and helping with regulatory compliance.

The banking lakehouse from Databricks comes on the heels of another industry-centric offering from the company. It launched its lakehouse for retailers in January, touting AI and real-time data management capabilities.

What’s the over-under on health care as its next lakehouse vertical?

– Kate Kaye (email| twitter)

Cool tech of the week: Go AI racer go!

Truly self-driving cars in the real world might be years away, but Sony has figured out how to get an AI to drive in the virtual one. In a recent paper in the journal Nature, Sony outlined how it has created an AI that is capable of playing the racing video game Gran Turismo. Sony said over two years that it was able to make an AI — called Gran Turismo Sophy — drive well enough to beat four of the “world’s best e-sports drivers.”

To competitively race in the game, an AI must be adept at controlling vehicles running at high speeds, which has its own set of challenges, Sony said. And beyond just ensuring a car can handle a track, the AI had to learn racing tactics, too. Those include methods to pass opponents, for example, or defend against an overtake attempt. The last component of the AI was getting the software to understand racing’s specific rules that include who gets blamed for a crash or potential time penalties for racing behavior.

Sony said that Sophy learns through “novel deep reinforcement learning techniques” and is trained on the company’s cloud gaming infrastructure. The company picked Gran Turismo because it is designed to simulate real-world racing conditions.

— Max A. Cherney (email | twitter)

Around the enterprise

Whoops! Yesterday we teased the Intel-Tower Semiconductor acquisition — then failed to include the story in the newsletter. Intel announced it was acquiring Tower early Tuesday for $5.4 billion. Here are all the details.

Enterprise spend on cloud services rose to $178 billion, or double the amount spent on its own data centers, CRN reports.

Google has opened early access to a new version of its Chrome OS called Chrome OS Flex designed to run on older PC and Mac hardware.

Content delivery network Akamai said it is acquiring infrastructure software provider Linode for about $900 million. The company also reported its earnings.


Dataiku is the only AI platform that connects data and doers, enabling anyone to transform data into real business results — from the mundane to the moonshot. Because AI can do so much, but there's no soul in the machine, only in front of it. Without you, it's just data.

Learn more

Thanks for reading — see you tomorrow!

Recent Issues