Customer support jobs could be some of the first to fall to AI.
So says Sam Altman, CEO at OpenAI, who made the claim on a podcast in September. When asked what jobs would be lost to the technology, he singled out call center agents at customer support companies.
“I'm confident that a lot of current customer support that happens over a phone or a computer, those people will lose their jobs, and that will be done better by an AI,” said Altman, in conversation with ex-Fox News host Tucker Carlson.
He is not alone in his prediction. Salesforce has slashed its customer support headcount from 9,000 to 5,000, revealed CEO Marc Benioff in September. He said that its army of Agentforce customer service bots should be handling around half of all conversations with customers by next year. However, he admitted that large language models can’t do everything – Agentforce currently only speaks English, with support for French, German, Italian, Portuguese, and Spanish in beta.
Nevertheless, a key focus for leaders looking to use AI to replace customer-facing roles is the potential to serve different regions without the need for interpreters and translators. Years after the likes of Google Translate began to wear down the language barrier, businesses are looking at using AI tools for more translation tasks.
Recent Microsoft research backs this up, with interpreters and translators listed among the top ten jobs at highest risk of being impacted by AI. But how useful will AI actually be for translation tasks and reaching customers in different regions?
Multilingual AI tools
One of the areas of customer support AI tools are promising to change is translation.
Language barriers can sometimes make it difficult for customer support employees to understand customers. This can lead to inaccurate information and details being recorded and customers not receiving the support they called up for.
In early October, Google rolled out Gemini Enterprise, a subscription bundle for businesses that includes video translation. In the big tech firm’s own words: “This goes beyond just words, capturing your natural tone and expression to make conversations seamless, no matter what language you speak.”
A couple of weeks later, Salesforce announced Agentforce Voice. The software add-on allows customers to process calls using AI, choosing their desired tone and speed of voices as well as adjust the pronunciation of specific words and phrases.
Former Salesforce CEO and OpenAI chairman Bret Taylor launched a similar service, Sierra, towards the end of last year, declaring “AI agents can now pick up the phone.”
The call center, the linguist, and the customer
Despite the ability of AI agents to answer calls and have conversations in multiple languages, there are some in the customer support industry who are hesitant about the technology.
John Campbell is the senior vice president of client services at Map Communications, an answer service provider used by businesses in more than a dozen industries across the US. He tells ITPro that AI translation tools in customer support can be “super helpful”, but warns that a “delicate balance” needs to be struck, to prevent damaging customer trust.
“Imagine an extremely disgruntled customer on the phone with an AI agent, and something gets lost in translation. Not only is their trust going to continue to be damaged, but you could risk losing that customer forever,” says Campbell.
There has been plenty of pushback on AI translation tools from the interpreting and translation community. The argument is that these tools struggle with ambiguity, idioms and cultural nuances that human translators wouldn’t miss.
A 2025 study, carried out by researchers at the Shahjalal University of Science and Technology and University of Oklahoma, found that AI translation ignores crucial cultural and linguistic context.
“Traditional machine translation systems, while effective in word-to-word conversion, often fail to retain cultural and historical depth, with up to 47% of contextual meaning lost in conventional translations,” the researchers wrote.
In 2023, The Guardian reported on complaints by translators over the machine translation tools used in the US immigration system. The paper cited cases in which asylum applications were denied due to “I” being mistakenly conveyed as “we”, or regional variations of languages such as Farsi Dari not being supported at all.
Applied in a business context, critics argue similar translation models may introduce unwanted errors into business processes, even carrying the potential to cause legal and reputational problems.
Annalisa Nash Fernandez, a cultural strategist and tech linguist, acknowledges the arguments over using AI for translation in a professional context, but offers up a counterpoint:
“Their lack of genuine reasoning and world-modeling means they can misinterpret context or overgeneralize where a human would stop and think. But the cons related to misunderstandings and cultural blunders could also happen with human agents.”
AI translation tools are “faster, cheaper, and always available,” adds Fernandez. “Companies can scale multilingual support and reach a more diverse customer base.”
Until AI translation tools are capable of understanding cultural context, some level of human oversight is going to be required.
Even if and when the tools have advanced to the point that there’s no longer need for oversight, the question is whether customers will want to speak to an AI agent anyway.
A recent survey of 1,011 US consumers conducted earlier this year by Kinsta found that almost half would cancel a service or subscription if they had to deal with AI-powered customer service, Meanwhile, 41.4% said that customer service has been made by worse by AI, while 41.3% would happily pay more to avoid having to deal with AI agents.
This data shows that customers still want a human-to-human connection, Roger Williams, partnerships and community manager at Kinsta, tells ITPro. “AI can mimic answers, but it can’t replicate the trust and growth that come from a truly human exchange.”
Companies looking to introduce AI translation tools need to be asking themselves why it is that they’re automating customer support. If it’s being done as a cost-cutting exercise, “customers will feel it – and not in a good way,” says Williams.
“The goal should always be to enhance the customer’s experience, not replace the human connection that makes support meaningful.”
