It’s not easy being a customer service representative, especially when those customers are so angry with a product that they want to yell at you on the phone. That’s the kind of wrath that Sonos, a maker of home audio systems, faced in May when it released an app update so riddled with bugs that its stock price plummeted.
One of the officers who dealt with the subsequent customer rants was a rookie. But not one human. Before the debacle, Sonos had hired Sierra, a startup co-founded by OpenAI chairman Bret Taylor, to provide the company with a customer service bot powered by generative artificial intelligence (AI). It could have been a disaster; the only thing worse than a defective product is being stuck in an automation prison by a robot that gives you the runaround. Still, the bot exceeded expectations. After digesting Sonos’ technical material, it came up with its own solution to one of the problems with the Sonos app.
Customer service is one of the few industries where the use of generative AI is already taking root. In a survey of customer service managers published earlier this year by research firm Gartner, nearly half said AI customer assistants would have a significant impact on their organizations over the next 12 to 18 months. Startups and established tech companies alike have launched a slew of new products in the industry that promise to transform customer service – and millions of jobs.
Customer service is a big industry. Most companies have some form of customer support, either in-house or outsourced to call centers. According to the Bureau of Labor Statistics, there are nearly three million customer service representatives in America alone. At an average salary of about $40,000 per year, that equates to about $120 billion in payroll costs. Many more people work in call centers in countries like India and the Philippines, where these jobs are seen as a ladder to the middle class.
However, in recent years, the industry has become infamous for driving customers crazy with the use of technology. The bad reputation is justified, says Andy Lee, co-founder of Alorica, an American contact center company with 100,000 employees. It’s expensive for companies to employ people to solve their customers’ problems, so they make the process as cumbersome as possible by forcing them to punch in a bewildering combination of numbers or chat with a bot that spits out generic answers – a strategy known as ‘deflection’. ”. Once human intermediaries are involved, it is in the financial interest of outsourcing companies to make the process as labor-intensive as possible, which increases costs and leaves everyone frustrated.
Now entrepreneurs and investors are betting that generative AI can make things less terrible. Funding for startups developing customer service tools that use generative AI reached $171 million globally in the third quarter, up from $45 million in the same period last year, according to PitchBook, a data aggregator (see chart). This month, Crescendo, co-founded by Alorica’s Mr Lee, raised funding worth $500 million. Sierra, which raised funds in January at a valuation of around $1 billion, is now said to be looking to raise more at a $4 billion valuation, raising eyebrows even among some AI-mad venture capitalists.
It’s not just startups entering the field. Tech titans like Alphabet, Amazon and Microsoft are bringing generative AI to their customer service offerings. Software companies like Salesforce are too. Some companies are also using large language models, such as those from OpenAI, to create their own customer service bots.
Unlike their rote-learning predecessors, generative AI bots don’t regurgitate canned answers to limited questions. Instead, they create their own answers based on the companies’ training materials and previous customer service interactions.
We apologize for any inconvenience
Providers are divided about the role these bots should play. Some advocate an approach where humans continue to manage customer conversations, but with an AI buddy in the background providing tips. But many others think that generative AI bots are now smart enough to handle most conversations on their own. This month, Twilio, another software company, announced it would launch a tool that allows customers to build a customer service bot that can listen and talk, rather than just read and type. A number of generative AI startups in the industry have adopted “outcome-based pricing,” where their technology is charged when a customer query is resolved, rather than per agent or minute of interaction, as is standard.
That raises two questions. One of them is how customers feel about all this. Proponents of the technology say customers will no longer have to wait endlessly for someone to answer the phone, pointing out that bots will be fluent in many languages and have easier-to-understand accents than foreign call center agents. However, customers still need to be convinced: 64% of Gartner respondents said they would prefer companies not use AI for customer service, mainly because they worry it will make it even harder to reach someone. People still value human contact, emphasizes Rob Goeller, co-founder of Clearsource, a customer service company based in Utah with employees in America, Costa Rica, India and the Philippines.
Furthermore, generative AI bots tend to exude complete confidence in their responses even when they are wrong, which can wreak havoc. Earlier this year, Air Canada was forced to compensate a customer who was wrongly promised a discount by the airline’s AI chatbot.
A second question is what all this means for the jobs of call center agents. Last year, Gartner predicted that generative AI would lead to a 20-30% reduction in customer service jobs by 2026. For now, Mr. Goeller says Clearsource is focused on using generative AI to train its human agents and help them summarize calls. But he adds: ‘I’d be burying my head in the sand if I said that [generative AI] would not replace people.”
Previous waves of customer service technology, including email and those pesky voice menus, fueled concerns about job losses but failed to materialize. AI could turn out to be different. And if it does, its effects can be beneficial. Human agents could be freed up to spend more time on creative and rewarding tasks, such as using feedback to make products better – and less time listening to irate customers.
Correction (October 18, 2024): The original version of this article incorrectly characterized Crescendo’s customer service tool as a tool that only worked in the background of conversations.
To stay up to date on the biggest stories in business and technology, sign up for the Bottom Line, our weekly subscriber-only newsletter.
