|
Welcome back! As the cost of running artificial intelligence models falls, the land grab in customer support automation has reached a new level of intensity. Nearly every major enterprise software vendor now offers a product for automating customer support, according to our analysis, which also breaks down how the race to automate knowledge work has bred new turf wars among these firms. Against this backdrop, a one-year-old startup, Crescendo, is taking a different tack: buying up contact center companies and using AI models such as OpenAI’s GPT-4o to lower their costs. The goal is to make it dead simple for businesses to automate customer support by not requiring them to switch away from traditional customer support vendors or try to create AI-powered services themselves, co-founder Anand Chandrasekaran said. “The biggest value is that we’re removing the risk from the enterprises’ hands,” he told me. That could assuage the fears of some chief information officers who are concerned about the expense and risk of automating customer support with AI—even though the technology could lower the cost per interaction by as much as 95%, as it did for fitness company ClassPass. After refining its product with some early customers, Crescendo in October bought PartnerHero, an outsourcing company founded in 2014 that employs thousands of contact center staff in the Philippines, Honduras and the U.S. The acquisition, whose price Chandrasekaran declined to share, brought in around 200 new customers, most of whom are based in the U.S. Crescendo then implemented a system that uses AI to resolve customer calls, emails and text conversations. That reduced how many interactions are sent to PartnerHero’s staff by more than 70%, Chandrasekaran said. Despite fewer inquiries going to human workers, Chandrasekaran said Crescendo didn’t lay off any PartnerHero staff because they’re still needed. “Our business is growing like crazy,” he said. Crescendo plans to buy other such outsourcing firms, Chandrasekaran said. Notably, Crescendo uses an outcome-based pricing model in which customers pay only when the AI delivers a result, like avoiding the need for a human to take the call. Chandrasekaran wouldn’t specify what Crescendo charges for each outcome. (For more on the outcome-based model, see my story on the subject.) Crescendo was incubated at venture capital firm General Catalyst and has raised a total of $50 million. It has generated more than $50 million in revenue since launching last year, Chandrasekaran said. Customers opting for outcome-based pricing pay the same rate regardless of whether AI or a human resolves the issue, though Chandrasekaran didn’t specify how much of its revenue comes from each source. Crescendo’s acquisition strategy means “we don’t have to go [to customers] one by one by one and sell the product.” Customers include Udemy, Grindr, Miro and Sweetgreen, he said. Crescendo’s playbook could be a model for other AI companies looking to grow through acquisitions. It bears resemblance to the practice of some private-equity firms that buy low-tech businesses like plumber or dentistry offices and use software to lower their costs. VC firms have increasingly been investing in or buying up mature businesses to turn them around using the same methods, which now include AI. Klarna, a lending firm that refers to itself as OpenAI’s “favorite guinea pig,” is walking back some of its bold claims about AI’s ability to replace human workers. A little over a year ago, the Swedish firm made headlines for weeks when it said in a blog post that an OpenAI-powered chatbot had taken over two-thirds of customer support interactions human contractors previously handled. The company declined to give interviews about its claims. Then this week, CEO Sebastian Siemiatkowski said the company is changing course and will look to hire more human customer support employees, citing customers’ expectations that “there will always be a human if you want.” Klarna’s about-face raises questions about just how cost-effective automation will prove for enterprises. The company disclosed a few months ago that the AI customer support chatbot had helped it save around $39 million in 2024. Maybe Siemiatkowski determined that lost business from frustrated customers who were unable to contact a human outweighed those relatively modest savings. Still, we’ve been hearing about CEOs across the business world using AI to automate roles and hire fewer humans. Some seem to have been more successful: PayPal, another fintech firm, recently told me it has automated 80% of customer support interactions and will save around $200 million annually from its use of AI. Those savings are robust enough to seem worth the risks, even if not every customer with a service complaint is happy to be talking to a robot.—Aaron Holmes | 
