ET
AI has become an engine for customer engagement. Illustration: Thomas R. Lechleiter/WSJ
AI agents are poised to make communication between businesses and their customers deeper, more open and more personal. That level of intimacy might be great, but as in any relationship, it creates the potential for hazard, too.
The latest generation of AI agents are capable of taking action on behalf of people in a range of functions, including coding, customer service, legal services and booking an appointment with a healthcare provider. Their migration into popular messaging services such as WhatsApp opens up even more opportunities for businesses in the areas of sales, support, marketing and branding.
There is a flywheel effect at work here. The AI agent has access to an enormous amount of data about users that makes it possible to tailor recommendations, information, and insights to their needs. And once they reside in a messaging app, they can create a continuing presence in the user’s life, just like a person would.
“Once an AI knows you and remembers your history, it stops feeling like a tool and starts to feel like a companion,” says Conor Grennan, chief AI architect at New York University Stern School of Business. “It starts to blur the line between an AI brand ambassador and just a friend who shares your taste.”
Newsletter Sign-up
WSJ | CIO Journal
The Morning Download delivers daily insights and news on business technology from the CIO Journal team.
But that is also where the potential peril lies. As AI agents become more like people, they are increasingly prone to the pitfalls of human relationships. Last year, for example, Air Canada was ordered to refund a customer after the airline’s chatbot gave the wrong details about its bereavement policy. Air Canada said the technology was less sophisticated than more current AI. But it would be rash to assume that today’s AI is beyond error.
Consumers themselves might not trust an AI agent’s answers, and could demand to speak to a human representative the way they do now when they reach automated customer-service phone lines.
WhatsApp owner Meta Platforms is betting that combining AI agents and business messaging can provide a new pillar of growth.
It has a lot riding on that bet. Meta, which paid $19 billion for WhatsApp in 2014, last month raised its projected companywide capital spending for the year to a range of $64 billion to $72 billion, largely on its artificial-intelligence investments.
But AI chatbots could help get more businesses using the company’s messaging products. It has a number of ways to cash in if they do, including the sale of ads on Instagram or Facebook that let users click to send messages over the regular WhatsApp app that they use everyday.
WhatsApp created a specialized app for small businesses and an application programming interface, or API, for large businesses back in 2018, and their usage has been soaring in markets such as India, Mexico, Indonesia and Brazil, according to Meta’s Nikila Srinivasan, vice president of product management for business messaging.
Verizon began a WhatsApp campaign for Spanish-speaking customers in the U.S. last fall, according to Robyn Parks, the telecom giant’s vice president of segment and local marketing. The campaign uses ads on Facebook or Instagram that give customers or potential customers the option to click to chat on WhatsApp, with an opportunity to move the communication to another channel.
The company would typically expect a “conversation start rate” for an ad to be in the single digits. The WhatsApp campaign has been in the double digits, she said. Verizon said it also has ads on WhatsApp that direct customers to the Verizon website and app.
And if a customer drops out of the conversation, Verizon can use the WhatsApp channel to try to re-engage via WhatsApp text, a process that she described as a “warm callback.” That wouldn’t be possible with advertising in a broadcast medium.
“It’s much more of a conversation,” Parks said.
Now imagine untold AI agents initiating those warm callbacks. It could be revolutionary—and risky for companies that don’t learn to manage those agents with the same care that they manage people.
Write to Steven Rosenbush at [email protected]
