Pulse News

**Meta Plans New Push for Business Chatbots**

By Kalley Huang

Mar 4, 2025, 10:00am PST

Meta Platforms is preparing a new push to woo businesses to set up and run artificial intelligence chatbots on Messenger and WhatsApp, according to a current employee. As a newcomer in a crowded market, Meta has pitched its business chatbots as tools to provide customer support, plus information and recommendations that will lead to sales, the employee said. The company has also suggested to businesses that running a chatbot on Meta's apps could help Meta better target ads from that business, the employee said.

But businesses that have tested Meta's chatbots have complained that the chatbots can take months to tailor responses to a brand's voice, according to the current employee. Also, Meta's chatbots don't learn from interactions between human representatives of a business and its customers. Rivals, such as customer support startup Gorgias, say their AI agents can learn a brand's voice from the outset and act to change, track, cancel or return orders.

These challenges reflect Meta's uphill battle in selling software to businesses, an arena where it historically has struggled. The company has had mixed results getting businesses to use its AI model, Llama. Last year, Meta said it would shut down Workplace, collaboration and productivity software it had been selling to businesses for several years.

Meta started testing business chatbots in October 2023 and has since quietly rolled out the service, now available in English and Spanish. The effort accelerated after Meta hired Clara Shih to lead its business AI group in November. Shih previously led AI at Salesforce, overseeing product development, research, sales and marketing for software including its business chatbot product, Agentforce.

In a sign of the urgency behind Meta's efforts, Shih's group has been assigned monthly goals rather than the half-year goals typical at the company, according to the current employee. The team, which includes company veterans who worked on Meta's earlier attempts at commerce and enterprise sales, is also trying to hire outsiders with more experience in business-to-business sales, they said.

Hundreds of businesses are now testing Meta's chatbots, ranging from American furniture retailer Wayfair to smaller retailers selling consumer goods such as clothing and beauty products, according to the current employee. That's a fraction of the hundreds of millions of businesses with accounts on Meta's social media apps. A spokesperson for Meta declined to comment.

In its tests of business chatbots, Meta has focused on businesses that use click-to-message ads—which send people who click on an ad directly into a text message with that business—on Messenger and WhatsApp, Chief Financial Officer Susan Li told analysts in January. Many of the tests have been in Southeast Asia and Latin America, where people often use messaging apps to communicate with businesses, she said.

"We know that we're going to need to invest more aggressively to figure out how to go to market in markets that have for various historical reasons not evolved down that path" of using messaging apps to communicate with businesses, Li said in January. Those markets include the U.S. and Canada, where Meta makes much of its revenue.

Businesses testing chatbots on Meta's apps have also asked the company for more information about how it uses their data and how their chatbots are performing, according to the current employee. Meta says it collects messages sent and received in conversations using business chatbots to improve its AI.

Meta has faced hurdles in its other efforts to get businesses to use its AI. Since it introduced text- and media-generation tools for advertisers in May 2023, more than 4 million advertisers use at least one of those tools, Li said in January. (Meta says it has more than 10 million advertisers in total.) Similar to how it has promoted the business chatbots, Meta is also pitching the tools as a way to improve the reach of ads on its platforms. In blog posts, the company has said the tools will help advertisers create more variations of ads, which will ultimately help them identify which ad performs best.

But some advertisers have complained to Meta about the accuracy of the tools and said that they had to rewrite or edit AI-generated copy, according to two current employees. Part of the problem is that Meta's tools can't yet provide the level of control and detail advertisers are used to and want, they said.

Meta also has considered testing AI models from Chinese firm DeepSeek in its tools for advertisers. The company currently uses its own AI models for the tools. But it has weighed such a move, Shih has told her staff, because DeepSeek performs so well and Meta wants to make sure it provides advertisers with the best experience.