Where to start with an AI agent on WhatsApp: what Meta's own product leaders advise

Fred Leach, Meta's VP of Product Management, starts from what a business already has: customers already message it, and the information an AI agent needs is the same information people need to run the business. In his telling, the businesses getting results did not wait for a perfect strategy. They picked one use and started, and the ones stuck in a trial are the ones that never put anything in front of customers. His advice is to try AI on something the business is already expert in, where its limits show quickly.

Meta's July 2026 report page, Beyond chatbots: the agentic economy is here, names what holds businesses back in plainer terms: channels that do not talk to each other, so a customer has to start over when they switch; poor data and loose internal processes; and, as a result, projects that never leave the trial. Its answer is to give the AI the business's own context and to treat messaging as the foundation of the relationship rather than a one-way channel for notifications. The same page says more than one million businesses use Meta Business Agent each week on Messenger and WhatsApp.

James Detmer, a Director of Product Management at Meta, gives a similar answer: find your “magic moments”, the experiences already working in a business that are ready to reach more people with AI behind them. Start there, he says, go deep, and let trust build as results follow.

Heidi Rhodes, Meta's Global Director of Customer Success, says the question executives ask her most is how to tell whether an AI agent is paying off. Her answer is to pick the moments that matter to the customer and use them to start a relationship, not only to close a sale. She also names a risk: as customers get used to talking to AI, a brand can lose its direct line to them. Her remedy is to meet customers in the apps they already use every day, so there is no gap to close.