Meta expands business AI
What happened
- Meta widened access to its Meta AI business assistant across major global markets with better local-language support. - The company is also reported to be tracking employee keystrokes and mouse activity to train AI models, sparking internal backlash. - Platform-level assistants will compress simple messaging needs, so vendors must sell workflow depth and unified pipelines instead (tech.yahoo.com) (cnbc.com).
Why it matters
Meta has widened access to its Meta AI business assistant for advertisers and agencies across the United States, Europe, the Middle East and Africa, Asia-Pacific, and Latin America. (tech.yahoo.com) The assistant launched inside Meta Ads Manager last year, where it answers account questions and suggests campaign changes using a business’s own Meta ads data. Meta said beta users resolved common account issues at a 20% higher rate with the tool. (tech.yahoo.com) Meta also said small-business advertisers cut ad cost per result by 12% after applying its “opportunity score” recommendations, which are now built into the assistant. The latest rollout adds stronger local-language support for users in the newly expanded regions. (tech.yahoo.com) The push fits Meta’s larger plan to turn its consumer AI into a business product for the companies that already sell through Facebook, Instagram, WhatsApp and Messenger. Clara Shih, Meta’s head of business AI, told CNBC in March 2025 that the company was targeting “hundreds of millions” of businesses. (cnbc.com) Meta has been moving beyond ad support and into customer service and shopping chat. In October 2025, CNBC reported that Meta’s Business AI assistant was being positioned for use in Facebook and Instagram ads and on Shopify-powered websites, with WhatsApp and Messenger launches starting in Mexico and the Philippines. (cnbc.com) At the same time, Meta is building the training data for more advanced AI agents by watching how employees use computers at work. CNBC reported on April 22 that the company’s Model Capability Initiative captures keystrokes and mouse clicks across sites including Google, LinkedIn, Wikipedia, GitHub, Slack, Atlassian, Threads and Manus. (cnbc.com) CNBC said the internal list of tracked sites originally included OpenAI’s ChatGPT and Anthropic’s Claude, though the list was still changing. A Meta spokesperson told CNBC the project is meant to gather “real examples” of how people use computers so its agents can learn tasks like clicking buttons and navigating menus. (cnbc.com) That pairing shows where Meta is heading: one product answers routine business questions inside its ad stack, while another effort trains software to imitate office work across the web. The nearer-term value is faster support and cheaper campaign tuning; the longer-term bet is that businesses will want AI that can do the work, not just chat about it. (tech.yahoo.com) (cnbc.com)
Key numbers
- Meta said beta users resolved common account issues at a 20% higher rate with the tool.
- (tech.yahoo.com) Meta also said small-business advertisers cut ad cost per result by 12% after applying its “opportunity score” recommendations, which are now built into the assistant.
- Clara Shih, Meta’s head of business AI, told CNBC in March 2025 that the company was targeting “hundreds of millions” of businesses.
- In October 2025, CNBC reported that Meta’s Business AI assistant was being positioned for use in Facebook and Instagram ads and on Shopify-powered websites, with WhatsApp and Messenger launches starting in Mexico and the Philippines.
What happens next
- (tech.yahoo.com) The push fits Meta’s larger plan to turn its consumer AI into a business product for the companies that already sell through Facebook, Instagram, WhatsApp and Messenger.
- In October 2025, CNBC reported that Meta’s Business AI assistant was being positioned for use in Facebook and Instagram ads and on Shopify-powered websites, with WhatsApp and Messenger launches starting in Mexico and the Philippines.
- The nearer-term value is faster support and cheaper campaign tuning; the longer-term bet is that businesses will want AI that can do the work, not just chat about it.
Quick answers
What happened in Meta expands business AI?
Meta widened access to its Meta AI business assistant across major global markets with better local-language support. The company is also reported to be tracking employee keystrokes and mouse activity to train AI models, sparking internal backlash. Platform-level assistants will compress simple messaging needs, so vendors must sell workflow depth and unified pipelines instead (tech.yahoo.com) (cnbc.com).
Why does Meta expands business AI matter?
Meta has widened access to its Meta AI business assistant for advertisers and agencies across the United States, Europe, the Middle East and Africa, Asia-Pacific, and Latin America. (tech.yahoo.com) The assistant launched inside Meta Ads Manager last year, where it answers account questions and suggests campaign changes using a business’s own Meta ads data. Meta said beta users resolved common account issues at a 20% higher rate with the tool. (tech.yahoo.com) Meta also said small-business advertisers cut ad cost per result by 12% after applying its “opportunity score” recommendations, which are now built into the assistant. The latest rollout adds stronger local-language support for users in the newly expanded regions. (tech.yahoo.com) The push fits Meta’s larger plan to turn its consumer AI into a business product for the companies that already sell through Facebook, Instagram, WhatsApp and Messenger. Clara Shih, Meta’s head of business AI, told CNBC in March 2025 that the company was targeting “hundreds of millions” of businesses. (cnbc.com) Meta has been moving beyond ad support and into customer service and shopping chat. In October 2025, CNBC reported that Meta’s Business AI assistant was being positioned for use in Facebook and Instagram ads and on Shopify-powered websites, with WhatsApp and Messenger launches starting in Mexico and the Philippines. (cnbc.com) At the same time, Meta is building the training data for more advanced AI agents by watching how employees use computers at work. CNBC reported on April 22 that the company’s Model Capability Initiative captures keystrokes and mouse clicks across sites including Google, LinkedIn, Wikipedia, GitHub, Slack, Atlassian, Threads and Manus. (cnbc.com) CNBC said the internal list of tracked sites originally included OpenAI’s ChatGPT and Anthropic’s Claude, though the list was still changing. A Meta spokesperson told CNBC the project is meant to gather “real examples” of how people use computers so its agents can learn tasks like clicking buttons and navigating menus. (cnbc.com) That pairing shows where Meta is heading: one product answers routine business questions inside its ad stack, while another effort trains software to imitate office work across the web. The nearer-term value is faster support and cheaper campaign tuning; the longer-term bet is that businesses will want AI that can do the work, not just chat about it. (tech.yahoo.com) (cnbc.com)
