Meta's Muse and the rise of AI agents that act for you

On 8 September 2026, Meta launched Muse, a personal AI agent designed to take actions rather than just answer questions. According to TechRepublic and Khaleej Times, Muse can send emails, book travel, fill in forms and make purchases, and is available at first only in the United States, to adults, through a dedicated app or through WhatsApp. A basic version is free, with paid tiers for heavier use.

Coverage also highlighted concerns. Reuters reported that Meta employees testing Muse had raised worries about security and reliability before launch, and Meta said it delayed the release to strengthen security and added monitoring that can review some actions before they run.

What an "agent" is

Most people have used AI chatbots that answer questions. An agent goes further: you give it a goal ("book me a flight to Mombasa on Friday under 10,000 shillings") and it takes the steps, searching, comparing, filling forms and paying, often in other apps, sometimes while you do something else.

Muse is one of several agents launched by large technology companies in 2026. Together they point to a shift in how people may shop and use services: from searching and clicking to asking an assistant to do it.

Why WhatsApp matters for Africa

Muse is US-only for now. But its arrival on WhatsApp is the detail African businesses should notice. WhatsApp is where much of Africa's commerce already happens: customers ask about products, negotiate and arrange delivery in chats. If agents become common on WhatsApp, some of those conversations may one day start with an AI acting for the customer rather than the customer typing.

What businesses can do now

It is too early to predict how agents will work in Kenya or when they will arrive. But some preparation is sensible whatever happens:

1. Make your information clear and consistent. Prices, opening hours, delivery areas and stock should be accurate everywhere you publish them. Agents, like people, are confused by contradictions.
2. Answer fast. A business that replies in seconds suits a world of automated customers as well as human ones.
3. Keep a human in the loop. Agents make mistakes. Be ready to confirm important orders with a person.
4. Build a recognisable brand. When an assistant recommends options, the brands people already know and ask for have an advantage.

Safety and trust

Agents that can spend money and send messages raise real risks: fraud, mistakes and privacy breaches. Businesses should be cautious about letting any automated system, theirs or a customer's, complete payments without confirmation.

In Tari Studio, our own AI features follow that principle. The in-app assistant proposes changes and you press Apply; Autopilot prepares posts and waits for your approval; automations answer WhatsApp questions from answers you approved, and hand over to your team. Nothing is published or changed without a person saying yes. The AI assistant.