AI Can Now Shop for Customers. Is Your Business Ready for a Buyer That Isn't Human?

What changed

Meta launched Muse, its new AI agent, in the U.S. on September 8, available through a standalone app and on WhatsApp. Unlike a chatbot that explains how to do something, Muse can go do it: browse websites, fill out forms, send emails, book travel, make purchases, and work across apps you've connected to it. It keeps working after you close the app and comes back when something changes or needs your approval, the way a human assistant would check in partway through a task. Meta's own example is telling: tell Muse to sell your car, and it can handle pieces of that process on its own. Commerce is built directly in. Muse can buy things using Stripe's Link, which issues a one-time card number so the agent never actually holds your real card details, and Meta says support for Shopify's Shop Pay is coming next.

The honest nuance

This isn't a clean launch. Reuters reports that Meta's internal testing turned up security and reliability problems before release, including cases involving private information, and that the issues were serious enough to delay Muse's original April launch date. Meta's response is a genuinely extensive security setup for a consumer product: Muse runs in its own isolated cloud VM, a separate system called Sentinel polices what it does on the internet, it asks for your permission before sensitive actions, it keeps an audit trail, and you can control which connected apps it's allowed to touch. That makes Muse worth watching for a second reason beyond the shopping angle. It's an early, concrete look at what safety guardrails for a consumer AI agent actually look like in practice, not just a company's promise that its agent will behave.

Why it matters for your business

This week has built a pretty specific case. On September 1, companies started selling tools specifically to measure whether AI can find a given business. On September 3, John Lewis reported AI-driven search traffic jumping from 0.3% to 2.5% of its total. On September 4, Adobe's data showed visitors referred by AI converting at meaningfully higher rates than typical traffic. Now Meta has shipped a mass-market agent that can carry out an actual purchase, not just recommend one. Put together, that's no longer a reason to treat agentic shopping as a distant idea. It points at a new kind of customer showing up at your website: software acting on behalf of a person, working through a request like "find a unique apple-scented laundry detergent under $25, check the reviews, confirm it's in stock, and order it" without that person ever browsing five different stores themselves.

What to actually do

You don't need to buy another AI tool to prepare for this. What you need is a website that's easy for software to understand and transact with, which happens to be the same thing that already helps human shoppers and search engines: clear product names, accurate prices, detailed descriptions, current inventory counts, spelled-out shipping and return policies, structured product data, visible reviews, and a checkout that doesn't require an account or a phone call to finish. Pick one product page this week and read it the way an agent would: does it answer "is this in stock, what does it cost, and can I trust it" without a click-through or a call? Fix what's missing there first, then repeat it across your catalog.

From Kindloom Labs

If getting your website and product pages ready for both human and AI shoppers feels like a lot to tackle alone, our AI Small Business Starter Kit walks through practical, no-tech-team-required steps for exactly this kind of website and workflow cleanup.

Sources

• Introducing Muse, a Personal AI Agent (Meta)
• Meta launches AI agent that can access other apps, send emails, make payments (Reuters)

Keep reading