Meta's Muse will book things for your customers. The hard part is the first booking.

Meta launched Muse this week, and it calls it "the world's first personal AI agent built for everyone". That's a big claim, so I sat down with the launch video and recorded my first reaction. I came away more impressed than I expected, and with one question I can't stop thinking about as a marketer.

What Muse actually is

Muse is not another chatbot that answers questions. You chat with it, and if it has the context of what you're doing, it goes and gets things done. It places orders on your behalf. It acts like a personal trainer that tracks your goals. It fills out forms. In the launch video it picks things up for school, does the groceries, and rebooks a flight that was delayed.

Under the hood, Meta says each agent runs on Muse Secure VM, a dedicated virtual machine that houses both the agent and your data. You talk to it in the Muse app or directly in WhatsApp, the same way you'd message a person. It's rolling out in the US first, it's free for most of what people need, and it asks for your approval before it completes a purchase.

The use cases are real, and that's the point

What I liked most is how ordinary the examples are. Nobody in that video is building a startup with an agent. They're dealing with school runs, groceries and delayed flights. That's the stuff that actually eats people's week.

And it's all about the connected apps. This is where Meta has an edge nobody else has. The Meta universe is strong, especially with WhatsApp, and an assistant that lives where people already talk to their family and friends doesn't have to convince anyone to open a new app. Meta also says Muse will be connected to people's Facebook and Instagram profiles for expanded context. That's a lot of context.

Meta isn't alone in this race

There's a lot happening in the agentic space right now. Google announced Gemini Spark at I/O in May, an always-on agent that runs on its own cloud machine and works across Gmail and Docs. Microsoft's Copilot isn't quite there yet, in my view. It will take some time before it makes sense for everyone.

The direction is clear, though. As one write-up of the launch put it, if you sell online, Muse is a third large company pointing software shoppers at your website, after Google and Microsoft. A couple of weeks ago I wrote about Claude Cowork getting its own browser, a tool for people who are comfortable with that kind of setup. Muse brings the same idea to people who will never install a professional tool.

Your website is about to matter less

Here's the part that matters for brands, and for anyone who sells a product or a service. There's going to be a world very soon where your website won't matter that much anymore. The discoverability of your products will have to happen somewhere other than on a static site, and not only on social media either.

If an agent can just go out and book an appointment for you, it needs to know your preferences and what you've done in the past. For repeat purchases that's easy. Your agent can work out from your card statements or your loyalty programmes which products you buy and buy them again. Your brand wins that one if you already won the customer.

The hard part is the first booking

The real challenge is everything you've never done before. Say you want an outdoor activity with the kids this weekend. You ask the agent where you can do it, and it books something that fits your general description: not too far away, not too expensive, roughly what you had in mind.

As a brand or a service company, what you're missing at that moment is the whole discovery phase. Everything that used to happen before the booking, the browsing, the comparing, the ad that caught someone's eye, the review they read on the train. The agent compresses all of that into one decision you don't get to see. The grocery example is the same. What does it buy for you when you've never bought it before?

I've seen what happens when people forget where demand comes from. On one account, the marketplace looked like it was growing while it was really harvesting demand we'd built upstream. Agents will make that mistake easier to make, because the upstream part becomes invisible. And for considered purchases, where people take weeks and many touchpoints to decide, the question is which of those touchpoints an agent will ever look at.

"Amazon is not happening to bookselling, the future is happening to bookselling." Jeff Bezos, as quoted by Porter Erisman in Six Billion Shoppers

It's a line I highlighted in Erisman's book. Swap "bookselling" for your category and it describes what's coming. Meta isn't happening to brands. The future is, and this time it shows up as an agent.

What about privacy?

Privacy is always going to be an issue with something that has this much access. Meta's answer is architecture: an isolated, encrypted environment per user, with the data separation designed with input from Signal co-founder Moxie Marlinspike, according to Social Media Today. Meta has built encryption at scale before with WhatsApp. We'll have to see how it holds up, but it's clearly the right thing to lead with.

What I'd be thinking about now

I don't have the full answer yet, and I don't think anyone does. But the question for marketers and sales teams over the next few years is simple to state and hard to solve: how do you make sure your company is the one that shows up when someone asks an agent to just handle it?

That's a positioning question before it's a technical one. Any AI tool will write you a checklist for "agent readiness". Knowing which parts of your offer an agent can actually read, and which story it needs to hear to pick you over the alternative, takes judgement and reps. It's the same reason I keep arguing that domain expertise is the only real leverage left.

We're all moving towards a world where agents just take care of things for us. The brands that think about discovery now won't have to ask later why the agent never picked them.

Sources & further reading

External
Meta: Introducing Muse, the world's first personal AI agent built for everyone
Social Media Today: Meta launches Muse personal AI
PPC Land: Meta puts an AI agent that buys things inside WhatsApp, US only
The Next Web: Gemini Spark, a 24/7 agentic assistant you can email like a colleague
Porter Erisman, Six Billion Shoppers (St. Martin's Press, 2017)

Related posts
We cut the top-of-funnel and the marketplace 'grew.' It was harvesting demand we'd already built.
People spend weeks choosing a sofa. Most furniture brands sell like it's one visit.
OpenAI built a whole browser for the agent. The agent didn't need one.
The Death of Generic AI: Why Deep Domain Expertise is the Only Real Leverage Left