What if you could hand an AI your inbox, shopping accounts, travel plans, and calendar and just let it handle things?
That’s essentially what Meta wants you to do with Muse, its new personal AI agent. Unlike a chatbot that waits for a prompt and gives you an answer, Muse is designed to take a goal and actually work through it. It can browse websites, fill out forms, send emails, book travel, shop online, and keep working after you close the app.
The idea is convenient. It’s also a pretty big shift in how we interact with the internet.
In the latest episode of The Daily Carnage Show, Shannon Sankey takes a closer look at what Muse can actually do, how much access it needs to do it, and what happens when an AI agent starts making more of our digital decisions for us. If you want to hear the full conversation and Shannon’s take on Muse, you can watch the episode here.
Muse Isn’t Just Answering Questions
The easiest way to understand Muse is to think beyond the usual chatbot experience.
You could ask a chatbot to help you plan a trip to Denver. Muse is supposed to take that a step further: give it the goal, and it can research flights, compare options, make reservations, fill out forms, and come back to you when it needs approval.
That’s the whole point of an AI agent. You’re not just asking it what to do. You’re giving it something to accomplish.
According to Meta’s announcement of Muse, the agent can send emails, book travel, make purchases, and continue working in the background after a user closes the app. It runs inside a dedicated cloud computer called Muse Secure VM, with another system called Sentinel controlling whether it can interact with the internet.
That setup is important because an agent needs considerably more access than a chatbot does. If you want Muse to clean up your inbox, it needs access to your email. If you want it to shop for you, it needs access to shopping sites. If you want it to book a trip, it needs to be able to interact with travel websites.
The more useful you want it to be, the more you have to let it see.
The More Useful Muse Gets, the More It Knows
That tradeoff became pretty obvious when people started testing it.
In one hands-on test, The Verge connected Muse to Gmail and Amazon. Muse deleted thousands of promotional emails, then used information about size, style, and color preferences to find workout tops on Amazon. It even noticed other products sitting in the tester’s cart and asked whether they should be removed. The impressive part wasn’t that it could answer questions. It was that it could actually take care of the task.
The uncomfortable part came when the tester asked Muse what it knew about them.
Muse was able to surface detailed interests based on information from connected Instagram and Facebook accounts, including data that wasn’t visible in the same form through the regular Instagram interface.
That’s a different kind of AI experience. The agent isn’t just responding to what you tell it in the moment. It can potentially build a much more detailed picture of you by pulling together information from the services you’ve connected.
WIRED’s testing raised a similar concern. Muse could find products on Facebook Marketplace based on specific criteria, offer to contact sellers, and continue working on the task later. At the same time, the app encouraged the tester to connect additional sources of personal information, including email, bank accounts, and passport details.
Meta says users choose which apps Muse can access and how much permission it gets. They can also disconnect services, tell Muse to forget information, and opt out of having interactions used to train Meta’s AI models. The company also says Muse conversations and VM data aren’t shared with its advertising systems. Still, there’s an interesting tension here.
The product is built to become more useful as you give it more information. So while you technically control how much access Muse gets, the experience itself encourages you to hand over more.
Meta Has an Interesting Advantage
There’s another reason Muse is worth paying attention to: Meta already sits across a huge part of people’s digital lives.
Instagram. Facebook. WhatsApp. Messenger. Marketplace.
That gives the company something many AI startups have to build from scratch: an existing ecosystem where people already spend time, communicate, shop, discover products, and share information about themselves and Muse can potentially connect those pieces.
That matters because an agent doesn’t need to replace every website or app to change how people use them. It just needs to become the layer sitting between the person and those services.
That brings us to the part marketers should probably be watching most closely.
What Happens When the Customer Stops Shopping?
A lot of digital marketing is built around the assumption that a person is going to do the work.
- They search for something.
- They click an ad.
- They compare products.
- They read reviews.
- They visit a product page.
- They sign up for an email.
- They add something to their cart.
- They come back later because you retargeted them 14 times.
You get the idea. AI agents start to mess with that entire journey.
If Muse can compare products, read reviews, browse websites, fill out forms, and make purchases on someone’s behalf, the customer doesn’t necessarily need to move through all those steps themselves.
That changes the question for marketers.
Instead of only asking, “How do we get someone to click?”, brands may eventually have to ask, “How do we get recommended by the agent?”. That’s a pretty significant shift.
We’ve already moved from the traditional search experience, where people sorted through ten blue links, toward AI-generated answers that summarize information for them. An agent takes that one step further.
Eventually, the experience could be less:
“Here are five products you might like.”
And more:
“I compared them, picked one based on your preferences, and bought it. Here’s the receipt.”
That means websites, product information, pricing, reviews, checkout flows, and even customer service could increasingly need to work for machines that are acting on behalf of people.
The agent becomes part of the customer interface.
Convenience Has a Cost
There’s a very obvious appeal to all of this.
Nobody wants to spend an hour cleaning out promotional emails. Plenty of people would happily hand over the annoying parts of booking travel, scheduling appointments, or finding a product that meets a very specific set of requirements.
But not everything we do online is just administrative work. Sometimes browsing is the point.
Finding a weird gift for someone. Going down a rabbit hole while planning a honeymoon. Comparing running shoes. Discovering a restaurant you didn’t know existed. Developing an opinion about what you like.
Those experiences can be inefficient, but they can also help us build taste, knowledge, and a sense of what we actually want.
It’s a little like grocery delivery. Having someone else pick up your groceries is incredibly useful. But there are also things you notice when you walk through a store yourself, see something unexpected, compare options in person, or change your mind at the last minute.
The internet has always been part of how we discover things and make decisions. If agents start making more of those decisions for us, the tradeoff isn’t only about saving time.
It’s also about how much of that process we want to outsource.
The Bigger Question for Marketers
Muse is still new, so it’s too early to know exactly how much of our digital lives people will eventually hand over to agents, but the direction is interesting.
For consumers, the question is how much convenience is worth giving an AI access to the accounts, information, and preferences that make its recommendations useful.
For marketers, the question is what happens when the person who visits your website isn’t necessarily the one making the decision anymore.
The customer may still be the person buying the product. But the agent could be the one searching, comparing, negotiating, filling out the form, and clicking “buy.”
That creates a very different internet and it leaves us with a pretty simple question: What parts of your digital life would you actually trust an AI agent to handle for you?
