Is the internet ready for Meta's Muse and its AI agent rivals?

Good afternoon.

First, this is your last chance to RSVP and join me and 70+ of your peers at The Aisle’s first subscriber party on Monday in NYC.

It’s going to be a great night with some good food and drinks, and of course healthy industry chatter. Looking forward to seeing many of you that night.

Moving on…I thought we were temporarily done with the AI personal agent craze after the last few weeks saw the Instinct and Grok Bot personal agents go viral among VCs, tech entrepreneurs, and other early adopters (including independent journalists like this one).

But on Tuesday, Meta launched Muse, its own take on a digital helper that will set out to accomplish tasks for you while you wait. Or sleep. Or do other work. Or go out in the backyard, lie down in the grass, and clear your mind for a few minutes away from all of your screens (or bickering tweens).

The social network formerly known as Twitter immediately blew up with an army of internal Meta supporters, as well as executives at partnering companies (hello, Shopify and Stripe!), telling us how much they’ve been enjoying their early access to the technology.

“muse is way hotter than the other personal agents (and also way faster, but mostly hotter),” Meta’s Chief AI Officer Alexandr Wang joked while retweeting that meme showing a boyfriend fawning over the new girl on the street (Muse) while ignoring his own girlfriend (Instinct).

So I logged into Muse on my desktop browser using the same credentials I use for my Instagram and Facebook accounts. I then asked Muse what it was good at. Actually, I asked Lana. Muse had asked me to name my agent, and I chose the first name of the singer who stole my last name for her stage name.

“Basically, if it’s research, monitoring, or coordinating, hand it over,” Lana told me.

Things like watching for price drops or a rare open reservation at a hot restaurant. Architecting and executing complex travel plans. Rummaging through an email inbox to summarize some messages or flag the most urgent ones. And, if you are so daring, drafting and sending those emails off for you too (def not for me).

I asked Muse about finding a certain size of pants for me—I know, weird—and the AI popped open a browser and started navigating to various pages on Amazon before selecting a pair of Amazon Essentials as a good fit for me. (I was insulted.)

Anyway, these are the three buckets of questions on my mind right now amid this craze:

“Resy does not currently permit third-party bots or agents to independently access or interact with the Resy platform,” company spokesperson Lauren Young told me in a statement. “These types of automated interactions can introduce risks to the platform and create an unfair advantage for some users. As agentic technology evolves, we’ll continue to evaluate opportunities thoughtfully while maintaining the trust and reliability that diners and restaurants expect from Resy.”

Resy said it does have integrations with ChatGPT and Claude to allow for authorized AI searches using its system. And because of the nature of a reservation business, Resy has had to combat bots for a long time, even if modern AI technology was not behind them. Now it has to figure out how to deal with this breed of AI agents that want to take actions on behalf of an actual Resy customer.

If these agents do become popular with a mass audience, some have suggested Resy auction off a portion of reservations to AI agents or simply charge for authorized access to its system.

But in a conversation with Sources founder Alex Heath, Meta CEO Mark Zuckerberg proposed kind of the opposite setup in the future: companies paying Meta a small fee if Muse completes a transaction on behalf of a customer.

Just a few months ago, an executive at Resy’s parent company American Express told me the company would stand behind purchases made by a cardholder’s AI agent if something went wrong as long as the agent registered with Amex first.

“[W]e’re big believers in agentic commerce,” Amex Head of Global Innovation Luke Gebb told The Aisle at the time.

We’ll see if Instinct and other agents garner enough traction to force Resy and the rest of the consumer internet to the table to cut deals with them. Who will have to blink, or adapt, first?

Now on to the good stuff…

Instacart has spent more than a decade trying to make getting groceries easier. Now it wants to help with the part that can come before that: figuring out what to eat.

On Wednesday, Instacart launched Clementine, an AI shopping assistant that can suggest meals and the ingredients needed for them, find deals, and reorder a shopper’s usual purchases with just a tap or two.

The goal is to get more shoppers to come to Instacart before they already know what they want or need to buy.

Clementine is meant to move the company “upstream,” Instacart CMO Laura Jones told The Aisle last week, beyond delivery and into the mental task of deciding what a household is actually going to eat. “That question coming every night for the rest of your life,” Jones said.

That ambition also shaped how Instacart designed the product. Product designer Shaun Harris told me the team wanted to avoid turning Clementine into a chatbot that requires a long back-and-forth before a shopper can get anything done. Call it conversational shopping with as little conversation as necessary.

“This can’t be something that feels more laborious,” Harris said. “It can’t be tons of exchanges where now you’re kind of getting in the way of my shopping journey.”