My experience with Meta’s new AI agent, Muse

One of the best ways to understand an investment is to get up close and personal: to get your boots on the ground, as it were.

Paperwork is important, mind you. But it can only tell you so much.

That’s why, when I was a real estate developer, I personally visited properties, inspected buildings, talked to tenants, inspected buildings… and occasionally even climbed onto roofs.

Considering all the real estate investment trusts (REITs) our team follows these days, I do still occasionally do all that. But that’s not always possible – or even helpful – depending on the company, product, or service I’m investigating.

Take Meta’s (META) new personal AI agent, Muse. My boots-on-the-ground approach here involved downloading it and actually using it this weekend.

It probably wasn’t necessary to give this new digital employee a name as well. But I decided to anyway: Dunk. And since she has a female voice, I’ll refer to Dunk as such from now on.

(I gave her a basketball emoji as well. While I’m not sure I could touch a net these days, and used to be able to dunk back when I was playing for my college team.)

Now, Muse can do quite a bit, including:

It can also work across connected applications and continue on task even after I close the application – all of which works for me. After all, I produce content across seven platforms: social media platforms X, LinkedIn, Instagram, and TikTok; videos on YouTube; and articles on Seeking Alpha and, of course, Wide Moat Research.

So instead of logging into multiple platforms every night, I began connecting accounts to Dunk to see how many followers and subscribers I was gaining or losing per day.

That might sound trivial, but it saves me more time than you might imagine. And that detail alone made me appreciate her.

I’m just not sure if I appreciate her enough.

I also connected Dunk to my calendar so she can give me a five-minute rundown every morning. Before I leave for my morning walk, I hear about the weather, news and market updates, and what's on my calendar.

This morning, I asked her this as well:

And here’s how she responded:

Eventually, I'd like to connect more of my Wide Moat Research workflow so Dunk can update me on subscriber additions and cancellations, engagement, social-media growth, YouTube performance, and newsletter statistics.

Muse can handle all of that – and more – on one dashboard. It’s essentially a digital chief of staff, so Meta has something quite valuable on its hands.

I know that investors have been worrying a lot about Mark Zuckerberg’s enormous AI spending this year. We’re talking about $30.1 billion on capital expenditures in Q2 alone, up more than 80% year over year.

Unsurprisingly, that helped push free cash flow down dramatically. So I understand the unease it caused.

Even so, Meta's underlying cash machine continues to get bigger. It generated about $31.9 billion of operating cash flow in Q2, up almost 25% year over year.

And it generated $130 billion in cash from operations over the trailing 12 months, compared with $89.3 billion of capital expenditures (capex).

That means Meta still has roughly $40.7 billion left over even in the midst of such a massive infrastructure buildout. This is just part of the reason I personally own shares.

Since the beginning of 2023, in fact, it’s generated approximately $342 billion of operating cash flow against $183 billion of capex. Facebook, Instagram, and the rest of Meta's advertising machine are essentially financing Zuckerberg's AI bet.

So he doesn’t actually need Muse to pay the bills. If all goes well, it will just add to his already very well-padded bottom line.

Muse offers three levels of interaction and capability. There’s a free tier, one that costs $20 per month, and another at $100.

Institutional research group Oppenheimer has estimated Meta would make $27.5 billion in annual revenue on 115 million subscribers paying $20 per month. And while that might seem like a tall order, the social media giant already has around 3.6 billion daily active users across its platforms.

That makes 115 million a mere 3.2% fraction of its potential reach.

Recognizing that, I ran some scenarios. Like how 36 million $20-per-month-paying Muse subscribers – a mere 1% conversion – would still generate $8.64 billion of annual revenue. And at 5%, or 180 million users, Meta would make $43.2 billion.

Applying a 35% operating margin in consideration of how much computing power AI agents requires, a mid-point 3% conversion could add $2.97 per share to the larger company’s earnings. Though that might be a conservative calculation.

After all, it doesn’t take the $100-per-month tier into consideration at all. Or business-agent revenue. Or commerce fees.

Then again, it might also be an optimistic assessment considering how much trust Muse requires in order to optimally operate… and how much damage Zuckerberg has done to his company’s image over the years. There was:

So while I’m perfectly fine letting Dunk see my calendar and my social media accounts, do I really want it accessing my credit cards or bank account?

Do you?

Apparently Amazon (AMZN) doesn’t. It banned Muse from shopping on it this month, alleging that Meta never asked for or obtained relevant permissions.

Don’t get me wrong. I’m exceptionally impressed with how well Muse can work. Yet I’m still hesitant to let it in too much further. And since I imagine I’m not alone in that, I’ll be closely watching not only how many people download Muse but also how many:

That's the funnel I'll be watching.

In short, I want to know if Meta has enough of a “trust” moat to make Muse as big as it could be.

Yes, it already has one of the largest distribution advantages in the world between Facebook, Instagram, WhatsApp, Messenger, and now AI glasses. And it has an enormous opportunity to create even greater digital real estate through Muse’s ability to help consumers get things done.

After using it myself this past weekend, I really can better understand why Wall Street is excited about this product. Muse might truly be capable of handling billions of consumers’ digital lives.

I know I’m willing to keep trying it out up to a certain point. For instance, I’m ready to let it explore more of my Wide Moat Research workflow.

Eventually, I want to direct Dunk to check my calendar and find the best flights for me. The same goes for hotels and rides.

But beyond that?

No matter how far Muse can go, I can’t escape the fact that it’s Meta behind it. And until I either can forget that fact or Zuckerberg can better earn my trust…

I just don’t think I can sleep well at night (SWAN) giving it my money.

Meta the stock? Yes.

Muse the app? I’m just not so sure…

Happy SWAN investing,

Brad Thomas
Editor, Wide Moat Research

Disclosure: Brad is long META.