Eli Tan Let META Muse “Take Over My Life:” Where AI Agent Outperformed & Failed

agents in everyday life with Eli Tan, technology reporter for The New Times. Eli, thanks so much for joining us. It's really great to have you on your latest piece says, I gave my life over to Meta's agent and was blown away. Now I'm blown away at the fact thate agent. But that's just me. But I don't want to give th away because obviously we want people to read your piece, but just explain to us you did, what you learned, and your some of your biggest takeaways. Yeah. meta. It's called muse. It's in the App store just like any other app I connected. It's kind of like a chat bot that can, you know, do for you and connect to the internet. So I downloaded it, connected it to my banknts, my emails, my calendar, and basically had to take over my life for, for a couple of weeks. It ordered my groceries,hedulede a podcast for me with these two , you know,hink there were a lot of features that worked really well, some that didn't, some thaould never use. But in the end, I found it to be like, I write in ece one of the most useful AI products I've ever personally experienced. So what made you reach that conclusion that it's one of the most useful AI products tha you ever used? What did it actually do for you versus what were your expectations of what it was going to do? Well, most chat bots you use, no matter how intelligent they are, you kn they're basically like these tools you can converse with. But with muse, I can g some examples. So it, it has a phone call feature where it cane two like personalities. Haley and Brett, a male and female voice. And, you know, for instance, I had call my dental insurance and it told them who I was, gavemy me ID past the security question, and then waited on hold for 15 minutes and then transferred me to the operator, saving me time. I gave it all my bank statements and had it look through them and put the in, you know, spreadsheets and it found, you know, duplicubscriptions or things I was paying for that it could then cancel for me and save me, you know, 40 something dollars a month. So those two things alone I found to be useful and more useful than even like I tried to do those type of t with something like rocket money or cloud and it wasn't able to do them. So thoseimple things alone I found really helpful. Eli, you cover mand its apps for a living, like Instagram, like threads like WhatsApp. I'm just wondering, what is it about that meta ecosystem and the user base that it has access tot makes this stuff so successful? And I draw comparisons to a company that I covered for many years, and that is Tencent. And I think I had WeChat before I had WhatsApp. And there has been some similarities ds to whether that Chinese company could actually leverage some of the capabilities. matter is to roll out something similar. When you think of the success of what we've seen in recent weeks when it comes to meta, what is it about this company that maybe others would not be able to replicate?en it s ecosystem of just look at Instagram, Facebd WhatsApp, billions of people use those apps every single day. Something like over 3 billion people. All of the the d that you have on Instagram and Facebook, you know, is conn to muse. You can ask it, hey, I can you go into my, my Instagram reels and look at like a recipe I saved and then grocery store. You know, it. The thing about these AI products is that the, the trade off is that the more they know aboutsedly from the companies, the more helpful they can be. And meta as ampanye known for a long time, right? It knows a lot about you and whether people find that, you know, intrusive with, with privacy and data concerns is whole problem potentially. But I think for meta, one of the upside that it's already connected to, you know, so many different parts of your life. And youtioned that some things it didn't do well. So, yourself having to step in? Where did muse not perform up to your expectations? Well, a lot of websites, they're just not yet m for agents to go on them and operate themmethingI was it was really helpful at booking me like a restaurant reservation or getting me a movie ticket to like a local penu basically just fill out a form online and pay, right? Butor some like getting a ticket to an AMC theater where I want to sign into my rewa account, it had a tough time. You know, it can sign in and it can go to your email and it can get the But sometimes it would get tripped up or it would get flagged as, you know, a bot the website. And I would have to go and manually, you know, put my password o to the email and get a confirmation code or something like that. The other issue is that there are a lot of websites that it's not going tor agents to use them. Amazon is one of them. I wrote using Amazo in the piece, and coming out, really the, youwas know, Amazon decided that they don't want muse agents to go on Amazon and want humans to, to go through the website. So they banned it. And I think that could be a problem going forward where all these companies like Amazon that has own agents is notgoinga company like Meta or Google or OpenAI able to use it really smoothly and cleanly., what's been the response to the piece? I mean, how have people weighed in? Because I think the comments are always fascinating as to, you know, what people areing. I mean, someone left a comment on one of Marley and my videos yesterday saying,r free? Which I suppose brings together the two opposing kind of crosscurrents that we're dealing with in this market t nod then the sort of macro environment and ability ches. But I'm just wondering, what has been the reaction from people who' read your piece? And for those who do find it perhaps a bit blie giving, you know, so much data an confidential information over to technology, there been, I would say, two buckets that cers fall into. A lot of people are very concerned about the privacy aspects. They're already concerned about it with Instagram and Facebook. And this is like yet another way that metacollecting our personal data and monetizing it and making money from it. And this way more personal than ever, right? Because it's reading your email and potentially, you know, seeing text messages and things like that. The other type of commenter, you know, also concerned,nd Ihare this concern about that we're automatingts of o life that we can automate. Then the technology is good enough, but why would we want to do that? One instance in the story was that I muse going on Facebook Marketplace and haggle for me, trying to b bowls and plates from this company called Heath cs. And a lot of people said, why wouldn't you just do that yourself? You don't want to talk to people. You don't want to have human connection. And I largely agree with them. Ink there are a lot of things that I, I was using muse for that. You k now that the experiment's over, I would much prefer to just probably do them on my own. Eli, the privacy conversation is something that comes up all the time when we talk abou AI adoption as arguably the primary barrier is this privacy concerns and confidence in turning over your information call it complicated historyet when it comes to privacy, how yn your bank accountmation? I mea had it paying bills, parking tickets for you. You gave it full access to your email. What was that experience like? Because a lot o the conversations Sam and I are having, and I'm sure you are too, is that one of the greatest impediments to adoption is this concept of turning everything over and what that means. It w leap of faith, for sure. It's going to be a huge obstacle forpeoplet of people, quite frankly, they will never be comfortable within it comes to the banking information, I think it's a little bit more secure than people realize. You know, it's with this integration with stripe. So y basically know, you give stripe and Plaid another, you know, financial services company, you give it your information. It's not like I'm taking a picture of it card o muse and having it manually do it. Yeah. Whe emails, I mean, I was comfortable connecting personal email to this, but not my work email, for instance, because, you known there, right? And I wouldn't want a third party agent to come in and somehow get access to. But yeah, this is going to be a major problem for meta. I think one, given their reputation with privacy, but also just with how comfortable people will be handing over that privacy, trading it for, you know, whatever convenience these agents can offer them. And Eli, we're tight on time here. But just one last question to wrap. Do you plan to keep using muse now that your your joutic experiment is over? I do, yeah, I will keep using it for all customer service needs. I get that people, you know, there's we're humans, right? We want to do things ourselves. We want to talk to people, but sitting on hold so that I can talk to another company's AI agent to do customer service. That's not something that I'm hoon to. I'll, I'll happily have muse keep doing that for me as well as the pa tickets. I have a feeling Sam and I may be investigating that process during the cial break here, but Eli, it has been such