Every live agentic checkout still has a human in the loop

Episode 20 - Why only 5% of distributors call ecommerce their top channel - Transcript

Matt: Welcome to Data Versus Commerce, where we explore the messy middle between database and doorstep. I’m Matt Johnson.

Floyd: And I'm Floyd Blaikie. Let's dig in.

Floyd: I'm really excited about this one, Matt. We've got what I think might be a niche e-commerce celebrity in the mix today. I’d like to welcome Paul Drecksler to the show. Paul, can you just kick us off by saying a little bit about yourself, and what it is that you do every Monday?

Paul: Hey, Floyd. Hey, Matt. Thanks for having me on. For the past almost six years, about to be 300 editions in two weeks, I produce a weekly e-commerce newsletter that recaps the most important industry news. I try to cast a fairly wide net across e-commerce, B2B, D2C, advertising, basically things that impact merchants and operators.

I read thousands of articles a week so that you don't have to, basically. If you couldn't follow any other news that week, you read "Shopifreaks" and hopefully it'll catch you up to speed.

Floyd: Nice. Very excited about this. And, although our show is named Data vs. Commerce, we think about it more like couples therapy and not a debate show, but this might turn into one in this episode 'cause we are going to all try to get to an agreement by the end of this episode on a question I think a lot of people are either struggling with or just kind of ignoring and hoping that it goes away, which is this: What even is agentic commerce?

So tell me how you're thinking about this question. I wanna hear from both of you on this.

Paul: Kick it off first, Matt. I already know mine, so I'm curious to hear yours.

Matt: Oh yeah. Oh, guest always goes first here, Paul. So yeah, please kick us off.

Paul: All right. So my definition of agentic commerce is when the entire transaction happens autonomously start to finish based on the user's initial input without having to re-engage the user at any point in the process. So that would be all the way from discovery to choosing which product actually gets purchased to making the transaction.

That's what I've had in my head when people talk about agentic ever since the term, you know, entered the greater conversation a few years ago. But as we'll get into later, that's not necessarily the definition that the industry has landed on right now.

Floyd: Okay.

Matt: I think the same way, Paul. And of course I'm coming at it from a B2B perspective, and I'm thinking about where agentic happens. And I'm also, I don't know, I'd love to hear your thoughts on this, that in some ways, and in some circumstances, B2B can actually lead the way when it comes to that definition of agentic commerce because it's not relying on the feedback and the preferences of a, say, a consumer who's shopping.

Instead it's just looking at business rules or replacing orders or, you know, replenishment. So, I'd be curious. Where do you see, in B2B, that definition of agentic commerce actually having some legs to it?

Paul: I think that's fair that there is a big place for it in B2B, and the overlap or the comparison to D2C would just be repeat orders, and that's where I'm most comfortable. I've been a traveler for a long time, so I've decided on the pants, the shorts, the shoes, and the shirts. Like, my whole wardrobe is just a repeat of the wardrobe I had the year before.

And so I buy the same things over and over again. I think a lot of consumers are the same way. And so reordering fits right into B2B in the way that you described. I think, though, the bottleneck is the platforms on either side of B2B. I don't work hugely in the B2B space, but I do have some B2B clients, and they are just, for the most part, decades behind the retail side.

And so while I do think you're right that the type of orders are conducive to agentic commerce, I don't necessarily, you know, again, not my world, but as far as I've seen, it's not as up-to-date on the platform side as it should, as the retail side.

Matt: Yep, exactly. And I would agree. We have some catching up to do, theoretically, possibly. Theoretically, we could go fast, but theory and reality, two different things.

Paul: What's interesting though about agentic on the discovery side of B2B though, is that right now a big problem has been taking the catalogs and all the information that's inside the sales reps' heads and putting them into data, and it gets really complex really fast. I'm working on a project now for a company that sells like the most niche B2B tools I've ever seen in my life.

I had to go through a huge learning curve just to figure out what exactly they sold and who used them. But it's like there's no perfect solution that exists on the market, and I feel like we've seen all of them to try to take their setup when it comes to inventory sharing across products.

They'll open a box in their warehouse, pull out an O-ring from one product, put it onto another, so they need to keep individual inventory counts of all that stuff. So my point is it gets very complex really fast, and I think what AI can be really helpful for, it doesn't just have to search product, structured product data in the same way that it does retail.

You can actually just upload all these disparate PDFs and spreadsheets and stuff, and it can use that as an underlying way for discovery. Whereas I feel like on the retail side, much of the agentic commerce on the market today is dependent on product enrichment and product data. They've been like, "Hey, here's our 64 rails. You fill in the information. Our AI systems will use those rails to extract the information."

With B2B, it's like, I don't know if we'll ever find 64 rails or however many rails that fit every single B2B business. So I feel like the end result is we're creating an AI level of discovery that works in a different capacity and in a more complex capacity than from the retail side. But again, that's not my world. I speak to it just from a few projects.

Floyd: I think what the risk of defining agentic commerce or where that risk lies is that if you ask a company what is agentic commerce, they're gonna tell you what their tool does, right? They're gonna say, "Well, agentic commerce is this specific tool I built that does this and this, and that's the definition."

But I'm more interested in what is the user's definition of agentic commerce. As a consumer, I think it could be, at the loosest definition, it's like any transaction where AI has touched it at any point. Is it like AI discovery and it leads you to the checkout page? A more strict definition could be, it is, you know, commerce where software is picking a product for you and it brings it back and says, "Eh, is this the one?" And then you give it the manual permission to buy. And then the strictest definition is you give it parameters and it goes off and does the thing and spends your money.

And I don't wanna make Amazon responsible for defining everything 'cause obviously they've got their own issues, but, to me, I feel like they've gotten way ahead of this. They patented one-click checkout in '95. They've always been a bit ahead of the curve.

And to me, I feel like how we've been talking about agentic commerce, like Subscribe & Save kind of almost meets that definition, doesn't it? I'm giving it parameters. I'm saying, "This is the product I want, and you can decide what day it shows up and how much I pay for it." Is that fair or is that way too loose to be agentic?

Paul: That's your definition? We hadn't heard yours yet. So that's maybe Amazon's.

Floyd: I don't know yet what my definition is. That's what I'm here to find out.

Paul: It's interesting 'cause obviously there's multiple stages of agentic that people are using the term to describe, and the biggest one right now is the discovery side of it. But really, discovery at this point is just glorified search. It's multi-step search. That's all it is. Then it's basically assisted search, and on the other side of it right now, we've got one-click embedded checkout.

And so it's great if it does a better job than human on the research and discovery side. The one-click checkout isn't really new. I mean, Amazon, they recently launched what they call, you know, a type of agentic commerce where you can set the price you're willing to pay for a product, and then if it hits that price, it automatically buys it for you.

And it's like, well, is that considered agentic? That's been around for decades. People have been doing that in the stock market for years. There's been other tools. It's still one click. It's just a pre-approved click happening at an earlier stage. But there's always a human in the loop at every single part of it, and human in the loop is actually one of the terms that they're starting to use to describe agentic.

So I feel like, like I said, my vision and maybe the vision that we were sold a little bit in presentations and pitch decks early on was completely human out of the loop, hands-off, personal assistant, super shopper, all these different terms that have floated around. And what you imagine is just being able to tell the AI, "This is my needs," let's say for a telephone or something less big of a purchase.

You know what I mean? Just anything. I need a shirt to go to work. You tell it what you want and then you receive it in the mail the next day. It knows your preferences. It knows that you have a Prime membership or a Walmart+, so maybe it gravitates towards those websites to get you the product faster.

So it uses logic, what it knows about the customer, to find you the best product and just you get it the next day. That's, I feel like, what we were pitched, but what the industry has kinda landed on is just using agentic commerce to get it to your screen, and then you make the final decision from there.

And like I said, it's still helpful. I think it's great. I use it all the time. It's a faster way to search. What's interesting, though, is there's been some surveys that have come out, where folks have looked at the results. It's easy to take the results for granted when you ask Google, "What's the best tuxedo rental service nearby? What's the best crockpot for my needs?" You see the results, you're like, "Okay, here it is. You just saved me a lot of time."

But the trade-off people are finding is that they're not saving a lot of money. This is anecdotal, but there are surveys coming out, quite a handful of them, saying that the results of agentic commerce isn't the same results that a human would have found if they went looking for it by themselves. Have you seen those as well?

Matt: Yeah, exactly. And I think it kind of speaks to this whole issue of trust and, I would say, the difference between capability and whether or not the people that are utilizing these capabilities or implementing them, if they're a practitioner, are actually ready for it.

And so by your definition, agentic is no human in the loop. The human is in the setup essentially, defining rules, defining preferences. But then when I prompt it, that's it, shows up at my door. And correct me if I'm wrong, but that's your understanding. But if that is your understanding, and that is where we're heading, I wonder if we're ever gonna be at a point...

'Cause human beings, just from an anthropological perspective, right, we crave control. We want control. And are we ever gonna be at a point where that agentic reality in our day-to-day lives is something that we're willing to outsource?

Paul: It's also emotion, I think, too. Fair point about control. I think the two go hand in hand, having control, feeling emotion. Shopping is a pleasurable experience for a lot of people. That's why malls exist. That's why e-commerce exists. Like, you enjoy the journey. You look at T.J. Maxx as a retailer, their entire business model is built on what they call that treasure hunt experience.

And they kinda coined the word treasure hunt, and they built their whole business around it, but they're not the only ones. I think most retail is built somewhat on a bit of a treasure hunt. You go in, you discover, you see things you weren't necessarily looking for. And yeah, AI, in that... I guess efficiency is the enemy of that emotional journey it takes away from you.

So I think there's a fit for some products. Like if I just wanna, like you said earlier, reordering, what a perfect example of using agentic commerce in action.

Matt: Yeah.

Paul: I buy the same brand every single time I go into a grocery store. I saw some examples of reordering at restaurants. I'm one of those people that go to a restaurant, and if I like the meal that I order the first time I go, then that's my meal. So now it's like I go back to the...

Floyd: You know it's good.

Paul: So it's like, yeah, order that meal, just don't forget to ask for the side of barbecue sauce and honey mustard.

And so I shouldn't have to tell you that every single time. Level one is I tell you what I want in my prompt. Level two is you learn from my prompts, and I no longer have to tell you. I order from somewhere new, and you know that Paul likes barbecue sauce and honey mustard, and you preemptively order the sides when the meal comes with french fries, for example.

Your integration layer holds your e-commerce stack together, and your best engineers spend most of their time making it work. Pivotree delivers integrations in half the time and at half the cost. Then we watch them twenty-four seven in production for a fixed monthly fee. Get started with a free integration sprint.

Learn more at www.pivotree.com/integration. That's www.pivotree.com/integration.

Floyd: I think that there's a really interesting possibility on the horizon because the B2B buying process, not in every case, of course, but it often is less of this treasure hunt emotional purchase, right? If I've got a bill of materials and I need to source all of these parts and I need to source them to a certain, you know, price point, that's certainly something that an agent could help with, if not do entirely, and I wonder if retail will ever get there.

And I've done this recently, right, is I'm renovating my kitchen, and I don't mean to turn this podcast into, like, a recurring soap opera about my dishwasher. I am guilty of that. But I did buy a new dishwasher and a new fridge and a new stove, and yeah, I used AI to help me narrow down the list of possibilities.

No way, no possible chance would I let AI actually go and check out for me before I had a chance to really absorb the emotional weight of this purchase. I'm wondering if I had done that, if I had said, "Okay, Claude or Gemini or, you know, any of the shopping tools, here are my requirements for a dishwasher. Go buy me one," and the wrong dishwasher showed up, who gets in trouble for that? Does the merchant absorb that? Is it on the customer who used an agentic shopping tool? How does that play out in agentic commerce?

Paul: There's gonna be, I guess, a lot of friendly fire in the meantime while the industry kind of decides that. There's a woman named Heather Hershey that I follow. She works for IDC. She was just talking about this recently, all the dangers of true agentic commerce, and one of the biggest ones she pointed out was talking about how the liability of chargebacks and returns ultimately lies on the merchant, not on the tool that the merchant is using.

And I think because of that, that's why the industry has sort of landed in the middle. Even though at the beginning it was very much like hands-off, autonomous, don't touch it, personal shoppers, we do everything for you. Where we've landed is that human-in-the-loop, consent-based, approval-based, and I'm fine with that.

When I talked about my definition of agentic commerce at the beginning, I guess I was going back a few years. That's what I was led to believe we were going towards with agentic commerce. That's not necessarily what I want as a consumer, and I would imagine that I'm not so different from a lot of regular consumers, too.

And maybe because of that, it's always, you know, a few steps back from what you promised, and that's okay because this is where the industry has settled on. So I don't know about you guys, but I'm happy with using it as a discovery tool. The actual checkout and decision process has never been the big bottleneck in shopping for me.

The decision process is the gratification part of it. I don't wanna get rid of that, and the checkout part of it couldn't be easier. I have absolutely no complaints with my browser storing my credit card or using a digital wallet to make a payment. For me, it can't get any faster than that, or I don't really need it to, I guess.

Matt: Yeah, I was just thinking about, so if we break up the term agentic commerce, and we think about it as agents executing workflows within different stages of the buyer's journey, I think that can be extremely helpful. And, you know, I would probably lean towards that's where we're heading because we can cut out the analysis paralysis, right, of the cereal aisle, as an example, right?

So we're able to narrow down, you're able to narrow down all of the refrigerators, all of the dishwashers out there to a subset and then enjoy the experience of the emotional attachment to that purchase. And that's tremendous in and of itself. And then, you know, the whole shopping cart process.

So I look at it as segments of workflows that are now you're able to completely turn over to a well-trained agent. And, I mean, there's value in that. But I'd also say that there seems like a lot of pushback around AI encroaching on human, you know, activity.

I even think about it in just as simple as this example. There is this weird thing happening. You know how we said at the beginning, okay, Paul writes all of his newsletters, right? There's no AI involved. He writes them. There's this weird pushback about, if you use AI to write anything, you're like the scum of the earth. You know what I mean?

And so it's like we're taking back viciously control of our writing and our content and I don't know. I am curious to see where this goes. And, with enough of that rant, Paul, I'd love to hear your thoughts on the future and particularly if you could get into some of how these different tools like OpenAI, like Anthropic, what are they doing today? I know you're on top of all of that.

Paul: Yeah. And honestly, I think we're here. I think the future arrived and we've kind of reached a stabilization point where we're not necessarily gonna be introduced to a whole new way to shop next year or the year after that. I think now we just need to catch up and improve what's already on the table while we wait for mass adoption.

Because I think you and I, like the three of us are in the world. We try things out intentionally. We're early adopters. We look for this stuff. We talk about it all the time. And so there's like this assumption that the whole world knows what's going on. But if you talk to some friends that are outside the industry... My mom had been sending old photographs to a Facebook group and people were colorizing them for her, and she got a kick out of it, and it was awesome.

She would send me the pictures on Messenger and I'd be like, "That's really cool." It looks like the pictures were taken yesterday, but they were from the '50s and the '40s. And I said, "Hey, Ma, you know, they're just using ChatGPT." She goes, "No, it's free." I said, "No, I know it's free, but if you want, you can just use ChatGPT too."

So my point is like, here's my mom, she reads my newsletter every week. She's, for someone in her 70s, more aware of the industry than other people, yet she had never used ChatGPT or really any kind of AI that wasn't already baked into the tools that she's already using. So she started going to ChatGPT, started using... putting up her own photos and then, oh my gosh, she sent me like 30 photos in two days 'cause she didn't have to wait for other people to use ChatGPT for her.

And I give that story just to say that I think that's where we're at with agentic commerce in general. We talk about it all the time. We expect it to be mainstream. The tools are there, the foundation is in place, but there's two things missing. One, the actual output still kind of sucks. It's getting there, and you can use stats and say, "Oh, well, 90% of answers, 80%..." That's not good. Like, anything short of 99.9 is not good.

So the output, I still don't trust it. It's saving me a lot of time, but I'm still doing my own follow-up research because it's necessary. That's one thing missing, and we'll see that improve. Look at just whatever you use AI for, whatever examples you have in your head of how you guys use AI in your work every day, I know for a fact it's better than it was two years ago, and that's what we're gonna see on the agentic commerce side too.

It's gonna be the same thing. It is going to get better. They'll build new tools, better comparison tables, better carousels, better charged multi-step whatever. But the actual thing that we're moving towards, we're already here. It just needs to improve. And then the other side of the coin, we just need, not us nerds to jump on board with this stuff, you know, just regular people.

It's happening, but not at a level that, I mean, just look at analytics of any one of your clients. You see how many sales are coming from AI versus coming from the other channels. That speaks for itself. You don't necessarily need my anecdotes to say one way or another.

Floyd: So I think that all three of us kind of have the same understanding that in order for something to meet the dictionary definition, although I'm sure it's not in the dictionary just yet, of agentic commerce, it has to be kind of everything from discovery to purchase has to be done by an agent.

And if we're looking at it through the consumer lens, you know, regular people who are buying from DTC companies, they're getting their groceries online, they're shopping for their widgets on Amazon. They may not use the term agentic commerce, but they're thinking about this new way. It's pretty limited to the discovery phase of that purchase and narrowing down your options.

Do you think that there's a fundamental disconnect, then, between what the average consumer uses AI for and wants to use AI for, and what the big e-commerce companies are actually building for? Are they all heading in the same direction, or are the tech companies trying to push people in a direction we just don't think they're gonna go?

Paul: Always a bit of an ebb and flow. They push. We don't want it. They take a step back or they quit too early on it. But I think we're aligned now. And it happened quickly too. It's like the alignment that used to take a decade now happened in a matter of years or even months.

I've been really impressed with Google's Universal Commerce Protocol. I remember the day that it launched, the first things that popped in my head about, "Oh, why doesn't it have this? Why is it missing that?" One example was being able to add multiple products to a cart. Seems so obvious. And I'm like, "Why didn't it launch with that?"

But the list of things that it should have or could have launched with would've meant that it would've launched months later. But then within several weeks or several months, they kept coming out with those new things. So I feel like, you know, the complaint with ChatGPT when they launched Instant Checkout, that was trying to push people way to a level that they hadn't experienced before, that no one necessarily wanted because they didn't know that they wanted it yet.

And the reason they didn't know that they wanted it yet is because they never reached a point with the discovery part of the journey where they're like, "Oh, I..." The day that a consumer gets to a point where they say, "Just why don't you just buy it for me?" That's the day that we're there. But until then, that means that the technology isn't good enough and it's just one of many tools that we use.

Floyd: Got...

Paul: Do you want me to give you some examples of what's live on the market? I came prepared.

Floyd: Oh, yeah. All right, tell us about what's out there. We'll cut it in somehow. Let's go.

Paul: Okay. And correct me if I'm wrong on all this stuff, 'cause I know you guys are using this too, but ChatGPT right now, purely discovery. You look for stuff, they help you find it, they help you compare it. The actual checkout handoff, either they send you to the merchant website or if you're like Walmart or you're running a ChatGPT plugin they call it now, they'll handle the purchase for you.

But technically it's just like a buy button. It's just an embedded checkout. That's what ChatGPT is at right now. Alexa, probably the closest thing to agentic commerce, just because you can actually give it instruction, to purchase something, but it's very limited use cases.

And so, the example I was giving earlier, you can say, "When this particular product hits this particular price point, you have my permission to buy it." So the actual purchase is happening agentically, but it's really just, like I called it pre-approval earlier to one-click checkout, but pre-approved at a certain price point.

None of these tools yet are taking your input, deciding what to buy for you, and then making that decision and finishing the transaction. Every single tool has a human in the loop at one point, either earlier in the process or later in the process to approve the purchase and say, "Yes, you can have my money. This is the product that I want."

And that's why you were asking about the risks. I think some of the risks are overblown right now, like with what Heather was writing about. Yes, the merchant is the merchant of record, and they're the ones that hold the liability for the chargeback and the returns. But as far as I've seen, there isn't a single AI agent that's making purchases autonomously. And so it's kind of a non-sequitur.

The other thing I wanted to talk about is you guys are familiar with the Amazon versus Perplexity case, right?

Floyd: Yeah.

Paul: Effectively, Amazon didn't want Perplexity's AI agents shopping on their website. They tried to block them. Perplexity would circumvent the block by pretending that they were a Chrome browser. So anything Amazon tried to do, Perplexity went out of the way to circumvent that and shop anyway.

It made it to court. Perplexity's whole argument was that we're a tool. We're like a wrench. You know, you can't stop shoppers from choosing their own tools. And that's what their whole argument is built on.

It went back and forth. First, they got a temporary injunction. They said, "All right, Perplexity, you can't shop." Then that was reversed, and it was like, "Okay, Perplexity, you can shop." And then I think in the most recent judgment, Perplexity is winning, and so they're allowed to shop on Amazon. But the whole basis of winning, it comes down to the legal definition that is slowly forming. It hasn't formed yet. We're still circling around it 'cause this case is still ongoing. But ultimately, the definition of agents are tools, not people.

That is ultimately the legal protection that they need. So the whole idea of the AI agent not being fully autonomous is what makes it legal or what is starting to make it legal. Does that make sense?

Floyd: Yeah, and they're taking up the same kind of fight against Muse, Meta's new kind of personal shopping agent. That might change by the time this episode comes out. It seems to be updating every day, but I think that's what they've...

Paul: But I don't think they're fighting though. I don't think Muse did any or Meta did any pushback about it, right? They just said you can't shop it and they're done. As far as I know, Meta isn't trying to circumvent Amazon's...

Floyd: No, no, but it's another example of Amazon kind of being selective about the tools, likely because they've got their own, right? They want you to use Amazon tools on the Amazon Marketplace.

Matt: Yeah.

Floyd: It's just interesting to me that the definition legally, like I said, it can't be fully autonomous in order for it to win these cases. And so it's like all these tools are... I don't know if that's the reason they're taking a step back from being fully autonomous, but both are happening simultaneously. So we'll see what happens.

Honestly, the laws are archaic. We're talking about 30, 40 or older year-old laws that have nothing to do with agentic commerce that they're trying to use to argue one way or another, and this all needs to change, but that's, I guess, a conversation for a different day.

Matt: It's the solving the liability and the legal definitions is probably going to be a turning point in adoption, especially from, you know, the larger retailers, because I think about it and I'm like, well, yeah, it's one thing to have a $2,000 return for a dishwasher.

But then I also was thinking about other businesses where more and more is on the line, right? It could be healthcare products, it could be, you know, the medication, refilling prescriptions, and B2B, it could be a six-figure order, right? That gets sent back to the dock. And so the liabilities around agentic commerce, I think it's something that we're still trying to figure out, and until we figure it out, right, the best we can do is human in the loop and a faster, more frictionless experience.

Paul: And beyond liability, there's also unscrupulous use cases that you can do. Imagine if you were like a WordPress developer and you were like, "Hey, AI, go buy this plugin for me. They have a 30-day return policy. Download it, fork it, and then issue a refund for me." You know what I mean? So you can do unscrupulous stuff like that more.

Or what if the AI starts doing that on its own because it has access to your wallet? So it's like, "Hey, AI, build this for me." They find it, "Oh, it exists already in this capacity. I'll just go buy it and then I'll return it and that'll make my user happy." And I would say it seems far-fetched, but it's not even really that far-fetched when you read about the stories of OpenAI and Google and Anthropic and how they're making decisions on their own to go outside of their sandbox. And they don't have morals. They don't have ethics, so it's like they don't know what's right or wrong. They look at a task as binary.

Matt: Yeah, I listened to Zack Kass quite a bit, and he was just recently talking about this. He said, in his opinion, we could pause our AI development right now and it would take two or three years for the guardrails to catch up. And because we're just... The ability is there, but whether or not it's safe, whether or not it is, you know, regulated, you know, the legal side of things, the government questions that have to be resolved.

These are all things that if you are following AI and agentic commerce, these are the type of things that you'll wanna have Paul's newsletter to help you navigate. So yeah, there's a lot changing, seems like week to week.

Yeah. But don't worry, I think we can trust the AI tech companies to make the best, most moral and ethical decisions for us, the consumers, right? Oh, yeah.

Floyd: That's what I think. And I would like to thank Matt for assuming that I'm the kind of person who can and would buy a $2,000 dishwasher. Sadly, I didn't go quite that high end. And, Paul, this has been fascinating. I think we could talk for three to six hours about this, but our listeners probably don't have that much time today, so maybe we'll get you back on the podcast at some point. But until then, can you tell our listeners where to find you if they're interested in knowing more about e-commerce news?

Paul: Oh yeah, shopifreaks.com. S-H-O-P-I freaks.com. You can sign up for the newsletter. It's free, and then you can hear my opinions for better or for worse week after week.

Floyd: Perfect. Thanks so much.

Paul: Thanks, guys.

Matt: Thanks, Paul.

Announcer: Thanks for tuning in to this episode of Data vs. Commerce. New episodes drop weekly. So if you're responsible for any part of how products get from a database to a doorstep, subscribe now on Apple, Spotify, or wherever you listen.