Customer experience is becoming autonomous - Interview with Andrew Bialecki of Klaviyo

24:31You know, I know what the pair I want is, but also I want something that's like, hey, we'll show up in the next couple of days, not in two weeks, right? Or something works in stock. I mean, I could go browse 10 websites myself, but I can also just have an agent go do it for me. And I think this is some of these are some of the use cases that we're already seeing. I think when you're designing experiences, it's smart to design them in such a way that, you know, agents can understand. And so practically, what does that mean? Well, we're envisioning a world where when you go visit a website, when an agent visits a website, it may not want to browse around the entire website.

25:00It may want to just connect to, say, your customer service agent, the customer agent product that Clivio offers, and just ask its question directly. It's like a form of search, but it may then want to say, like, okay, well, I'm ready. Can you tell me availability of this? Can you tell me what colors it's in? Maybe you can do some of that research. Hey, Andrew, he's in Boston. He's in the Northeast. It's kind of hilly. How are these trainers where there's lots of ups and downs? Can I ask all those questions? Come back and bring that information back to me to make the final decision. So we expect that's going to happen.

25:39Now, to your question, I'm like, hey, humans, does that mean that we're just not going to do any buy anymore? It's interesting. It starts to press on a lot of these things of like, well, especially let's take, you know, kind of shopping as an example, right? I think there are some purchases that we make that are more transactional, right? If I'm going out to buy, you know, milk, bread, groceries for the week, like they're actually, I probably know what I need. There's a little bit of serendipity, but a little bit, you know, less so. Certainly we've seen like, I mean, one of the reasons that like social advertising works so well is a lot of folks are in their kind of like in lean back mode.

26:12They're browsing content. And, you know, they're looking to be entertained, looking to be a little bit that serendipity, as you said. And there, I think, you know, we as consumers really want to just sort of get surprised, see something you weren't standing for, a little less direct intent. Even there, I think agents are going to do some really cool stuff in terms of, you know, maybe showing us things we didn't see or in a format we didn't expect. So I'll give you a little sneak peek. we have a little project we're working internally that is like basically alternatives to how an inbox, like think about your email inbox or text message inbox works.

26:44Or is that like, you know, hey, a lot of people like, you know, Instagram or TikTok or some of the other short term video sites. What if we could take all the messaging and the products, you know, that you got from the brands that you love that you're, say, subscribed to. And instead of having to fit into the form factor that we're all kind of are used to, right, with, you know, email or text messaging or WhatsApp, what if we could say, hey, an AI actually went and read all through all that and it generated an Instagram style feed or a TikTok style feed of short form video, pulling from assets that those brands already have.

27:15It's just a more engaging way to browse and shop. So maybe instead of seeing a static image of a product, it could show you a video that was on the website, but it wasn't contained in the message, but the AI could easily find. Right. So I think we're about to see an explosion of some real creativity in the kind of the consumer experience when we're browsing and shopping. There's this really interesting thing. As I show this clip from probably the early 90s from one of like the major U.S. news outlets. And they were showing the first websites that were doing retail. And I remember they showed this video where it was like you could be a little avatar.

27:51And it was if you were walking around a mall, you know, like a physical place where there's all these shops. and so you'd click to walk in front of each shop, right? And they thought, oh, this is the future. Instead of walking through the mall, you're going to literally click your way through the mall, right? And we kind of laugh about this now because it's like, that's such a terrible user experience. Like, I mean, a mall is great because it's all these physical places, you know, physical retail places in one spot, but the walking around is kind of painful. Like the whole fun part about having it on your phone or your laptop is you just pop open a new tab and you instantly teleport it into, you know, a whole new store, right?

28:26Yeah, we're going to see similar things with AI where the current website, the current experience is going to feel a little archaic. And we're going to be able to use AI, though, to repurpose, repackage that into experiences that are really great for when we're, you know, we have less intent and we're sort of we're looking to be either entertained or maybe, you know, we're looking for a little bit of that serendipity, as you said. excellent i mean yeah no it's interesting because you know you're talking about kind of like go back in the day and about kind of that that internet shopping and kind of like almost like you have to click to walk around a mall and i just it reminded me of almost the the real estate market where you end up with these virtual kind of like tours and you've still got to click around to visit different rooms i'm like there's that's so clunky and it's like there has to be a better way to that why can't i just put my finger on a trackpad and just move it around you know that sort of thing yeah i'll give you one more example like one thing we're playing with with our customer agent is this kind of like you know a little bit chat interface we're actually plugging it into the search result pages for some of these brands so let's take my you know um you know my uh you know my uh hey my trainer example of like i'm buying a new pair of shoes you know and say i say i'm looking for a new you know a couple of new kits or something right looking for some new shorts some new shirts right to go with it i mean now instead of you think of a lot of retail sites, especially if they have a lot of products, you have to search.

29:49And then I might have to search by tops, by shorts, by shoes, all separately. I mean, we're playing with things where what if we just rewrote the way a search results page looks so you could search straight up and say, hey, I'm looking for a whole new kit, new outfit, new gear for running this season and let it put together all the pieces. And oh, by the way, I want you to show them on a model. and oh, I want the model to be yay tall, right? And can you show me a couple different variations? And like, that's it. I mean, this is all, I think this is, it's not only possible, it's coming in the next few months this year as a result of AI.

30:29And I mean, it's brilliant. I mean, we will, I mean, excited to see all the kind of the developments and definitely there's a couple of watch this space on their kind of marker on that one. The final thing I wanted to ask you about is, and I want to circle back to that sort of vibe marketing kind of idea, because you did say you were moving from this idea of vibe coding to vibe marketing. And the big question I had around that is how are enterprises, particularly larger kind of, well, any business, but particularly larger businesses going to govern that sort of like stuff? Because that feels quite real.

31:09I mean, it's a bit like you can speak to the thing and it's going to go and then go execute and then it's off and going and you're like going risk and compliance is going to be all over that yeah like catching butterflies yes so what's interesting i think we're you know i think where everybody's learning quickly is is like you know when you have this explosion of you know we talk about generative ai and vibe coding and now vibe marketing i think we're seeing vibe a lot of different you know kind of jobs i think a lot of our minds immediately go to it's like well how do we prevent, to use the AI term, slop, like low quality, poor things.

31:43And especially, look, if you're building on behalf of a business or brand, like, look, this is, I mean, how I represent myself to my customers, to my end consumers, this is everything. Like if I make a mistake, that's almost way more painful than just doing things the old school way. So within our composer products that helps build marketing campaigns, one of the features that we built in there is this thing called guardrails. Right. And what guardrails do is, I mean, a little bit like what sounds like if say you're on the highway is it says like, yeah, it's like, look, if you have a self-driving car, like you want to make sure that it stays in its lane.

32:15It doesn't bump into the thing in front of it. So what we've done is within guardrails, we've built, you know, dozens of guardrails that basically say, hey, look, we just, we won't allow or constrain the agent and tell it to check itself against the following checks. And these are not probabilistic checks, like a lot of AI. It must pass them. And then as a business or brand, you can add more checks. And again, these are things that are like, hey, hard rules in these situations. We would do this or we would never do this. Or you need to put a legal disclaimer here. So interestingly, what we found from our guardrails feature is it was very popular for AI-generated content.

32:56It actually was equally popular for human-created content. So I'll give you an example. I was talking to, you know, when I was over in Europe, you know, last fall, I was talking to the second largest athletic retailer like in the world. Right. And one of the things they told me was they said, Andrew, like if you can build these guardrails for AI, let me tell you that actually will speed up a lot of our marketing teams because we're a, you know, international global, you know, athletic retail brands. You know, we're in 50 different sports. We're in, you know, dozens of different countries. And we have teams for every single category and region.

33:40And right now we have this huge approvals chain where if you want to send out marketing, you need to sort of like, you know, you need to pass it through a whole bunch of people that will check it. And it's very slow. So these marketers that are working on, say, you know, say it's like, you know, basketball here in the States, right? Or maybe football in the UK, like they have to wait for these like kind of central authority to decide if their stuff is right. And so the queue gets very long because it's all human validated. They said, if you can build this, we can give more control to those local teams and we can allow them to be more creative and we can give them.

34:18And they said, you know, things that we used to have to spend time for and pull people together physically to teach them about our style, right? Our brand, our voice. We can actually just have the AI grade how on brand they are and give them feedback and tell them, hey, they're not allowed to ship until it meets a certain quality bar. So we've found that actually these guardrails are equally important for AI-generated content and especially for larger businesses, right? That may be multiple product lines, multiple brands. They want this just as much because it enables their individual teammates, right, to go faster.

34:52So I think you're going to see that. Like, you know, one of the beauties of AI is we can use it to generate things, to be creative or take our creative traits of being extended. We can also use it to help us, you know, do a lot of the auditing and checking and reviewing that, frankly, we as humans either don't have time for, or in a lot of cases, just don't have the patience for, because it can get pretty rote. And we can use AI to help us, you know, make sure things are correct. And then basically, you know, either block or flag or, you know, escalate things that, you know, maybe look iffy or it's not true about.

35:21Yeah. I really like that. That idea is like, rather than using kind of AI to create autonomous content, unless it automatically, it can actually be used as a tool to, I feel like, accelerate something which is a very human kind of process. And that approvals thing is like, I'd imagine there's people listening into this and going oh getting approval for things is just it's the it's a blocker in many cases you have to wait for the right kind of like time get time in people's diary don't know what their other priorities are and then you might have wanted to send something out which is going to capture the moment or something and then you have to wait for approval and then you get approval then the moments pass and you're like going yeah i phoned me sort of thing and so I'd imagine as people listen to this going, ooh, that might be quite interesting.

36:17So that's kind of an interesting kind of feature. So it's not just about the software doing the software sort of things, but it's actually also how can it help the very human side of organizational kind of life as well, which is great. That's really interesting. A couple of final questions before we can do some quick fire ones, Andrew. Give me an example of a brand or something that you're working with. Well, obviously the ones that you think are killing it, right? Or they're leading the way. and just to help us understand a bit more about kind of like what they've done what sort of challenges they've faced and what sort of returns they're generating because all this stuff it sounds great but then sometimes kind of like you know you've always kind of it's good to help people rather than just inspire people but also help people understand what they can be aspiring to yeah so a couple of examples i mean we talked about vibe marketing and we want people to feel with AI that they have more safety in being able to be more aggressive with the ideas that they try and know that AI, that software has got their back.

37:19So with that, my big push for folks is if you have an idea, what we hate at Klaviyo is if those ideas kind of die, maybe, I don't know, in your head, right? They die in a spreadsheet. And instead, you should try to go for them. So let me give you a couple of examples of businesses we think are really pushing it in some of the results we're seeing. Okay. So one is a health and skincare brands, Allison's. So they were one of the early adopters of actually the first generation of our composer agents, you know, our first generation marketing agent. And, you know, with skincare, there's a lot of, you know, people have skincare routines and people want to understand like, hey, you know, well, how, like, not only which products are right for me, but how should I be using these?

38:00So content marketing matters a lot. So for them, you know, they had experimented with content marketing, but it was hard to find the right person to write it, to generate it. And so with our first generation, you know, composer, they would use that to generate, you know, hey, let's go summarize skincare routines that they had curated and tell the story, give people practical how-to tips, right? And then actually personalize those to people based on, you know, different skin conditions or different skin types. And they built these campaigns and then they, you know, they assign them to different customer segments based on, you know, different skin types, right?

38:35And maybe how new somebody was to the brand, right? And the kinds of content they were interested in. And they sent those out to their subscribers. And this was, this was, you know, in this case, augmenting, kind of replacing what they had been doing with, you know, where they'd have somebody on their marketing team would try to write up, you know, hey, here's some, here's some tips, right? You know, once a month, right? Instead, they were using AI to generate this content. Those content campaigns generated 50 % better open rates. So not only was the open rates better, it generated on a per recipient basis, 40 % more sales.

39:11So what's amazing about this is, I mean, we have to pause, like 40, 50%. These are like, I've talked with customers that have iterated on content and segmentation to try to grind out 10%, 5%, 10 % changes. I mean, these are like step function changes in engagement. And why? Because, I mean, they know, they like, well, they didn't, you know, they didn't have the budget to hire a full-time, you know, content team. You know, they had limited time. They knew what they wanted to do, but it was an idea. And now they were able to use AI to actually execute on that. And this is one of the patterns I love is Allison's is, hey, it's not just about the quality of their products.

39:50They've got great products. but also we can be value add to our customers in terms of their everyday life. Show them how to use their products to maximize their efficiency. It's actually a pattern we see quite a bit. There's another brand, a swimwear brand called Kulani Zikinis that has taken our customer agent, so customer service, and is able to say, hey, not only can we help you with the size and fit of a swimsuit, so if you can't try it on the store, we can tell you what the sizing is similar to, But also they're doing some things there with their customer agent where they're helping people.

40:23Like if somebody has questions about planning out a vacation, and maybe they're trying to decide, hey, I'm going away for a week and a half. Should I bring one swimsuit or two? They're actually helping also pack their luggage, right? Right. So there's some really great things I think where brands in the future, they're not just going to be about their products, their services. It's also going to be about their expertise. And so you're able to offer that up. So last one I'll give you is Lifestraw. That's a filtration company that we work with, right? So it helps with cleaner water. Well, their products are actually used in two places.

40:59They're used sometimes in the home, right? So a lot of us, you have a home water filter, things like that. But they're also used when you're out, like people, hikers will use these. When they say, hey, they're out, they're camping and they're not near like say a clean water source and they need a way to like get clean water from the lands. Well, because they have these customers that are very different, one of the things they wanted to experiment with was should they segment out the content in their product launches tailored to these two different audiences? It's an example of the kind of thing that people would try out, but they had an idea around, but it was very hard to test.

41:36So they just ran that experiment of, hey, if we have different kinds of content, how much better is that performing? And again, similar to Allison's, they saw much better results. And then that same persona, we talked about marketing and customer service and how it's all the same. They've now embedded that into their customer agent so that they can answer questions both from like people that are very technical camping, hiking, as well as folks just for everyday home usage. And, you know, it's because, you know, it's water filtration. It's a little more technical. They've actually been able to get increased sales because people like to ask a lot of questions before they buy.

42:08And they haven't had to do that with scaling out their customer support team. They can just use AI to answer a lot of those questions. you know and even if it was some detailed home setup right you can ask a question about like hey here's specifically how in my flat in my home i use this like is this a good fit right and those are things that might not be in a product detail page but in ai can sort of do that you know reasoning for you awesome that's really interesting and some there's some great results in there as well so andrew that's it for my main questions um but before i ask you some of these quick fire questions.

42:39Anything that you'd like to add or highlight that I've missed out? I don't think so. I mean, look, I think I push everybody to think consumer first. And look at Klaviyo, we think there's so much human creativity that's been untapped. When we think about how do businesses, brands deliver stunning customer experiences, the trick is that we want people to feel like if you can dream up an experience, if you have an idea, you can make it possible. And so I think you should try to implement more of those dream experiences that you think your consumers want. They will result in more engagement and more sales.

43:18And also, I think AI has gotten to the point now where it's good enough that if you're not taking the time, whether it's through Klaviyo or otherwise, to ask AI what ideas it has for your business, I think that's going to replace or not replace, it's going to augment a lot of the conversations we have, like the peer groups we have, where we try to collect and understand information. I think that's something that we've gotten in the habit of the best brands. They're just constantly asking AI, hey, what's the best you've seen for how we should launch products, right? Right. You know, what if you've seen for how to deliver a great customer service, or, hey, when we should, you know, when we should do discounting or incentives, or we should offer a refund, all of these things, I think you can use AI to help you get a better sense of, you know maybe best practices where you're a little unsure awesome so on to my quick fire questions and so i've got three of them and the first one is a bit is a best advice question and i've been i asked people to to help with that i've been asking people to complete this sentence and the sentence is if you want to improve your customer experience andrew says do this dot dot dot complete that sentence excellent all right well if you want to improve your customer experience, you need to make sure that you find software that has access to your data.

44:29And I would say, I would also make sure that that software either has now AI agents built into it. It's obviously some of the things we're trying to do, or at least that software can be connected to chat, GPT, Claude, something that can be your reasoning partner. So make sure that software has everything you know about your customers in it, and that you have another brain that you can connect to it in case yours is busy doing something else. Awesome. So my second question is a punk one, obviously. And so which company or brand do you think takes a more punk approach to customer experience and why?

45:04Awesome. So I mentioned, you know, in the UK, Loop Earplugs, one of my favorite customers, great, you know, amazing products, you know, with their earplugs and very creative. Like they, I mean, they look at the world as, hey, the same way that you might have different belts that you wear with different pairs of pants or different sets of sunglasses, the same should be true, you know, for the earplugs that you wear, and they become like almost a lifestyle accessory. So there's different ones I wear to a concert versus, you know, when I go to the beach versus on my way to work. So one of the things that they've done is, you know, it's obviously like you have to help you all sort of build out their, you know, closet, so to speak, their wardrobe of like, what are what are all the different earplugs and varieties they should have.

45:43So they've used a bunch of quizzes and surveys and prompts and now increasingly the AI to help people map that out. So it's a little bit like planning your closet, but in this case for earplugs and earbuds. And I think that's just, it's crazy. It's very creative. And it's a fun thing that consumers like to do because it sort of gives them a chance to a little bit envision the future. And then they might not realize, they're like, oh, shoot, I actually do want a pair for this concert, this festival I'm going to this summer. Right. Anyways, because of that, they've had this kind of two-pound effect.

46:18One, it obviously results in some more sales to them. But two, they're finding that it's just getting their customers more engaged, where a little bit like we think about the clothes we wear as like, well, it's always on. I think about the brands that I buy from and for each occasion, they've now kind of taught their consumers that, hey, actually, you don't need just one pair of earplugs, you can actually have multiple pairs. They've done it through this very creative, kind of like fun, engaging customer experience. Nice. Then my final question is a request to end on a good news story because there's not a lot of them around.

46:55And so Andrew, tell me a good news story. Tell me something about something that's the most interesting, positive, or exciting thing that you've seen in, say, the last week. Oh man. Well, I'll tell you, like, so this is not, But as I mentioned, I was a physicist, astronomer, you know, before we got into Klaviyo. And I'm very much, how do we scale humanity, you know, across the solar system, the galaxy? How do we scale ourselves in terms of like, you know, share more of our product services, our insights with the world? So I'm very excited about, in the news, you know, there's just a, you know, a rocket launch, the US, actually international team that's going to, you know, go orbit the moon.

47:33moon and with the plans of like building, you know, putting human beings back on the moon, which we haven't done for over 50 years. I think that's fascinating. It's very fun. Like it's a very, you know, it's really makes the earth really feel like, you know, hey, we're one, one society taking on like, you know, the really hard, you know, physics and engineering that it takes to be in outer space. The other thing I think it's very cool is, you know, relative, to where we were 50 years ago, because I studied this stuff, we're relying more and more on software and AI. So actually when that rocket lifts off, it used to be back in the day when they first went, there really was a person, a man, right?

48:15Back then that was responsible for basically driving the rocket, like make sure it was aimed in the right direction. So the analogy is a little bit dragging, like a very fast race car. I mean, now we can do that, but it's actually not human guided. They obviously did all the studies And, you know, much like we talked about like self-driving cars, it's actually like, hey, the error rate of humans is just it's low, but it's not as low as with software. The software can adjust faster. So I think that's very cool that like we've been able to take some of the technology gains that we've developed over the last couple of decades.

48:45And it also just goes to like, you know, if you think about like, hey, what is humanity going to work on? What are we going to see in the next, you know, 50 years in this century? You know, there's so much, you know, in this case, literal exploring the world, the universe that we haven't done. And, yeah, I think that's an awesome story. And I'm excited to see what happens in the next couple of years. Awesome. Well, thank you for sharing that. And thank you for sharing your time and your insights and your expertise with me today. That's been super cool. Excellent. Well, thanks for having me, Adrian.

49:16Wow, what a great interview. I hope you enjoyed it. I know I did. Find out more about me and the work that I do at adrienswinsko.com. Do leave a review on your favorite podcast platform. and if you have any comments, feedback or questions about the podcast then feel free to send me a message to podcast at adrianswinsco.com and do tune in again. Thanks very much.