 Hello. Hello, everyone. I think I'm the last one up. So we're going to bring a lot of energy to this one. I'm talking about AI, so it's perfectly programmed. Everybody wants to talk about AI right now. Show of hands if you're using AI in your products today. It's like everybody's raising their hand. The live stream, everybody's raising their hand. It's a big topic for us too, and I want to walk you through a little bit about how we've been thinking about AI at Instacart. They've already introduced me, so we won't- I look after the product organization, and as part of that, figure out, do I need the hand mic? Sorry, I think they're going to switch me up. Thank you. This work. Can you guys hear me okay? All right. So as part of this, I'm looking after all of the aspects of our shopping experience of what shoppers who actually pick your order do behind the scenes, of our retailers, of our advertisers, and the truth is that we're thinking about AI in every single part of that. So let me walk you through a little bit about how we're doing this. Now, the reality is, AI has been hiding in plain sight for years, and the reason is that AI actually means many things, and one of the things that it means is deep learning, and neural networks, and machine learning, and lately we've been talking about this as AI, but I want to take a moment first of all to just walk you through how we think about all those other aspects of AI that have actually been around for years, and you might find some of these things familiar, both in how you've used Instacart and in your own products. So first example, someone's shopping on our platform, and we make recommendations for more things that they might want to buy. Pretty straightforward use case, heavy machine learning going into every aspect of this, and I think we can all agree this makes the shopping experience better. If you're buying salmon, you might want to buy a lemon with that, you might want to buy dill, you might want to buy certain other things that you always buy. If you're buying milk, you probably want cereal, maybe eggs, pretty straightforward use cases, and ones that we see in every product these days. We take this a step further, by the way. In search, when you go in and tap search, you could see the screen on the left, which everyone would see, a completely new user might see that. But you might also see the screen on the right, which is actually reorganizing the categories and the search terms that we're showing you based on what we think you might want. Now, there's a really cool thing we do here, which is this changes as you build your cart. So literally as you're adding items to your cart, we're using machine learning to inform what other items you might want, and I don't know about you, but when I use Instacart these days, I often find that I'm just tapping those. I never end up actually typing anything because the computer is able to figure out what I'm going to want pretty much faster than I would have typed it myself. There are other examples, and when I say that this has been hiding in plain sight, I really mean it because every pixel inside of Instacart is influenced by machine learning today. So great example on the left, looks like a pretty straightforward screen, right? It's just listing products in a store. Actually, there's a lot of science that goes into what you see on that screen. Not every item is available at the store right now. Some of them might be out of stock. Some of them might be available in one store near you, but not another. And we can't always trust the data that we get from external sources, including retailers, because they themselves don't always know exactly what's on the shelf. So we use a lot of machine learning to predict which items we should show you. I'll give you a specific example that's really exciting. If you're not getting your order delivered until tomorrow, do you want to see the inventory that's there today? Or do you want to see the inventory that we predict will be there tomorrow morning when we actually pick your order? This stuff makes a big difference to people's experience while shopping, and it's only possible because we use machine learning throughout this experience. Now let's just say for a moment that one of those items is not available. Maybe it's likely out of stock. You might also want us to make recommendations for other items that you would want instead, or if you're actually at the point where the shopper is going to have to make a replacement, we want to give them a signal as to what the best replacements would be for those items. And we're not just going to guess. We're going to look at millions and millions of orders that we get every single day and use all the data we have there about replacements and around what people were happy with and what they weren't happy with to make the next shop better than the last one. Everything I showed you up until now is the part of Instacart that you probably see every time you shop. But this applies behind the scenes as well. So our shoppers, people who earn a livelihood on the platform are extremely important to us, and we want to make the best use of their time so that they can earn the best on the platform. And so the app that they use, we call it the Shopper app, also uses machine learning for a whole bunch of different things. One of those things is we want to be able to predict when orders are going to come in and where they're going to go so that the shoppers can start navigating in that direction because the closer they are to the store, the more money they're going to make. So a great example of something that's happening behind the scenes every single day, which we might not notice. Once they've actually shopped an order, we also need them to deliver those orders, and often they're shopping multiple orders at the same time. So if we can help them deliver in the most optimal routing and with multiple orders that are actually all going to the same neighborhood, again, they're going to make more money, customers will get their orders faster. Machine learning is rampant in every aspect of that experience. And then when we look at receipts that the shoppers upload after they complete a purchase, we really want to make sure that there's no fraud going on and that customers are getting the things that they paid for. And so we use machine learning and vision to figure out what exactly was purchased and see if there's any sort of anomalies that we should take action on. Last example here. We spend a lot of money on incentives. A new customer that's trying Instacart for the first time probably needs a little help to get over the line to get really comfortable ordering their groceries this way. A customer that may not have shopped with us in a while might need an incentive to give it another try. Maybe they had a bad experience and maybe we want to make sure that they feel taken care of. So a lot of money goes through this system, which means we need to be really smart about who we target with these incentives. And so if we're going to do that, we need machine learning and AI throughout all of that. Okay, lots of examples there of what I would prefer to call machine learning because we're now truly at a turning point. And I hear this a lot and I think you probably hear it too. A lot of companies are talking about using AI but what they really mean is they're using machine learning to do the kinds of things I just talked to you about. But there is a difference and it's a really, really important difference. With machine learning, we used a lot of data to figure out patterns that we might have had a hard time noticing ourselves to better target products, to better organize information. The difference is that with AI, we're actually creating a brain. We're creating things. It's a creative process. If we're successful in using AI, if we're using it the right way, then we create use cases that weren't possible before. That's the thing that excites me most. Now, we're here to solve a pretty basic problem. What's for dinner? It's actually a really hard problem to solve and it's a problem that people have been trying to solve for decades. Now, in a pretty basic form, I could even argue that recipe books have been trying to do this for years. The reality is that if you get a recipe book and you get the same recipe book, you're seeing the same recommendations but you probably want different things for dinner and you probably have different habits and different family structures and all kinds of things in your household that would make you different. So how do we solve this at scale in a way that feels extremely personal? It's not an easy problem because there's quite a few different things that differ between different households. Your taste preferences, allergies, the kinds of dietary restrictions you might have or preferences that you might have around prep time. You might not have a lot of time to cook. Another person here might have a lot more time to cook. So we need to figure all of these things in to what is actually a journey, a journey that begins with inspiration. And this is where the creative aspect of AI really matters and that goes all the way through shopping and ultimately enjoying the food that you've ordered. This doesn't just apply to food either. The pantry needs to be refilled and you might run out of paper towels at a different pace than someone else might run out of paper towels. In fact, think about what would happen if we just averaged everybody's experience and said, we think you probably need new paper towels because you last ordered them three weeks ago and our data says that people replace them every three weeks. We'd actually be wrong more than we'd be right because one household, I have two kids, two little girls, I can tell you, we go through paper towels quite fast. We're gonna go through them faster than someone else might. So what we really need are personal answers to these questions and those answers need to be informed by a lot of data, not just averages. Really, if you look back at what we've done at Instacart for the last 11 years, we've taken the physical work out of shopping. We've made it easy for you to replace that effort that you would have had going to the store by bringing the groceries to you. But now with AI, we're going to start replacing the mental load associated with your grocery shop and make that debt simple. That's really fundamentally our goal. And so I talked about AI being the difference between AI and ML being formulas and averages and pattern recognition becoming creativity, actually generating new things. That's really what we're after here. So let me give you a few examples. Now, how many people think my next slide is gonna have a chatbot on it? You're right, it's gonna have a chatbot on it. But I just need to get this out of the way because so many people when they think about AI, they think about chatbots. And that's cool. And maybe it'll work and maybe it'll make the shopping experience a little bit better. So we've got a chatbot. And if you wanna chat with it, give it a shot. It'll give you good answers. But I actually think the real magic in AI isn't creating a manufactured use case that may or may not be the way that people truly engage. It's actually to change every aspect of our products to use AI in a way that produces a much better outcome than you would ever have otherwise. So just like I said that AI has been hiding in plain sight for years, it's possible that the future is still one where it's not front and center but actually it's making the whole experience better. So let me give you a few examples. How many people here know what the words hundreds and thousands refer to? Show of hands? Do we have any British people in the room? Are you British? Yep, okay. That's why he knows what it is. I had no idea what this was. And I ran the words hundreds and thousands through our search engine, our normal machine learning powered search engine. And this is what it came back with, Thousand Island Dressing. Now, even those of you who don't know what hundreds and thousands are, probably know that it's not Thousand Island Dressing. This is the part where it gets crazy. I then ran the same search through our new AI powered engine and it came back with this. Hundreds and thousands, probably one of the strangest words I've ever heard. Do you agree with that? Yeah, no, totally normal to you. You've been saying it your whole life. Blew my mind, hundreds and thousands refers to sprinkles. I went down the wormhole on this. Then I started Googling and looking at Wikipedia for how this name came about and I learned that sprinkles actually have many names. In the East Coast, some people call them jimmies. There you go. And there's actually quite a few other names. Now, we've never done anything in our product to teach Instacart that hundreds and thousands refers to sprinkles. And in fact, on the West Coast, if you're looking for these, they won't be called that. So they won't even show up in the catalog this way. You would get Thousand Island Dressing, but actually the AI-based system is smart enough to know that these are sprinkles. It's smart enough to organize them in a better way that actually reflects how you would shop for them. And it solves this divide, this language barrier that we might never have solved any other way. That gets me really excited about the power of AI. I'll give you a few other examples. Search is a really, really big part of how Instacart works. Now, traditionally what we would have done in search is expect people to type very basic search terms, almost computerized their asks, and then we would try to surface the best possible results for those. But actually, if you're trying to do a barbecue for a bunch of friends, well, it's probably perfectly reasonable for you to think for you to do to just search for that and then let us organize it. Tell us if there's a vegetarian person there and we'll orient the results for that. These are really simple evolutions of something that people do dozens of times every time they shop, but we can now return results that are really well organized. And the beauty of it is in the old world, we would have actually had to create every one of the carousels that you see here, but in the new world, no one created any of them. It was done automatically. And what you see might be different than what I see because it's taking into consideration not only everything it knows about the world, but also everything we know about how you like to shop. Here's another example, shopping lists are a pain. Some people use the notes app for them, some people have their own favorite app, some people just write them on a piece of paper, but they're cumbersome and they live in many places. Sometimes they live on a piece of paper, but sometimes they're your recipe book that you're looking at. With AI, we're able to just make that whole process of searching and building your cart dead simple. You take a picture, you paste a list, you speak it, you take a picture of a recipe book, it'll just load it up and put items in front of you that you can very quickly add to your cart. Really, really simple, but the kind of thing that's not possible with machine learning, it's only possible with AI. I mentioned that we organize search results much better now that we are using AI, and we truly are about to power every single search with AI. And so when you do a simple search for something like healthy snacks, you're gonna get back an organized list of results that's so intuitive that it will make it very, very easy for you to shop. This is what I mean when I say taking the mental load out of shopping online. In the future, you're gonna see interesting things like when a shopper is looking at a shelf and all the items are out of stock or some of the items you're looking for are stock, maybe they'll just take a picture of that shelf and you'll get back, a version of that picture has all the products right there for you to tap and replace in real time. Much, much easier than any other way that we would have solved this problem. But again, not something that can be solved without AI. Okay, a couple more examples for you here. So I don't know about you, but I'm pretty forgetful. And the process of buying things almost always result in me forgetting something along the way. And the recommendations that help you figure out what you've missed are just as important sometimes as the recommendations that show up during the shopping experience to tell you what it is that you've already searched for and what products represent those things. So really simple things that we can do on Instacart. In fact, one of the coolest things about Instacart is that when you've completed your order, it's not too late to add more things. You can keep adding more items to your order all the way until the point where the shopper has checked out. So this idea of forgetting, which is normally a stressful experience is no longer stressful. So I shared with you that to me, the inflection point between what we use to talk about with machine learning and neural networks and what we're now talking about with AI is really about creation. It's about content didn't exist before and content that might be different depending on who in this room is accessing it. I think that's extremely magical, but I'm also conscious that I'm surrounded by a bunch of brilliant product managers and it's really going to be up to you to figure out how to use this technology and your products in a way that is truly additive. So my recommendation is don't force it in. Think instead of the magical ways that your product could get better if you truly make every single piece of your product work in a new paradigm, in this new AI driven paradigm because I think you'll find that you'll build things that you never imagined possible before. Thank you.