Latest / Talk Commerce / Live from Ecom Forum: Sharon Gee Is Transforming Ecommerce with AI and Agentic Commerce
Transcript
- Speaker 2 (00:01.696) All right, welcome to this episode of Talk Commerce Live from Ecom Forum in Minneapolis today. I have Sharon Gee from Commerce. Sharon, go ahead and tell us what you do day to day. Tell us a passion you have in Colorado. Awesome. Okay. Well, thanks for having me. So, I've been in commerce now for six years, and in my current role, I'm the Senior Vice President of the Product Organization focused on our AI offerings across the product portfolio that includes BigCommerce, e-commerce platform, Feedonomics the enterprise data feed management solution, and then Makeswift, which is the market leading react, next year's front end component, visual editor. And in the role there, it's really about how do we bring the data together that can help merchants become more discoverable and make it easier for their shoppers to find them wherever they are looking for products, whether that's on ads or social or marketplaces or agentic channels, and then make it as easy as possible for them to consider those products and make a decision to purchase, whether that's on a third-party channel or an owned dot-com experience that might be powered by BigCommerce platform with a visualization layer that might be powered by makeswift. So in that role, it's been fun. I helped lead the acquisition strategy for Fedonomics four years ago. I was the GM of that business and we've integrated that business successfully into the larger kind of portfolio. I think, know, Identic's one of the most exciting things that's happening. For many, many years, those of us who've been in commerce for, you know, the couple of decades that I have been, we've been kind of noodling and fixing the data that has been going to ads channels to try and eke out additional conversion or additional return on advertising spend. But the machine and the game, Speaker 1 (01:45.136) board game rules have been about the same. It's been, you know, acquire customers on Google or Meta and then try and push them to your dot com and hopefully you can track their behavior, understand enough about them, capture their interest, remarket to them, and increase their conversion rate and their average order value or their lifetime value. You know, we're up with a conversion rate of somewhere between two and five percent if you're in the data sea space. Somebody came along and bopped the board game and now we get to reset all the pieces because now all of the eyeballs are going to answer engines to answer our most basic questions. Those queries are no longer sort and filter criteria based on price or size. They're now long form queries about I'm going on a vacation in Italy and I want to wear a dress to my friend's wedding and it needs to be made of this color and it needs to be this size and it needs to be delivered to me by tomorrow. What can you recommend for me? And all of that data that our merchants have access to across their marketing channels, internal systems, their content teams. all of sudden they have to bring it all together and really get their house in order because data is the new storefront. If the customer is the channel, wherever they are, our job is to surface the data that allows them to make a good decision. And increasingly, these answer engines need really deep context to be able to answer long-form query. So that's really where we spend a lot of our time is figuring out how do we surface, how do we make sure that our merchants' products are discoverable anywhere their shoppers are and make it as easy as possible to shop. consumer wants to whether that's clicking through to a gorgeous you know customized personalized product detail page where I can visualize a couch in my living room or whether it's like let me buy it with a thumbprint because it's mascara that I know that I want you know we need to offer shoppers that flexibility because we are all shoppers using you know tattoo BT and and Gemini apps in order to be able to make our decisions and and so we need to move with urgency and have our merchants there on the channels where their customers are the most relevant so and that's what I We do it commerce nowadays and I think what I'm excited about you asked a question about what do I do in Colorado? I own a flower farm and a coffee shop and that's been a really fun change of pace since the days I was agency side in New York City. Speaker 2 (04:00.37) Wow, flower farm and a coffee shop. That's fantastic. Okay. So I want to unpack a little bit of what you said. We talked a little bit before about the PDP page, but it sounds like it's really about the whole site, not just about the page and how people are finding the products they're looking for. But as a merchant then, you almost need a map to help the LLM find the parts of the information you need because it is going to come. it's gonna find something and then it's gonna lead down a path just like a crawler would but maybe talk a little bit about the differences in the way SEO found people and word stuffing which is like way back when to now how LLMs are finding people and you hinted at it when you said I want to find I want to go to Italy and I want to have a nice dress and blah blah blah. Well, I think what's really interesting is so these new channels are Consumers are starting to adopt these channels faster than they've ever adopted any consumer technology ever including cell phones like ChatGPT got to opening I got to a hundred million users faster than any other technology we've seen it's this asymptotic line It's really amazing and and we all see the value, right? We've got we've got this thing that knows everything sitting in our pockets And so if we can answer if we can ask it questions that it can give us these amazing answers. We're gonna use it The challenge is to give a good answer an answer engine needs data. And they can get that data in a few different ways. One of those ways is they can scrape websites for it, but those websites aren't really optimized for agents. So it's full of HTML, it's noisy, it's unstructured, it's not, you know, the way that you present information to a human visually with HTML and images and all that, you know, we have these brains that like, don't like to be overwhelmed. But the way that you present that information to a user that is now an agent or, you know, is very different. And so we're finding this new you know, this challenge in the AEO and GEO world is how do you get good data, structured, contextual data into the veins of the answer engines, into their mind, so they remember to say your name when Sharon's looking for this product, right? And so part of this is not how do you do a real-time call to see if that product's available in stock at that price, which is what a lot of the current protocols are focused on. Rather, you know, at Fedonomics we focus on what data do you need that you have that you can serve Speaker 1 (06:23.83) to these channels to provide the most relevant context. So an example would be, you know, we talk about like levels of data. So like level one data is the data you need to like power an ad on Google, Title, description, image, size, color, know, weight, those kinds of things. That's structured data. The next level of data is like what you send to a marketplace where they're the merchant of record, right? Like to list on Amazon requires significantly more data than to advertise a product on Google. That's like level two data. Level three data, manufacturer is all that data that's sitting inside your PIM, right? So product specifications, what's it made of, where did it come from? You know, all of those products, if you're a consumer electronics company, you've got product specs for days around memory and all of these, you know, screen size. And if you're in apparel, it's hill size and fit and style guides and all of that, you know, trend information. And then you move into unstructured data, which is like anything that lives in a PDF on a website where you like, you go to a PDP and you like download a PDF to learn more, all that unstructured data can now get ripped, contextualized as data and sent to these LLMs so that they can answer questions more effectively. But that's not the kind of data you usually show on a product detail page. So all of a sudden we have a bifurcated experience that we need to provision because agents are customers and shoppers too now. And our job is to give the data to the shopper that they need to make a decision. Well sometimes those shoppers aren't humans anymore, they're agents and we need to batch up that data in nice ways for them to consume. So, you know, whether you see the Cloud Flares of the World talking about how to do that. know, LLM's dot text or robots dot text files where, hey, we detect that this is an agent that's coming to this website. Serve them this kind of website instead of this one that's built for humans. We're seeing that kind of bifurcation happen because what they're trying to do is offer the data that is the most relevant to answer the question so that they can be discoverable. Because the problem is, oh my gosh, I've spent the last 10 to 20 years optimizing my data for an SEO world for Google. where I stuff search terms into my site so that I get traffic to my site, that's not what's happening anymore. Now, it's how do you send the right context to the LLM where the decision and the discovery might not even make it to your site because it's all happening in that answer engine. And so now it's much more not only how do you provide the content but the context that is required for someone to know what is the best pair of shoes for a man with size 10 feet in blue when he has wide feet and plantar fasciitis. Speaker 1 (08:53.37) That kind of information is not how merchants are usually thinking about what they should put on their PDP, right? How do you identify which products are not good for certain things? Like we know that the shoe is not for running. We actually know it's for walking or for lifestyle or for travel. You know, how do you, that's not something you want to put on a PDP, right? Like you want to say what it's for, you don't want to say what it's not for, but that becomes just as important. Right, yeah. So when we think about what kind of data do you have on those levels, one, two, three, levels four and five are like unstructured data. we're user reviews. That is an amazing data set of real humans saying words about your products. It's a treasure trove of search terms that should be integrated into your product detail page descriptions. So the opportunity that we have is really deeply understand where our consumers are. That's always been the problem. Who loves our product? Where are they shopping? What questions are they asking? And what data, content, expertise do we have as the merchant or the retailer to offer that content to them in a way that is relevant to help them answer their question? And so, you it used to be a very ads-focused game, and now a lot of these channel entrants, they're not even monetizing with advertising yet. You know, and so there's this huge opportunity for the people who get their data ducks in order faster to get outsized channel opportunities. with visibility because if you show up with relevance in a place where all the eyeballs are compared to your competition, you stand to benefit. So I think one of the biggest opportunities we have seen in this kind of how do you get involved in the identity era, it's invest in the data, get it to the channels where your shoppers are which increasingly are identical channels and then use that data as a pipeline on your own experiences to offer an even richer experience like a brand agent that can answer questions in a semantic conversational way. Speaker 2 (10:43.47) When you're talking about putting all this new data onto a PDP page, are we talking about a whole bunch of data below the fold? Are you talking about adding, I think the reviews is a no-brainer, right? But if I have FAQs, if I have all these extra things that I want to add, and I want to add it in also a human coherent way, are you talking about a completely different experience that a human would never even look at the set? Exactly. So there's like when you sense that this agent is coming to your site and it is not a human, do you render them a different version of your website that's actually full of data and not full of pictures? It's full of links to pictures that they might want to reference, but it's not necessarily, it's not even presented in the same way. So we have to, you know, the three truths, the customer is the channel, data is the new storefront, and agents are customers too. have to think about how do we build technology? How do we build the protocols that allow the rails to pass this data back and forth in a new world where the people accessing the site don't have eyes? You know, yeah, I mean can say that last year I kept saying to people that eight like this whole idea of agent of commerce or an agent goes and it gets deployed and buy something it was like I don't even want to think about that where hey people are gonna buy things for people and people are the ones are gonna shop and then All of a sudden we have these ideas of this agent coming to your site He's gonna buy something for you. Yeah, and no and now it's not people anymore But it sounds like we're gonna go to one step further and maybe your site doesn't need to have a people-facing portion. Is it just going to be an API then, do think? Speaker 1 (12:20.832) I think it's going to be both because the brand site is still where people it's just one of the channels where people are interacting with the data Right so and it's the one that you as a merchant control right where you control what that experience is But the data becomes the data that you're surfacing to channels and how they choose to visualize it That's up to them in many cases if it's an agentic channel, right? We've all experienced that when we like look for products or are shopping for something on you know Whether it's a open AI or on a perplexity shopping experience or within you AI answers on Google, across shopping surfaces. So if you don't control the way the data is visualized, but what you do control is the data, you've got to invest in making sure the data is really good if you want to show up on those third-party channels that you don't control. It doesn't mean that you don't invest in the channels you do control. In fact, quite the opposite. It means you invest deeply. And if you're investing in data that you send to those third-party channels, the way that advertisers have forever, because it gives them the best. Any marketer knows that if you send better data to Google, they're going to charge you less for your clicks, right? Because you're giving them more relevant data to answer the ultimately the searchers query. So, we're kind of watching the data nerds of the world in here at the Earth. They started as the heads of organic, then they turned into the heads of paid, and now they're the heads of agentic. Because it's like when the creative director used to run the website, and then they got replaced by the person who could read the analytics on the website. We're seeing that same kind of thing happen here where those who understand how optimizing data for AI, whether that AI is third party, like OpenAI or Perflexity or Gemini, whether it's first party owned by a brand with a brand agent. So, you we hear this use case, I want to offer a shopper assistant on my dot com. That's somebody, you know, like a chat bot experience. That's kind of what I would call like level one, like baby version of what's going to happen. But think in the future when you have a B2B sales assistant who is completely trained on all the same PDFs that the sales people are. Speaker 1 (14:19.79) Imagine how much more time a human sales rep gets back if three quarters of the sales cycle was able to get through in the middle of the night while they were asleep and they can do all the high human in the loop, high touch important kind of capabilities that are necessary. It's going to revolutionize that. think the B2B with agentic and brand agents is going to be one of the biggest opportunities for B2B to leapfrog consumer experience compared to B2C experiences right now because they're the ones with the data. They're the manufacturers. They're the one with the blogging. they're the one with the pricing power because they've got, you know, custom price books and customer groups. So I actually think that, you know, when we think about how do we build experiences for shoppers, be they B2C or B2B, whether it's an agent or a human, these are the questions we have to ask ourselves when we think about, okay, and how does that data need to flow in order to deliver the best experience? The experience for an agent needs to look like data. The experience for a human ideally looks conversational. and expert and trained on the same data that the agent has. It's going to be agent assisted. So that's how we think about that problem at Commerce. I'm a big, I'm a big quad code user. I realized, mean, interacting with quad every day now, I've realized how stupid quad is in real life and how illogical it can be. How long do you think it's going to take for us to get, and if this is happening at the coding level, it's going to happen at the consumer level that, that, that it's going to make a lot of recurring mistakes and it's going to make a lot of assumptions that aren't correct. And one thing that drives me absolutely crazy is it gives you three questions and then it just goes and does one of them on its own. So I see at some point we're going to have this agent to come up. We're going to have searching and you're going to say, give me the shoe that I can wear in Italy for going on a hike in Tuscany. And it's going to give you, it's going to say, here's your three choices. by the way, I just already do choice number three. Like we're going to have some guardrails that are going to have to happen. Speaker 1 (16:28.748) those protocols is really where the commerce platforms and the channel partners and the payments partners are really getting together. You've seen more open protocols released in the past six months than like the previous 10 years combined. It's because we know that none of us want an internet that isn't safe, that doesn't keep us, that we can't trust. Trust is so important when it comes to like I want to buy a thing, you know, so I want to be able to authorize my agent to shop on my behalf, but I want to be able to say yes, right? And so that human in the loop piece, these are deep, like the protocol. protocols that are being, whether it's Stripe ACP, we're involved with, or PayPal, or whether it's Google, then the AP2 and some of their agentic protocols. All of these things are being established because we need to make sure that the guardrails can trust all four parties who are now in the mix. There's a shopper, there's a shopper agent, there's a merchant, and there's a merchant agent. Nathan, yep. All four of those have to be in cahoots and agree and trust each other. And that gets complicated. And so the technology companies are like, yes, let's lean in and figure this out. It's a really fun problem. And all the attorneys are like, my god, data security. Right? In a good way. But the new rules of the internet are being written in the Identic space. And I think it's a really fun space to be in right now because the opportunities are so huge. Imagine what you can do when you can rewrite your entire catalog anytime you click a button with General. of AI based on the search terms that came from the channels on which the searching was happening. Like, if you could rewrite your entire product catalog with a click of a button to be all focused on Halloween, like, what's limiting you from being able to do that now? You know, it used to be, oh, I don't have enough copywriters. I, you know, like, I don't have enough humans to be able to do those creative things that would deliver an amazing experience to my shoppers. That's not the case anymore. Those limitations are removed. And so a lot of people are talking about AI as a way to Speaker 1 (18:19.906) reduce, you know, get more profit for less cost. That's totally, you know, a fair assessment of like, let's become operationally efficient. But it's also like one of the biggest enablers where it's about your growth for the amount of humans that you have working. It's about how do you give the humans that are in-seat, you know, jet fuel in order to unlock capabilities and power and scale that has never been able to be realized because they're all in the scuttwork of the operational things that robots can do better than humans. So I I'm actually extremely optimistic around what this can mean for better experiences. I would love it if our generation is the last one to use a mouse and a keyboard. Yeah, I with that. I agree with that as well. I think one of the biggest things that we're going to see from AI for users, for merchants, is to find things that they're not doing instead of what things they should be doing. Like they shouldn't be worrying about generating content. They should be worrying about the content they're not generating and asking it to find patterns within your data to find that content that you need to be doing. lot of our customers are leveraging tools that allow, know, are tools that allow them to define simulated personas and understand what queries those simulated personas based on their actual users might be asking on these channels. And then based on that content, well what kind of content should you have based on the questions that someone like Sharon who's in Colorado, who's going on that vacation in Italy would be asking about the vacation she's gonna go on. And do you have the content to show up the right way so that she says and references your products as your competitors' products. That's where I think a lot of the marketers and the integrated, let's call them athlete generalists, who understand what their shoppers actually want and what their unique value proposition in the market actually is, I think they're going to win because AI democratized the tooling. All the platforms are working really quickly to try and make sure that we continue to have an open, trusted transactional experience because we all want to make sure that that data is presented in a good way but is also secure. But I think when it comes to like, Speaker 1 (20:22.992) What an amazing opportunity to be a brand marketer right now. What an awesome chance to be able to take advantage of an army of agents that can help support you and your goals in achieving. You no longer have to be an engineer in order to be able to get something shipped. It's like if you can think it and you can dream it and it would deliver good outcomes, you can do it. That's an amazing time to be alive. Yeah, absolutely, Sharon. It's been a great conversation and I know we've gone over our 10 minutes so I appreciate you taking the time here. We should continue this again in the future. This has been so great and I really appreciate you taking the time. Last question, what are you getting from the e-comm forum? Yeah, thanks for having me. Speaker 1 (21:03.98) Well, I think Darin and the Titans are one of the most heartfelt group of humans in commerce. And I love coming to the show. He always curates a really great group of people who are thought leaders in the space who are dealing with real problems and how to implement them. And I think you're no exception. And it's been wonderful to spend the time with the group. They gather every year. And I think that it's one of the best shows with the most personality. And ultimately, think, despite the fact that AI is everywhere, it's kind of still about the humans. It's about the humans. Thank you. Yeah, appreciate Sharon Gee, Commerce, thank you so much. Thank you. you