Name: Part 2: US Agentic Commerce Readiness Series | Accelerating Work and Productivity: Preparing To Move at Machine Speed
Uploaded: 2026-09-02T18:31:29.663Z
Duration: 34 min 40 s
Description: Part 2: US Agentic Commerce Readiness Series | Accelerating Work and Productivity: Preparing To Move at Machine Speed

Transcript for "Part 2: US Agentic Commerce Readiness Series | Accelerating Work and Productivity: Preparing To Move at Machine Speed": Okay. Hello, everyone. As we get people in the room, I wanna welcome everyone to our second out of three webinars in our agentic readiness commerce readiness series. The goal here is to give you a practical playbook to repair your products for agentic shopping. without further ado, I'm gonna hand it over to our director of product marketing, Nathan Teplow. Thank you, Rae. Hi, everyone. Excited to be here. Nice to see a lot of, familiar names and, the places you're all joining from. So I'm excited to, to kick things off. I wanted to set some context here, about the landscape and just sort of tee up our our discussion for the day. And if you joined our part one, you probably recognize this slide. But just to start things off at a super high level, like, we are clearly in the agentic era. And the question that you're likely faced with is where is your growth gonna come from over the next few years? Now there's still the physical shelf, right, in store shopping, but most projections are showing that that is gonna be minimal growth and or potentially declining growth. So most of the growth is gonna come from those digital channels. And you still have the digital shelf, right, your traditional ecommerce channels, and you should be working to squeeze out every last bit of conversion on those those channels. Right? But we have this new shelf, this agentic shelf, and it's fundamentally changing the way that products are being discovered and purchased. The challenge here is that we're not necessarily getting more consumers. People aren't buying more because of the agentic shelf. It's only changing the mix, and it's changing the way that products are being discovered and recommended across this journey. So growth in this next era really means winning more at bats. As consumers traverse every stage and every level of this of this funnel and these different shelves, you need to be winning more at bats. And this means you have to be relentless at optimizing your product content and refining your strategies to ensure that your products are being recommended and purchased across all three of these different shelves. The challenge here is that you're faced with a math problem. There's a lot of work that's involved here and it's putting pressure on teams. And just to go through some simple math here, let's say you have a thousand SKUs. Right? Best practice is to update those about once per quarter, so four times per year. Let's say you sell across four different channels. I know many of you probably sell across more of those. But, sorry. Just making sure people can hear me. It looks like they can. I just saw one note on that. But across all all different channels, 16,000 updates per year. That equates to about twenty five hundred hours of just purely updating your content across the digital shelf. And just for reference, there's about 2,000 working hours in a given year. So this is like one and a quarter full time employees dedicated to just purely keeping your PDPs up to date and optimized in a best in class process. If we look at the agentic shelf and add that on, let's say you have the same a thousand SKUs. LLMs are actually prioritizing content that's fresh and up to date. The the the the more that you're updating your content, the more likely it is to be recommended by, AI agents. So let's say you move these a thousand products to now being updated monthly or 12 times per year. That's now 48,000 different updates you have to make across these four channels. That's about four full time employees that have to just, again, purely manage the basics of updating product content across across every one of your your different channels. And the challenge is, like, we're not getting more people. Like, teams aren't growing, headcount's not growing. So, like, how do you absorb this, all this new work that has to get done? In addition, there's new content types that you have to manage, content that's that's purposefully built for LLM consumption, things like structured FAQs or use case guides. LLMs also prefer localized content. So how do you ensure there's, the right localized versions of content for all your different products? Like, there's just more data you have to manage on top of more of the overhead of just maintaining your different product pages across every shelf. So this creates a math problem. Like, how do you keep pace with all this new work that has to happen when your teams aren't necessarily growing? And on the last webinar, we talked about what PXM must become for this new era of agent of commerce. And it has to enable you to manage that entire library of product data and product context. It has to enable you to work at machine speed, and it has to be able to power experiences across every one of these three different shelves. And, you know, today's webinar, we're gonna really focus on how do we work at machine speed. And in the age of agent to commerce, really, the only way to keep pace is to use AI to your advantage. And your platform must enable you to orchestrate this work across both the humans and those AI powered agents to operate at the speed and scale that's required for this new era. And it can't just be one off actions. Like, you can't just do ad hoc prompts to different platforms. The way you implement this has to be programmatic, it's gotta be repeatable, and it has to be structured so that you can ensure the right consistent outputs every time you're asking AI to do something for you. You ensure the proper human review is taking place throughout that process, and it helps you drive real measurable results to really prove the value of the AI investments and the, you know, the transformations that you're making across your organization. So we wanted to do a quick poll just to get a sense of where the audience is, when it comes to adopting AI. So we've got a few different statements here. I think, Rae is going to launch this poll, and we wanna ask you to to to choose the one that best describes your current use of AI within your, your existing organization. Some of the options we have, once you launch the poll, you are not using AI potentially at all. You might be using AI for ad hoc tasks, but not necessarily a repeated process, like, it's not a proven out process. Perhaps you're experimenting with AI, but, again, it isn't a fully established process yet. Maybe you are using it for one or two different business processes, like, let's say, like, running a translation, workflow or maybe creating, product imagery, or maybe you have it implemented for multiple, processes. So we'll give it, just a few more seconds to let everyone, complete the poll, and then we'll, we'll launch the results here. And I'm seeing the polls come in. So, hopefully, you all can see the results. It looks like most are in the bucket of ad hoc usage and oh, here we go. Perfect. Ad hoc usage and experimentation. And that's pretty consistent with what we're seeing across the board. Like, everyone is, you know, interested and eager to use AI, but it's not quite proven now. Like, we're still sort of experimenting. And in this webinar, we really wanna help provide you with best practices and recommendations for how to move beyond that experimentation phase and really put AI into action. So just to bring it back at a high level to Salsify and as we shared on our last webinar, Salsify provides a unified platform that powers it helps power this process acts at scale. We have an intelligent engine that enables you to work at machine speed, and we serve as that system of record, the system of action, and that system of delivery in one complete platform. And this is what creates that trusted foundation of product data and intelligent automation that you need to ensure your products win in this next era. And on the last webinar on part one, we really focus on the system of record and and what you need to do to establish a trusted foundation of product data. And for this webinar, we wanna focus on the system of action and how you can use the platform to automate processes, to apply new AI powered processes, and accelerate work across your team and increase your team's productivity. So we have a fantastic customer who's joined us for today. She's been able to take her foundation of product data and product workflows and apply new AI powered processes on top of them to achieve some really incredible results. So I'm excited to introduce Jessica Thomas. She is the director of PXM and data solutions at the The Legacy Companies, and she's gonna tell you more about her journey and how she's transformed her organization and their ways of working for this next genre. So, Jessica, take it away. Well, hey. Thank you, Nathan. Super, super stoked to be here. And kinda like what you said, I wanted to just kinda give some real world, I guess, approaches of what our company did to really set the stage for getting us ready to, onboard AI within the platform of Salsify. So I'm gonna just really get it down into the weeds here, of where we started and what our key stepping stones look like. Right off the gate, just at the beginning, we, did a clean audit, not a clean audit, but we audited our dataset. We needed to look at what we had available to us, what our categories were, what our assortment look like, establish what our core data was versus, like, enhanced data. So we knew every product, because we also work with our ERPs and, our tech stack across the business. We knew every product needed SKUs, UPCs, what the classification was, where the exclusivity was, very, very specific to the product across our tech technology. Then our enhanced data. What did it look like from the retailer side? What our customers needed to see? And I do wanna be very clear, like, we all know content well, maybe not. But in my world, I look at content as data. At the end of the day, it sits in our system, as data that our customers would also need. So that included the enhanced dataset that we established that audit. Next, we, we standardized and cleaned what gaps were defined from that audit. Meaning, like, getting down into what kind of data value is being entered into the system. Was it a text? Was it a number? Was there inch signs? Was there you know, we got very, very specific and granular, because consistency is key across the board. Regardless if you're gonna use the Salsify dataset to then publish out into the retail, space. The less you have to manipulate that data, the better. So less touch points if you can standardize and clean. We then went, to our teams. We had I I joke with my my company here that I almost annoyingly so interjected myself into all of these depart departmental conversations, not because I was necessarily nosy, but I also wanted to just understand the business needs, what, customers we serviced, what teams did what, who owned what. And so we got everybody kind of on the same page, to communicate. That was our product managers, engineers, IT, and so forth. Then we defined ownership. Who owns these datasets? Is it if it's certifications and compliance, who's actually providing that data to us or putting it into the system? If it is product specifications down to sizes, color, again, content, who actually is owning these different datasets. So we got alignment from our teams there, but then we also empowered it, empowered these teams to own those datasets. We moved from one single data entry person, to 24 strategic data owners. We were able to use the tool of Salsify to really hone in on the the proper ownership and basically empower these teams to put that data directly into the system as they needed to do it. It took away the the need to email a million people to get the data entered or put into the right spot. So we just empowered our teams to do that and execute appropriately. So because we defined all of this, out the gate, we were also able to do process improvements. We moved from a reactive state to proactive. So I kinda mentioned this, but I went to basically all of our top retailers, our customers, and what they needed for product setup. So if you think of new product development or new SKUs that we're trying to launch, instead of waiting to launch that product to then go and hunt the data because we knew what that enhanced dataset looked like, we caught it right out the gate for anything that was launching. Originally, we were in a I think it was, like, hundreds of columns deep, Excel sheets. And then we were able to move that into the workflow optimization using the tool of Salsify to actually, assign certain tasks and use, like, our product life cycle to then say, this is now the data that we need. We are planning to launch this to Walmart or Home Depot or Amazon or whoever. Let's go hunt that data down. But we moved it right in the system and kinda let the system do project management for us versus Excel and, you know, figuring out who does what. And then we had the go to market data ready to go because, again, we moved out of this proactive or reactive state into proactive if it was a commercial appliance versus a countertop appliance. We knew what we needed at at the start of that product launch. And we were able to also, optimize any products that we had not yet used this workflow, you know, optimization for. So because we had set that foundation, we were then able to take advantage or not take advantage, but really jump on the intelligence suite train, earlier this year. So I wanted to just kinda walk through what that pilot looked like for us, and how we got it started. So knowing that it was coming and that we had kinda set the stage for having the data in our in our platform that needed to be there, we decided to pilot the intelligence suite in January of this year, starting with copywriting as our main use case. So knowing that that was coming, the leading up to it about one to two weeks prior to that pilot, I worked directly with our creative and marketing teams to kind of give us a brand and a category that we wanted to kinda test with and use our brand guide, to establish, like, tone of voice and what that setup would look like. Week one. So pilot begins, and now we're getting to work. So we tested this out a few times, built some prompts. We I know every company is very different here, but for us, we don't really write retailer specific copy, but because our copy suite is what I like to call it, is so robust. So right out the gate, we're asking for three different titles, three different descriptions, all with different character count limitations. We have minimum of 10 bullet points. Amazon, we do have I have to correct myself. We do have Amazon specific copy because, you know, it's Amazon. And so we have this really robust copy set, but we were able to use the prompts accordingly and the brand guides to then kinda set that up. We use, like, maybe one or two skews to test, go back and forth with our marketing teams to just make sure that this is on brand. We have that the tone of voice. We're talking to the right people. And then once that got approved, weeks two and three was refining that, building it into our workflows of new product development. We know that the product is launching on this date, so we know that we have to have copy at this time. And so we kinda built that into our processes and project management workflows. And then once we figured that out, weeks three and four of the pilot was just rinse and repeat. We just hit the ground running, brought in more brands. We were able to really go at full speed at that point. And what that looked like from process and implementation, because of these workflows being built into our processes, we were able to use it both ad hoc and automate. So we set up kinda rules in the system to say, again, if a product is launching at this retailer at this time, if it's a refrigerator versus a, ice machine versus a toaster, it gets this kinda copy. It has this voice for specific brands, and we run that with the marketing and creative teams. We then moved kinda from a copywriting perspective to a copy editing. So with with us, we're very we had to build, like, a add, a hybrid version of this because we definitely didn't wanna lose the human element of this. But we also have many people that work in different departments that want to have some hands on approach still. It definitely and we'll get into the numbers here, but I'm very excited to show you the time saved even with those review processes still in place. But just to kinda show you, it is a hands on, you know, approach. Every department, has a little bit of a hand on this. And then the turnaround because, again, we set that foundation was so quick. We were able to rinse and repeat rinse and repeat. And let me I oh, I do wanna give a shout out though because we did all of this data gathering and optimization. We were able to use Salsify IQ, to bring in that data without having to use an outsource, LLM. And, we had all of that data readily available. So not only was it accurate, copy that we were getting, it was still on brand. And so I was that was a huge shout out to Salsify for the IQ there. Alright. So let's look at those results. We've had this, like I said, since January. We started with one brand, and we are now up to 10 brands that are live through Intelligence Suite with, with copywriting as our main use case. Alright. It's a little small to see, so I apologize. But, we have basically doubled the amount of production time as well, the amount of SKUs that were run through the intelligence suite. The Legacy Companies path, we only had about 683, SKUs that were run through for copywriting with a copywriter. And it took, on average about twenty six days to complete. With the use of intelligence suite, we were able to do about 1,600 tasks have run through that with six days turnaround time for completion. So that's from the start of the copy being written even with the review process, and then it being published into Salsify. Excuse me. So we had about 833 SKUs through the intelligence suite. So those were tasks run because of the hybrid, situation. But, let me move on to my next slide. I'm sorry, guys. There we go. 833 SKUs went through the intelligence suite. 300 and, six SKUs through the legacy path. Overall, we have about 8,000 products that have gone through the the full workflow. But with the brands, you can see that we've have quite a lift across certain brands, based off of the time that they were onboarded. We have the Amazon title creation. We were able to kind of do a hybrid solution on that, using the copy suite or the copy, intelligence suite, with all of the new Amazon titles that have launched as well. So we're just the production time has gone from I think it was, we've about 74 increase of production, that we're getting work done in our in our space here. So here you go. Yep. 74.8% increase in production time. And, again, we've only had that since January. You can see there that we've got a seventy seventy percent increase in time it's taken to, to produce the the outputs. So what used to take fifty two days, this is overall of everything that's gone through this week. So it's just the numbers don't lie. It's been wonderful to see, how much work has gotten done staying, you know, speed to market to get the products across the finish line. So, yeah, it really speaks for itself at the end of the day. So all in all, I keep, forgetting to move my slides here. So all in all, the way that I see it is organizations that succeed with AI won't be the ones that adopt it first, but they'll be the ones that prep for it the best. And for us, in my experience, we are definitely seeing the the success, in in our space here. So getting that foundation set, knowing what datasets you're working with, and just setting your foundation, up for the success of the tool. Jessica, I might invite you on stage, and, we can talk through some different, questions. We've got a number from the audience here. And, maybe Steven, you can take a minute, and we'll, we can circle back and and finish things off. So thank you for some of the questions that have come in. Please add more questions, as we start talking through these. I, so one of the first ones here is on we've got a few questions actually on SEO. Should we be generating SEO descriptions, and should that be a focus in this agentic age, or is it less of a focus? I've got a few initial thoughts. Jessica, you might be able to chime in as well. The I think the thing with SEO and AEO, you might hear the term AEO, answer engine optimization, is like SEO doesn't go out the window. You don't want to give up your SEO playbook. A lot of the building blocks of AEO is off of SEO. You shouldn't just forget about SEO, but some of the ways that agents are looking for data is different. And the way you might structure some of your answers and structure some of your content is a little bit different for for AEO. One methodology I've seen is, like it's called bottom line up front. Like, the way that that answer engines look for stuff is they just want the answer. Like, if you have a question, is this backpack good for, you know, wet weather hiking? Like, yes. This backpack is good for wet weather hiking, and then you can go and answer the rest of the context and details there. So structuring some of your content a little bit differently to give those answer engines what you need what they need is a way to help influence them better. But that doesn't mean you just throw out everything you're doing from an SEO standpoint and you can build upon it. Jessica, I don't know if you have other thoughts or ideas around how you're approaching it, but I would love to hear what you guys are doing from an Well, yeah, I mean, I agree with you. SEO is very important because. it still triggers those keyword searches for the the average user in the web space and even how products are classified on these retailer sites. AEO comes into play though when you're thinking about how you as your as your day to day shopper, when you're going into the chat GBT, you're asking the agent questions about what works for me with whatever product you're searching for. So your copy needs to speak to those use cases. So AEO comes into a big play in this new era because now not only are you, like, popping full keywords, you know, fluffing it up, you're actually speaking to the day to day user of that product and what they may be looking for to use that product. So AEO is very important, and it is how we have started to structure our content now is we are moving into we don't wanna lose the keywords, obviously, but we wanna speak to the day to day user that would be purchasing our product, both from a brand perspective, but also that use case, making sure we're answering those questions within our content. Awesome. Thank you, Jessica. Another, I guess, easier question quicker question to answer, on the Salsify Accelerators and Intelligence Suite are related to SEO. There is one for SEO. There is one for AEO. Just to clarify with Intelligence Suite, we have a number of different accelerators that can do a whole lot of things. There are ones for content creation, whether that's SEO content creation, whether that's optimizing for a certain type of persona or a certain buying season, let's say. There's also, like, data extraction. If you wanna take a PDF and have it extract the copy off of that, you know, structured PNF and turn it into product content. So there's there's many different accelerators with the within the intelligence suite, and it's about applying it in different ways based off of how you wanna use AI, but it can absolutely be used for SEO, content creation. We also have new accelerators available for AEO content creation. So it will look at your existing, product data and recommend FAQ content that you can store in the PIM, like that structured content that, agents are typically looking for and it's easy for them to index. It will autogenerate that content for you. It will autogenerate use case guides. So the kind of content that's specifically built for agents, we can autogenerate through Intelligence Suite as well. So just wanted to clarify some of the different use cases we support and the ability to support those, SEO, AEO, capabilities. Here's one. Since we sell furniture, product geometry needs to be very precise. I've tested AI generated images before. Sometimes the shape and product details change. How would you ensure product accuracy with variant image generation? Jessica, I don't know if you've done anything with image creation yet. I know, like, these number one, these models are evolving very quickly, and, like, Yeah. the content creation stuff is a little more, call it, sure footed and, you know, easier to control some of the outputs. Images can be tricky. Like, there's just a lot more variables there. Different models are better at at image creation and different types of scenarios. So, I can't say this is like a, you know, plug in and just, you know, super easy way to use AI. That's one of the reasons why it's good to have this review process sort of baked throughout. But I don't know, Jessica. I know this is something I I believe this is something you guys have experimented with. I don't know if you have any guidance on how to hone some of the AI outputs. Yeah. I mean, so we've tested the we're in the testing phases for the image generation. I think we're getting to a really good spot. One thing I loved about that workflow is it does allow you to use your own images. Like, for us, we have dimensional drawings. And so we are able to actually use those dimensional drawings to help kinda scale the life cycle or the lifestyle images that are being generated. But, also, in the in the with the question, it sounds like some of these details can change or, like, for us, some of the specs on the product itself change. And so we've automated our workflows that that specific workflow to then be triggered upon a dataset change. So if the width changed, if the height changed, or we have a new drawing or something that is specific to that spec of a product, it will automatically trigger that workflow to then generate updated content. And then we also have the automation to then publish that updated content to the respective users or retailers or customers. So there's many, many ways to kinda just automate that so you don't you're not, you know, missing anything or something doesn't or falls through the cracks. But then use the content you already have in your system to help generate that, those images as well. Awesome. Thank you. Another question here on content on visual content creation. Is it a built in feature, and does it help with AEO discoverability as well? Again, I'll I'll take a quick pass out and answer then, Jessica, if you have any additional thoughts. So content the visual content creation through Intelligence Suite is an add on just to be clear clear within Salsify. So once you have Intelligence Suite, you get access to all the different accelerators, and use cases. It's included within Intelligence Suite. I I don't know if it helps with, AEO discoverability. If you ensure that the images have alt text on them, that helps the crawlers and agents kind of understand the image better alongside the alongside the actual, like, graphics of the image. So that would be one thing I'd recommend is making sure you have alt text alongside all of your images to help those agents understand the imagery. But I don't know if there's any data that proves, like, having AI generated imagery or whatever helps drive AEO discoverability. I don't know if you have thoughts on that, Jessica. Not not really, actually. I think he said everything that that I would. have said, to be honest. Okay. Alright. Another question here from Jen. So for marketing data, let's say bullet points, marketing data, let's say bullet points, we'll look at how your other product bullet points are configured and suggest bullet points in the same format. So, yes, you can again, it's all sort of in the power of the prompts. So you can have it look at, previous prompts. You can feed it additional context and say, I want my bullet points structured like this, or I want it based off of Amazon best practices or whatever. And it informs the the the content generation, you know, action on how to produce that. Jessica, I don't know if you have, experience, like, messing around with different prompts and formats and context, but you might have some additional insights here. Yeah. I mean, so we have, like I said, we have 10 brands that have that we're using, intelligence suite for content generation. And each brand works differently, not only with the brand guides, the tones, and everything, but some of them do require, that categorization of bullets. So we know that bullet one is gonna be, the size, bullet two is gonna be the construction, prompt. Some of our some of our brands' prompts are, like, a 100 rows deep. We have one that's about 300 rows deep. So we actually built that directly into the prompt. And the great thing about the the, I guess, the workflow is you can actually point to additional attributes that you have in your system, and it will crawl your existing attributes to then help feed it and guide it to how you want this new set to look like. So there's many, many ways to customize it. It will be very much a case by case basis just based off of your own needs. But it the the accessibility or the capability is there. Awesome. Another question here for you, Jessica. Different customers, I think, also meaning retailers, have different names for the same attributes. When you were creating your data model, did you implement any strategies for managing those sort of different, cross references? So for us, I think I said kinda out the gate that we didn't get retailer specific with our content if it assuming this is content specific or if it's data. Either way, because we standardized let's take data, for example. We standardized the dataset, making sure it wasn't a free form, and we got to the that it has to be this kind of value input right out the gate. We were then able to manipulate the data at the channel mapping level because our data was clean on the back end. We can then use a formula or whatever to manipulate the data at the retailer level the way that that retailer needed it. From a content or copy perspective, we kinda use that same case. We have this huge copy suite, multiple different titles, multiple different descriptions, character counts, and so forth for that very same reason. So when you get down into the retail level, we can then build a formula or use some sort of way to manipulate the existing data that we have based on the customer's needs. So because our data was playing on the back end and standardized, we can then manipulate the output of that to the customer's needs. I hope I answered. the question. I I think so. Thank you for that. Yeah. Another question here that I will take, it's about, I think, Salsify capabilities. But this person, they need to capture technical product information during, product development. They don't have that in a central system. They're curious if Salsify can offer those capabilities through a workflow and help them collect and structure the product information. Couple ways to do this. If it is like I mentioned earlier, one of the use cases for Intelligence Suite is you can actually extract data from documentation. So if you have a a PDF file of those technical specifications, we can create a workflow that basically extracts that from the PDF and turns it into structured product data in the platform. That's one option for that. There are integrations as well. You can set up feeds from one system to another, and you can add those into Salsify. You can then set up other workflows that help you, make sure that content's reviewed and stored in the PIM properly. We also have, a a supplier portal as well. So if you wanted to have I don't know if those some of those manufacturers might be external or suppliers or or data source from other sorts of, vendors, but you can you can have them enter that information in in a structured way that then feeds into your, into your PIN. So there's a few different ways you can tackle that. But, again, it's one of the exciting things with AI is that before, it was very difficult to take, like, a PDF, and now you can do that pretty easily. Like, it's very good at at scanning that content and turn it into the the data you wanna use and manipulate within the PIN. I think we've got to most of the questions, and we are coming up to the end of the time. So I think with that, I, you know, I just wanted to say thank you to everyone for joining. Thank you, Jessica, for, taking the time to share your story with us and share some of the results. I think that the numbers you shared are so powerful, and it's really inspiring to see that the work you've done. And, I guess just to close this out too, we still have part three coming. Rae, if you're there, you might wanna keep me honest on the date. But but keep an eye out for part three, and, we'll talk more about how to actually go and publish this data out to market and influence those, agentic shopping channels there. Thank you guys so much.