Building on the talk of AI in recent years, Agentic AI is an exciting area for retailers. But beyond talk, finding real-world examples and best practices can be tricky.
Recorded live at eTail Boston, and I am excited to talk today about a company that is innovating in this space, not a traditional retailer, but an innovative aggregator using AI to standardize product data across multiple retailers.
To talk about Furniture.com’s approach to agentic AI and more, I’m joined by Dan Russotto, General Manager at Furniture.com.
About Dan Russotto
Over 30 years of extensive startup and Fortune 1000 experience related to running tech companies throughout the entire lifecycle. Focused on completing high quality projects, products, and business objectives. Detailed understanding of the Furniture, Real Estate, Hospitality, Software, High Tech, Consumer Business, eCommerce, and Telecommunications industries.
He is currently the General Manager for Furniture.com, a privately backed business that’s using AI, data and technology to evolve the way furniture shoppers discover new brands and products. Prior to Furniture.com, Dan was the VP Product for Apartments.com and the Homes.com (CoStar brands), aggregator marketplaces in the real estate industry. Before CoStar, Dan spent several years across multiple tech startups.
Dan Russotto on LinkedIn: https://www.linkedin.com/in/danrussotto/
Resources
Furniture.com: https://www.furniture.com
The Agile Brand podcast is brought to you by TEKsystems. Learn more here: https://www.teksystems.com/versionnextnow
Don’t Miss MAICON 2025, October 14-16 in Cleveland – the event bringing together the brights minds and leading voices in AI. Use Code AGILE150 for $150 off registration. Go here to register: https://bit.ly/agile150
Connect with Greg on LinkedIn: https://www.linkedin.com/in/gregkihlstrom
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more: https://www.theagilebrand.show
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology: https://www.agilebrandguide.com
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. https://www.missinglink.company
Transcript
Greg Kihlstrom (00:00)
Building on the talk of AI in these recent years, agentic AI is an exciting area for retailers. It’s certainly something that’s coming up a lot in conversations. But beyond talk, finding real world examples and best practices of agentic AI can be tricky. Today we’re here at ETAL Boston, and I’m excited to talk about a company that is innovating in this space. Not a traditional retailer per se, but an innovative aggregator using AI to standardize product data across multiple retailers.
To talk about furniture.com’s approach to agentic AI and more, I’m joined by Dan Russotto General Manager at furniture.com. Dan, welcome to the
Dan Russotto (00:38)
Thank you so much, Greg. I appreciate that introduction.
Greg Kihlstrom (00:41)
Yeah, yeah, looking forward to this. And yeah, it was fun doing the fireside chat yesterday. we’re going to do a little bit of a recap of our conversation here for the show. So glad to share that. Before we dive in, for those that didn’t catch our session yesterday, why don’t you tell us about your role as well as what you’re building and what makes the business model of furniture.com unique?
Dan Russotto (00:46)
Absolutely, yes.
Yeah, great. Thank you. As mentioned, as you mentioned, I’m the general manager of furniture.com. I’ve been spending actually the last 15 plus years in marketplaces like furniture.com. I was formerly at homes.com and apartments.com. I also spent some time at Cvent in the travel space. And so at furniture.com, we really trying to create that one place where shoppers can find everything that they’re looking for that they love.
Brands that they know, give them the confidence, almost a confidence engine of furniture to know that everything’s going to be there in one place. And so we set out three years ago to bring all the brands across the US in one place, ⁓ as opposed to vendors that you don’t know or don’t have the confidence to know that they’re reputable and good quality. Yeah.
Greg Kihlstrom (01:54)
So let’s talk about how and why you’ve used AI as a foundation for what you’re building here. And what was the original problem or the pain point that caused you to rethink how retail operations should be run?
Dan Russotto (02:08)
Sure, right now we’re pushing about 65, we’re right at about 65 retailers on the site. When you think about it in terms of 200 geographies, 200 categories, that’s 40,000 distinct pages across two million SKUs, so there’s close to actually 80 billion combinations of ways that we can show product on our site. so the old ways of traditional ways of manually merchandising product just don’t work. You need some sort of scalable way to make it look and feel curated. And so that’s why that’s the use case for us and why we need different technologies, including AI, to help us really create an experience with a shopper that is personalized and unique.
Greg Kihlstrom (02:55)
Yeah. And so it’s not just built around AI. You’ve said your team is built around agentic tech. What do you mean by that? And how is that different from either some of the traditional automation that might be used or, if I can say traditional AI yet, some of the more traditional AI tools?
Dan Russotto (03:12)
Sure, sure. So as mentioned with my marketplace experience, I’m used to building point integrations from point A to point B. And it means, so like in the furniture world, for example, when we first got started three years ago, before all of the new AI technology took off, you had to one by one build feeds and integrations to each retailer to furniture.com. And so you had to map everything, you know, very strictly, very structured.
And if anything changed, you know, the feed or integration would break. What we mean by agentic AI in this scenario is instead of mapping it precisely, you can tell the agent what your intent is and what the context is. instead of saying this field that I need, you know, I need reviews for this product, instead of saying exactly where it is on the page or within the context of a feed, you basically tell the agent, go, go find me reviews on this product. And then if things change and if the website’s dynamic or wherever the reviews are captured change, the agent can figure it out and understand the intent of what it is you’re trying to do.
Greg Kihlstrom (04:21)
I mean, as someone who has integrated or been part of integrations with a lot of things over the years, that sounds amazing, first of all. So can you share, how does this work in the real world? Can you share a real example of how an agent works inside the org?
Dan Russotto (04:36)
Sure,
sure. there’s two really, we’re doing a Gentic AI in lots of different areas, but the two examples I wanted to share with you today first has to do with product enrichment. And so that the first example is we get basic product feeds from each of our retailers, but that that’s all it is. It’s basic. get name titles and pictures, maybe some dimension information. What we’re doing is we’re using an agent to not only go back to the websites of our partners with their permission to validate that basic information, but also pull additional information. I gave you the example of reviews, but care instructions, videos, any additional rich data that websites on these furniture websites have, we’re able to then bring over to furniture.com and make the shopper experience way better. And then the second example is, which is something that, when we first started, we were more of an aggregator where we would create a great way for shoppers to compare furniture products and then get sent over to our partners. Now, with the way that agentic AI is taken off, we’re actually able to take the transaction as well. And so what we’re doing with agents in this regard is imagine you had a cart that you wanted to pick three or four different products across three or four different retailers, maybe a sofa from retailer X and a rug from retailer Y. We will take the information on furniture.com and then behind the scenes we’ll actually use agents to go make those purchases on the shopper and furniture.com’s behalf.
Greg Kihlstrom (06:12)
That’s nice. Yeah, that’s, I mean, not only does that save time, but yeah, that’s, would be very difficult without a Gentic, right? If not impossible.
Dan Russotto (06:20)
Yeah, we looked into it. We looked into various e-commerce platforms, EDI exchange, and you know, with 60-something partners times maybe six months per partner to really make the integration work. It’s just not scalable.
Greg Kihlstrom (06:35)
Well, and also with the ability to, you know, with their permission to get additional information, that makes you more competitive than others that they may share that information with. Absolutely. you’re actually getting full information, care, whatever that.
Dan Russotto (06:50)
Yeah, absolutely.
Greg Kihlstrom (06:51)
So we talk about AI quite a bit on the show, of course. And so I’m sure most listening have heard the term large language model or LLM. You’ve said that you’re optimized for LLMs in your architecture. What does that mean and why does that matter now more than ever?
Dan Russotto (07:07)
So the best example for those listening as a comparison point, it’s very similar to what happened with Google 20, 30 years ago where you had to set up your page for, you know, to optimize for SEO so that Google could crawl. Well, some of those rules still apply. know, speed is very important. You want all the LLMs to be able to read your site very quickly so they can get information quick. However, there’s an additional layer now with the open AIs and perplexities of the world where
They have specific things that they’ve recommended to two websites to create in order to optimize the experience and make sure you have a chance of having your content end up in that LLM. The example with OpenAI is that something called MCP server, model context protocol server. You put your information, we put our information, furniture.com within the context of this server. And then it makes it easier for OpenAI to get that content and show it to the shopper.
Greg Kihlstrom (08:07)
Yeah, and definitely, mean, we were talking about GEO the other day and all that stuff. Definitely new frontiers that a lot of companies are… Whole new industry. Right, right. I know. Yeah, they’ll be GEO marketing firms or whatever. So, you know, as you mentioned, furniture.com is operating at a massive scale. I you two million SKUs, all the different combinations and permutations.
Dan Russotto (08:17)
free.Right, exactly.
Greg Kihlstrom (08:36)
What would you say to a retailer that maybe has 2,000 products, not 2 million? How can a model like this apply to them?
Dan Russotto (08:43)
Sure, building on what we just spoke about, definitely you should optimize, even whether it’s 2,000 or 2 million products, you should optimize your content for the LLMs. So that’s one big area. So you wanna make sure if you’re a retailer X that anytime somebody has a question about a comfy leather sofa in my world, that you’re gonna show up within the context of that LLM. So that’s definitely one area of focus. And then, you know.
In the same way we work with retailers at scale, most of our retailers work with vendors at scale. And so you hear all the time that they have trouble getting product feeds from their vendors and some of their technology is archaic and it’s hard to get all the information that the vendors actually have on products. As a retailer, you could be doing the same thing as us. You could get that initial product data from the vendors, but then you can also augment it by using agents to go.
get additional information on those products either on the vendors websites or other places on the internet.
Greg Kihlstrom (09:46)
Nice. And so, you know, to talk about agentic, I mean, again, there’s a lot of talk about it. There’s less real world case study. you know, it’s great to hear that furniture dot com is is using agentic AI as you know, as an example beyond the technology part and the data part, which, you know, critical critical components. But what about the mindset shift that’s required? Because it is a different way of thinking beyond traditional ways of doing things. Where should an organization that’s considering the start and what mindset shifts were required for furniture.com to really embrace it?
Dan Russotto (10:24)
Yeah, I think this is a really good question. think when we were sort of first evaluating all of this, we were somewhat in paralysis. Oftentimes, companies don’t know where to start. There’s just new versions of all of these AI tools coming out every week. But what we did was we said, you know what? Let’s take the latest and greatest version of OpenAI at the time. Let’s give everybody and the whole company training. So we actually brought OpenAI a couple people from OpenAI into Furniture.com. We did a half day session. After that session, we did some hackathons, which is a way to use the tools and build things, kind of take a day or two off from your normal day job and try to be creative and use things. That was step one. And then more recently, we’ve gotten more more technical over time. So we’ve gone from just regular chat GPT training to more how to use AI when you’re coding. We’re using a tool in this regard called cursor.ai or cursor.ai and we had everybody in the company, developers and non-developers get trained up on cursor.ai and then we did another hackathon and some of the ideas that came out of it were really amazing. And I think just giving access to the tools and training people up and then doing actual events where they’re somewhat forced to use them to see what they can do has been really successful for us and now it’s just kind of part of our fabric of how we work.
Greg Kihlstrom (11:57)
Nice, nice. Yeah, definitely. think getting your hands on the tools, I think is really is a critical part. So zooming out a little bit, where do you see this all going? mean, obviously there’s been tons of change over the last couple of years, but where do you see all this going in the next five to 10 years from the consumer level? What does a AI native retail org look like from the consumer’s perspective?
Dan Russotto (12:23)
Sure, so there’s a lot of opinions on this particular question and nobody really has the perfect crystal ball, but based on everything I’m seeing and reading, there’s one school of thought that says, agents are gonna do everything on your behalf and you’re not gonna have to do anything. And while I do agree that agents could do everything on your behalf, I still think that there’s that sense of ⁓ discovery that shoppers and users want to participate in. It’s part of the fun of of buying a piece of furniture or a car or a house or what have you. And so from this regard, while I do believe websites will speak to other websites via agents, what I believe will happen is that shoppers will create their own user experience through the LLM. So for example, instead of chatting back and forth with ChatGBT, you might tell ChatGBT, create a user interface that looks like apple.com and has a cart like Best Buy and then fill it up with furniture from furniture.com and that’s how I want to shop. And so I believe that there’s going to be some explosion of shoppers telling these LLMs how they want to shop because up until now, you sort of are beholden to the physical store you walk into or the website that you’re using and this, I think maybe where it’s all headed is the shopper will tell and describe how they want to shop. then as long as we’re all, we meaning, retailers, marketplaces, anybody that has content is optimized for these LLMs with our data, then that might create the experience that the shopper wants.
Greg Kihlstrom (13:57)
Yeah, that sounds good. mean, having navigated many an e-commerce platform where the UX was less than optimal, like would be nice to have some agency over that as well. And yeah, don’t think that, I mean, maybe at some point we’ll have AI furnished houses or whatever, but I agree with you. Like, I think people are gonna want some agency over what they buy and as well as how they buy it. That sounds amazing.
Dan Russotto (14:22)
Right, right,
Greg Kihlstrom (14:23)
So we’re here at UTel Boston, and obviously we had our session yesterday. What’s been a highlight for you so far, or maybe something you’re looking forward to?
Dan Russotto (14:34)
For me, the highlight so far has been connecting with other people that are going through the exact same issues and problems that we’re going through. know, it feels when you’re operating on a daily basis within your own company, it feels like you’re the only one that’s faced with the challenges. And I actually, from this show, I feel like we’re not alone. And I actually feel pretty good about where we are on the on the spectrum of the knowledge we’re gaining, et cetera. It’s still scary times and who knows what will happen in the next couple of years, but I feel it feels good that we’re sort of all kind of going through this technology revolution together.
Greg Kihlstrom (15:11)
Well Dan, thanks so much for joining and for sharing your insights. One last question for you before we wrap up. What do you do to stay agile in your role and how do you find a way to do it consistently?
Dan Russotto (15:22)
That’s a good question. I’m a little bit, the average age at my company is about half of my age. And so I’m constantly faced with what you just asked. And I read a lot. I speak a lot to other people. I engage with everybody in the company on a regular basis. I try to have one-on-ones with, we have close to 75 employees. I try to speak to every single one of them over the course of ⁓ the year.
And I just try to stay engaged with younger people and it keeps me sharp.
