In this episode
Gary Lombardo, Chief Marketing Officer at Athos Commerce, explains what breaks inside a large marketing organization when 60% of shoppers start their research on AI platforms the brand doesn’t own — and why only 14% of consumers find product information consistent across the channels they move through. Athos works with more than 2,700 brands across 50 countries, including Adidas, Burberry, New Balance, Clarins and Marks & Spencer, and Lombardo’s argument is that the visible symptom looks like a content problem but traces back to product data almost every time. He and Greg Kihlström work through why generative engine optimization is a different job from SEO, what has to be true underneath a conversational shopping assistant before it goes live, what a leader can credibly measure when attribution stops working, where the line sits between AI that helps a merchandiser and AI that quietly makes the merchandising call, and where brand preference survives when the first evaluator of a product is an agent comparing structured attributes.
Key takeaways
- The org chart breaks before the technology does. Marketing teams are still structured around channels they can own, measure and tag. When 60% of customers research through tools that can’t be instrumented, no one’s job description covers “what does AI say about us.”
- Inconsistent product information is a data problem wearing content’s clothes. The wrong price in a marketplace or a missing size on a social post is the symptom; the cause is that most brands have no single canonical structured source of truth — a PIM that’s 80% right, a hand-patched feed file, and a merchandiser spreadsheet that’s more current than either.
- Framing the problem as data changes who owns it. Called a content problem, it belongs to marketing. Called a data and decision-logic problem, it becomes a joint mandate across marketing, e-commerce ops and IT, with one team accountable for the feed pipeline itself.
- Feed accuracy moves revenue. Athos client Accent Group saw a 60% revenue increase over eight weeks after tightening feed accuracy and data consistency.
- SEO is a retrieval game; GEO is a synthesis game. SEO optimizes to be one of the links a human scans and clicks. GEO optimizes to be the fact a model chooses to restate — often with no click at all. The failure mode is porting keyword thinking across; models synthesize structured facts from catalogs, reviews and feeds rather than ranking phrases.
- A conversational assistant needs three things before it goes live: underlying data that is trustworthy in real time, clear guardrails on what it may decide versus escalate, and a named owner who reviews, audits and corrects drift the way you’d own a policy document. An assistant that confidently says something is in stock when it isn’t does more brand damage than a slow website.
- When attribution breaks, run a visibility audit, not a new attribution model. Systematically prompt the major AI platforms with the queries real customers ask and track whether and how accurately the brand appears. Pair it with branded search trends — people arriving already knowing the product name or SKU signals machine-mediated consideration.
- The honest line between assistance and automation is whether a human can explain and defend the decision afterward. Flagging feed errors and drafting attribute fixes for approval is assistance. Silent repricing, re-ranking or suppression across live channels with no review step is the system making the call. Accountability stays with whoever owns the channel’s commercial outcome, not the vendor or the algorithm.
- In agentic commerce, differentiation shifts from persuasion to precision. Agents aren’t swayed by hero images or clever headlines; they’re swayed by accurate, complete structured attributes. Athos customer David Jones saw a 66% performance uplift from restructuring product data alone, with no creative changes.
Chapters
- 0:00 — Why your website may no longer be your most important marketing asset
- 1:50 — Gary Lombardo’s background and what Athos Commerce does
- 3:10 — 60% of consumers now use AI tools while shopping
- 3:43 — What breaks first: the org chart, then confidence in the funnel
- 5:18 — Only 14% see consistent product information — content, data, or decision logic?
- 6:20 — A data problem wearing content’s clothes, and who should own the fix
- 8:30 — GEO vs. SEO: retrieval versus synthesis
- 11:23 — What has to be true underneath a conversational shopping assistant
- 14:06 — Measuring visibility when attribution breaks
- 16:11 — Where AI assistance ends and the system making the call begins
- 19:18 — Agentic commerce: where brand preference and differentiation live
- 22:13 — What it takes to stay agile as a CMO
What breaks first when discovery moves to platforms you don’t own
Lombardo’s answer is organizational before it’s technical. Large marketing teams are still built around paid, owned website, email and social — channels that can be owned and measured. The moment a majority of customers are researching through a tool that can’t be instrumented with the usual analytics tags, there’s a structural gap with no owner. The second break is confidence in the funnel itself: budgets, KPIs, calendars and campaigns were all planned against a linear top-of-funnel-to-purchase model, and when discovery starts somewhere invisible, that model doesn’t just get harder to measure — it stops being true.
Why only 14% of shoppers see one consistent brand
It’s tempting to call inconsistency a content problem, because content is where it shows up. Trace it back and it’s almost always product data. Most brands don’t have one canonical structured source of truth: there’s a PIM that’s roughly 80% right, a feed file patched by hand for one channel, and a merchandiser spreadsheet more current than both. That reframe matters operationally, because it moves ownership from marketing alone to a joint mandate across marketing, e-commerce ops and IT — with one team accountable for the feed pipeline that serves every channel, including the brand’s own site.
GEO is a synthesis game, not a retrieval game
SEO optimizes for ranking among links a human will scan. GEO optimizes for being unambiguous enough that a model can lift an attribute and repeat it correctly — often without any click occurring. Teams get it wrong by porting keyword thinking over and stuffing pages with phrases they think ChatGPT, Gemini or Claude are looking for. Models aren’t ranking keywords; they’re synthesizing structured facts pulled from catalogs, reviews and product feeds. If those sources are incomplete or contradict each other, the model gets the brand wrong or leaves it out entirely. GEO starts with data hygiene, not keyword research.
What a conversational assistant needs before you turn it on
Three conditions. The underlying data — inventory, sizing, policy, product detail — has to be trustworthy in real time, because an assistant that confidently promises stock that doesn’t exist causes more brand damage than a slow site. There need to be clear guardrails on what the assistant is allowed to decide versus escalate; a return exception should have a defined boundary rather than being improvised. And someone in the organization has to own the assistant’s answers the way they’d own a policy document, reviewing and correcting drift, because once it’s live it speaks as the brand’s voice and commitment at scale.
The first credible signal when attribution stops working
Rather than rebuilding attribution for channels that can’t be observed, Lombardo’s first move is a visibility audit: systematically prompt the major AI platforms with the category comparisons and product questions real customers ask, and track whether and how accurately the brand shows up. It’s a leading indicator, not attribution. The directional companion is branded search: when machine-mediated consideration is happening, more people arrive already knowing the product name or SKU rather than from a generic category search. It won’t produce a clean path, but it answers the more urgent question of whether the brand is in the conversation at all.
Where the line sits between assistance and automation
The test is whether a human can explain and defend the decision after the fact. AI flagging feed errors, surfacing underperforming products, and auto-generating attribute fixes for a merchandiser to approve is assistance — compressing hours of manual audits into minutes, often more accurately than a person would. Athos customer Early Settler saved 10-plus hours a week through automated merchandising, and Top Tiles roughly halved time spent on manual search-related tasks; in both, a person still approves what ships live. AI repricing, re-ranking or suppressing products across live channels with no review step is the system making the call whether or not anyone decided it should. Accountability doesn’t move because a machine acted — it stays with the merchandiser or e-commerce leader who owns the channel’s commercial outcome.
Where brand preference lives when the buyer is an agent
Brand preference doesn’t disappear; it moves upstream and downstream of the exchange. Upstream, it determines whether the agent considers the brand at all — you can’t be preferred if you’re not in the candidate set, which is the GEO and structured data work. Downstream, it lives in the human who looks at what the agent surfaced and decides whether to trust it; recognition, price memory and past experience still shape whether someone accepts the top pick or asks the agent to look again. Inside the exchange itself, differentiation shifts from persuasion to precision — which is why David Jones saw a 66% performance uplift from restructuring product data with no creative changes at all.
FAQ
Why do only 14% of consumers see consistent product information across channels? Because most brands have no single canonical structured source of truth for product data. A PIM that’s roughly 80% accurate, a hand-patched feed file and a more-current merchandiser spreadsheet produce contradictory information across the channels shoppers move between.
Is inconsistent product information a content problem or a data problem? Content is the visible symptom — a wrong marketplace price, a missing size on a social post — but the cause is almost always product data. Framing it as a data and decision-logic problem moves ownership from marketing alone to a joint mandate across marketing, e-commerce ops and IT.
How is GEO different from SEO? SEO is a retrieval game: optimizing to be one of the ranked links a human scans and clicks. GEO is a synthesis game: optimizing to be the fact a model chooses to restate, often with no click at all. Models synthesize structured facts from catalogs, reviews and feeds rather than ranking keywords, so GEO starts with data hygiene.
What has to be true before a conversational shopping assistant goes live? Three things: real-time trustworthy data on inventory, sizing, policy and product detail; clear guardrails on what the assistant may decide versus escalate; and a named owner who reviews, audits and corrects the assistant’s answers the way they would a policy document.
How can a marketing leader tell whether their brand shows up in AI-mediated consideration? Run a visibility audit — systematically prompt the major AI platforms with the queries real customers would ask and track whether and how accurately the brand appears. Pair it with branded search trends, since machine-mediated consideration produces more visitors who already know the product name or SKU.
Where is the line between AI assisting a merchandiser and AI making the merchandising decision? The line is whether a human can explain and defend the decision afterward. Flagging errors and drafting fixes for approval is assistance; silent repricing, re-ranking or suppression across live channels with no review step is the system deciding. Accountability stays with whoever owns the channel’s commercial outcome.
About Gary Lombardo
Gary Lombardo is Chief Marketing Officer at Athos Commerce, where he leads marketing, brand strategy, communications, and thought leadership. With more than 20 years of experience in retail technology, ecommerce, and SaaS, he helps retailers navigate the rapidly evolving world of AI-powered commerce. Gary is a frequent speaker on topics including artificial intelligence, digital merchandising, product discovery, customer experience, and the future of retail.
Gary Lombardo on LinkedIn: https://www.linkedin.com/in/garylombardo
———- Resources ———-
Athos Commerce: athoscommerce.com
The Agile Brand podcast is brought to you by TEKsystems. Learn more here: https://aglbrnd.co/r/2868abd8085a9703
We’re proud to be a media partner for #MAICON26 – Oct. 13-15! Learn how AI can power your marketing and business and help you grow smarter. Use code AGILE150 to save! https://aglbrnd.co/r/7fe458ced0f04658
Reach your customers with Reddit. Spend $500 in ad spend, get $500 back in ad credit! Learn more: https://advertalize.com/r/491818c79fb1873f
Chaser is the only Slack-native project management platform that helps teams turn messages into tracked tasks, automate follow-ups, and maintain team-wide visibility, without adopting another tool. Now integrated with Claude and other GenAI tools. Learn more at trychaser.com and use code AGILEBRAND for a 3-month free trial (normal trial is 14 days).
The most influential minds in software, AI, and engineering leadership will be at WeAreDevelopers World Congress North America, September 23-25 in San Jose. Learn more: https://aglbrnd.co/r/60a7299222a7bcf1
Start building your own apps with Replit and get $20 off. Learn more: https://aglbrnd.co/r/93531742a7625a20
Enjoyed the show? Tell us more at and give us a rating so others can find the show at: https://aglbrnd.co/r/faaed112fc9887f3
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://aglbrnd.co/r/35ded3ccfb6716ba
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
[:00:00] Greg Kihlström: Hi, I’m Greg Kihlström, host of The Agile Brand, and here’s a question for you. If a growing share of your customers now begin their shopping journey on a platform you don’t own and can’t see into, is your website still your most important marketing asset? Agility here isn’t just about reacting to new customer behavior, it’s about building an infrastructure that can meet customers wherever they show up and still have the brand behave like one company when it gets there. Today we’re going to talk about the great fragmentation of product discovery and why traditional marketing funnels are becoming obsolete, strategies for maintaining brand and data consistency across countless channels when only 14% of consumers say it’s done well today, and the role of AI and automation in not just personalizing experiences, but optimizing for new discovery engines. To help me discuss this topic, I’d like to welcome Gary Lombardo, CMO at Athos Commerce. Gary, welcome to the show.
[00:01:40] Gary Lombardo: Hi, Greg. Thanks for having me.
[00:01:42] Greg Kihlström: Yeah, looking forward to talking about this with you. Before we dive in, though, why don’t you give a little background on yourself and your role at Athos Commerce?
[00:01:50] Gary Lombardo: Sure, happy to. I’m the CMO at Athos Commerce. We help brands connect the right shoppers to the right products across every channel they show up in, not just on their own website. that’s become a much bigger job than it used to be because, every channel now includes AI platforms, social platforms, and marketplaces, that brands don’t control. And, to give you a sense of scale for us, we work with, more than 2,700 brands, over 50 countries, with brands like Adidas, Burberry, New Balance, Clarins, Marks & Spencer, and others, spanning across all different industries from beauty, sports, and, and fashion, amongst others. and our roots go back to 2007, so we’ve had a long runway, to watch this discovery platform, this discovery problem, rather, evolve and now accelerate
[00:02:35] Gary Lombardo: with AI. before this, my background has really been in the intersection of commerce, e-commerce, search, and data. I’ve been in the industry for, for a while, and I think these skills that I bring to the table sort of are, are at the right moment, right? Which, which this sort of confluence of, AI kind of changing the landscape of commerce, is an opportune, time for me to be at, in, at Athos, and in part of the industry. So excited to be here and, yeah, we got some good data that I think we’ll delve into as well in a recent report that we, we, released with Draper, so…
[00:03:10] Greg Kihlström: Yeah. Yeah. Let’s, let’s dive into that actually right now. And so your recent connected consumer research, well, does a few things. One, one of the things it highlights, though, is a major shift with 60% of consumers now using AI tools while shopping. I know we saw a, a big part of that and that growth last holiday shopping season, but it’s, it’s certainly continued. So when discovery starts somewhere a brand doesn’t own and can’t truly observe, what’s the first thing that breaks inside a large marketing organization?
[00:03:43] Gary Lombardo: Yeah, that’s a great question. usually it’s the org chart that breaks first, I think, before the technology actually does.
[00:03:49] Greg Kihlström: Sure.
[00:03:49] Gary Lombardo: Most large marketing teams are still structured around channels that they can, own and measure, like paid, the website, owned email, maybe social, maybe some others. And, the moment that 60% of your customers are researching through a tool you don’t own, don’t control, and, can’t, can’t instrument with your usual analytics tags, for instance, you’ve got a structural gap. There’s no one, no one’s whose job it is to, to own what does, AI say about this. Although that’s starting to change as well as, as we become smarter with AI. we see this gap constantly really, a lo- lo- with a lot of brands, including in our own customer base, right, across the 2,700 brands that, that we, we run into. But like I said, they’re getting smarter about it as well, and but certainly than they were, say, like 18 months
[00:04:39] Gary Lombardo: ago. I think the second thing that breaks is the confidence in the funnel itself. So teams-
[00:04:44] Greg Kihlström: Mm.
[00:04:44] Gary Lombardo: .. built their whole planning cycle, like budgets, KPIs, calendars, you know, campaigns, et cetera, around a linear top of the funnel to purchase model. when discovery starts on a platform you can’t see into, that model doesn’t, doesn’t just get harder to measure, but it stops, like, being true, right? So organizations don’t org– reorganize around a new, this new truth quickly. That’s really the, the first crack. So I really think it’s about the, the org chart and sort of the funnel being changed, in, in terms of, what we’re seeing today in, in marketing organizations.
[00:05:18] Greg Kihlström: Yeah. Yeah, definitely. I mean, the, the funnel, it’s, it’s breaking a lot of what was held to be [chuckles] just a constant and true, right? So it’s definitely… I, I think the other thing is the, and, and your report certainly says as much, is, you know, only 14% of consumers find product information consistent across channels, and, and what we know is consumers are jumping from channel to channel. I mean, they’ve been doing this for, for a while, but it seems to be just the norm now, and yet, you know, again, less than 15% are finding consistency in, in product information. When you trace that back to its source, is this inconsistency, is it a content problem, a data problem,
[00:06:03] Greg Kihlström: a decision logic problem? You know, what, what, what is it? Maybe all of the above. Like, you know, what, what is it– And does the answer change, who inside the company owns fixing it?
[00:06:15] Gary Lombardo: Yeah, for sure. I think it’s tempting to call it a content problem because that’s the visible symptom, right?
[00:06:20] Greg Kihlström: Yeah.
[00:06:20] Gary Lombardo: Like, you get the wrong price on, in a marketplace, or maybe there’s a missing size on a social post, whatever it might be. but if you trace it back, it’s almost always a data problem. I’d say kind of wearing content’s clothes, so to speak, right? So most brands don’t have one canonical structured source of truth for product. They have, you know, PIM, product information management system, that is 80% right, right, a feed file that gets patched maybe by hand by one, one channel, and a merchandiser spreadsheet that is more current than the others, that than either, right? So i- it’s this, this data, this product data problem specifically, that’s at, that’s the core of it. And we’ve seen this with our clients as well. Like, one of our big clients called the Accent Group, helped tighten
[00:07:06] Gary Lombardo: up their feed accuracy and data consistency through Athos, right? They saw a 60% increase when they, when they did this in, in their revenue over the course of eight weeks. Really, it’s kind of being able to ensure that that product data is, is true, is, is solid, right, so that you don’t have this, this disconnect, right, across what consumers are finding, across channels. And yeah, kind of around the question of ownership, you know, really who owns it, it, it– I think it, this, if we think of it as product being the center, it does change who owns it, right? I think it’s, if you try to frame it up as a content problem, marketing owns it, right? It’s pretty clear but if you, if you frame it up as a data and decision logic problem, which I think it is, becomes a joint mandate across marketing, e-commerce ops, and IT with, sort of one team accountable for this, this feed pipeline, right, itself, the, the pipeline of that data that, that needs to go off to the various channels, whether including your own website, to ensure that it’s, it’s accurate. So that, that team needs to really be on top of it. And that, I think that reframe for organizations really is, is, is really the, the key unlock when they start to think about, how can we get better and, and not have this, this, as we saw in our research, this 14% of consumers not finding the product information consistently across the channels? it’s been a problem, as you said, for a long time, but I think it’s, it’s exacerbating
[00:08:26] Gary Lombardo: now, particularly with the advent of AI.
[00:08:30] Greg Kihlström: Yeah. Yeah. I wan- I wanna get back to the, the funnel conversation a little bit here and talk a little bit about, GEO, generative en-engine optimization, sometimes called AEO, or there’s probably other acronyms out there too, but you know, certainly getting a lot of attention. We’ve talked about it a few times on the show already as well. I think one of the common, misconceptions i- in some organizations is that it’s mo- it’s more similar to SEO than it actually is. [chuckles] And so, you know, just what, what do you see that’s structurally different about that as compared to just saying, “Oh yeah, let’s put the SEO team,” or “Let’s, you know, kind of apply the same lessons we learned from SEO to
[00:09:15] Greg Kihlström: it”? You know, what do, and what do teams get wrong when they treat it as, as such, as just sort of another, another flavor of SEO?
[00:09:24] Gary Lombardo: Yeah, for sure. And, and I think this is– it’s still early on, even though we’ve been talking about GEO, AEO for seems like forever now, but in reality-
[00:09:32] Greg Kihlström: Right, right. [chuckles]
[00:09:33] Gary Lombardo: … it’s been a pretty, pretty short timeframe, and kinda similar. You know, we think back, I’m old enough to remember when SEO, right, was the hot term, right? We’re all trying to figure that out, and I think that’s kinda where we’re at as well. But SEO is fundamentally like a retrieval game, right? You’re optimizing to be one of the, sort of 10 blue links or whatever it is on the, on the cert page that a human will scan and, and click. GEO is, is really a synthesis game. You’re optimizing to be the, the fact the model is choosing to restate, right? Oftentimes there’s no click at all, right? So those are really different jobs, really vastly different jobs. One’s rewarding the ranking, right? The other’s rewarding being un- unambiguous enough that the model can lift your attribute and repeat it correctly, right? So that’s two,
[00:10:18] Gary Lombardo: I think, very fundamentally different things, and we’re all trying to figure out, okay, great, what does that mean? How do we best do it? And I think the mistake teams, make is porting over the, the, the, the concept of SEO, specifically around keyword thinking, right? Stuffing the pages with phrases they think ChatGPT, Gemini, Claude, et cetera, is searching for, right? But these models aren’t ranking the keywords necessarily. They’re synthesizing s- structured facts pulled from, the catalog, you know, from reviews, from your, your, your product feed, right, wherever they’re trained or grounded. So, and again, it comes back to that sort of central problem we talked about earlier around the, the product data being at the core, and if that’s inconsistent, incomplete, or, or contradicts itself across sources, and again,
[00:11:03] Gary Lombardo: remember that 14% consistency number we talked about earlier-
[00:11:07] Greg Kihlström: Right
[00:11:07] Gary Lombardo: … the model’s gonna get you wrong or just leave you out entirely. So I think GEO ultimately starts with data hygiene, right, not keyword search, so that’s the– or keyword research. That’s the, the fundamental difference.
[00:11:23] Greg Kihlström: Another thing, not, not necessarily a new thing, but it’s certainly enabled more, to be more robust with, with AI is, you know, adding conversational assistance to, to websites, and now with the benefit of, of LLMs and, and, and all the other things, to kind of… I mean, I think it, now it makes a lot of sense because consumers are used to interacting with Claude or ChatGPT or something, and so let’s, let’s make a similar experience on the site. But again, similar enough issues, di- different, different solution perhaps, but similar enough issues of, okay, where is that assistant on, on your site getting the data? Like, what has to be true underneath the, the surface for a, a conversational
[00:12:08] Greg Kihlström: assistant to be truly helpful and answer accurately?
[00:12:13] Gary Lombardo: Yeah, that’s a great question ’cause we’re seeing, you know, conversational- … commerce, as we call it, right, an assistant or just a transformation, at least on the front end, how consumers are interacting from a shopping perspective ta- really taking hold now, and I think it’s gonna accelerate over the next few, few months, for sure. And I think three things really need to be in place before you, before you kinda ha- you know, hand over, hand over is maybe not the right term, bit strong, but-
[00:12:38] Greg Kihlström: Mm
[00:12:38] Gary Lombardo: … but in a lot of cases you are, right? But sort of enable, right, this-
[00:12:41] Greg Kihlström: Yeah
[00:12:41] Gary Lombardo: … this agents, this, this, this conversational assistants to be on your website or wherever it may be. you know, the first thing is that, that back to the data, right? It nee- that underlying data has to be trustworthy in real time. So, the, the inventory, the sizing, the policy, the product information, right? Because an assistant that confidently tells a customer something is in stock when it, when it’s, when it isn’t, or does, can, can do more brand damage than a, a slow website-
[00:13:07] Greg Kihlström: Right
[00:13:07] Gary Lombardo: … ever would-
[00:13:07] Greg Kihlström: Right
[00:13:07] Gary Lombardo: … for instance. So I think secondly, you need like clear guardrails in what the assistant’s allowed to decide versus escalate, right? A return exception or an, for, for example, should have a defined boundary and not be improvised in, in the moment, for, for fairly obvious reasons, right? and then I think third, someone in the organization has to actually own the assistant’s answers the way you’d own a policy document. So reviewing, auditing, correcting drift, right? Really taking a look at, you know, what, what’s been said because once it’s live, it’s your brand’s voice and your brand’s commitment, at scale without a human in the loop on every conversation. So you wanna have some level of control over that.
[00:13:52] Gary Lombardo: I think we’re a ways away from, you know, the perfect conversational assistant being there and providing the right answer all the right time. I think there always needs to be this, this human element that, that ultimately is doing that review. So I think it really comes down to those three things.
[00:14:06] Greg Kihlström: Yeah. Yeah. Well, and I wanna talk about, measurement here as well, and maybe, maybe going back to the, the, the new or the most recent iteration of the funnel. You know, a customer starts on an AI platform, they go to social media, and then they end up in a direct purchase. Obviously the, the, the traditional brand funnel as well as traditional attribution models break, frankly. You know, ra- so rather than trying to rebuild attribution for channels that are difficult to observe and, and sometimes not able to be observed directly, what’s the first credible signal that a leader can use to tell whether a brand is even showing up in some of those machine-mediated consideration steps?
[00:14:53] Gary Lombardo: Yeah. I think the, we’re all so used to saying, “Okay, what converted?”
[00:14:57] Greg Kihlström: Right.
[00:14:57] Gary Lombardo: Right? It’s, it’s, I think with these, these AI platforms, it’s like, do we even exist? Do we [chuckles] do we even exist in the answer? So I think the most credible first signal is systematically prompting the major AI platforms with queries your, your real customers would ask, like category comparisons, you know, or specific product questions. You know, what’s the best product for X or Y or whatever it might be, and tracking whether and how accurately your brand shows up. that’s a visibility audit, right? Not attribution, but it’s the leading indicator everything else depends upon in, in this new world of, of thinking about how, how your, your brand is coming across on AI and never mind your, your products. Are they being, being surfaced? I think the, the second is, the second signal is directional but pretty useful. I think it’s pairing that with direct branded search traffic trends. So
[00:15:47] Gary Lombardo: if a machine-mediated consideration is happening, for instance, you’ll, you’ll often see people arrive already knowing your product name, arriving to your site-
[00:15:56] Greg Kihlström: Mm-hmm
[00:15:57] Gary Lombardo: … or SKU rather than arriving from a generic category search. So it won’t give you a clean attribution path necessarily, but it tells you whether you’re in the conversation at all, which frankly is the, the more urgent question right now.
[00:16:11] Greg Kihlström: Yeah. Yeah. Well, and you know, ta- talking about operationalizing this as well, certainly, you know, we, we’ve talked about it from the customer perspective, but you know, managing product feeds, merchandising for dozens of channels, you know, there’s, there’s an operational strain here, and the, you know, the, the governing consensus seems to be to, you know, quote unquote, “apply AI to it,” which, you know, AI can mean many, many things, but where’s the honest line between a system that helps a merchandiser work faster and one that’s actually making the merchandising call itself, you know? And, and who’s accountable when it’s that second scenario?
[00:16:52] Gary Lombardo: Yeah, for sure, and this is a problem that we’re, we’re dealing with every day at Athos. We got some pretty cool, technologies and, and agents that we recently released to make the merchandiser work faster, right? But with, with, with human intervention, of course. So I’ll, I’ll touch on that a, a little bit too. I think the honest line at the end of the day is whether a human can explain and defend the decision after the fact. So if AI is the f- is like flagging feed errors or surfacing which products are underperforming on a channel, auto-generating attribute fixes on a merch- for a merchandiser to approve, that’s assistance. Like, that’s really the right way to do it. It’s compressing hours of manual audits into, to minutes, really. It’s that efficiency gain. Not to mention accuracy, quite honestly. a lot of the, the, um,
[00:17:37] Gary Lombardo: the, the, the machine, models, right, could do it a little bit more accurately than, than a human. So however, contrast that if, if AI is like silently s- silently, like working behind the scenes, repricing, re-ranking, or suppressing products across live channels with no review step, that’s, the system making the call, whether anyone officially decided that or not. That’s exactly the model, you know, that we, we don’t necessarily wanna have in play, right? I think recommendations of that and making sure there’s that human element. Our cu- like customers at Athos, like Early Settler, for instance, ha- has done very well with the, the, the formers I described, around the, the, the time savings. They’ve saved 10, 10 plus hours a week through automated merchandising. we’ve seen others like, Top Tiles,
[00:18:22] Gary Lombardo: where they’ve roughly halved the time it takes, for them to spend– that they’ve spent on manual search-related tasks, for instance. So in all of these cases, a person’s still approving what actually ships live. and accountability doesn’t Just doesn’t move because a machine made the decision, right? [chuckles] No matter where it is, it stays with whoever owns the channel, channel’s, commercial outcome. So typically that’s the merchandiser or e-commerce leader, not the, not the AI vendor. It’s, it, or-
[00:18:49] Greg Kihlström: Mm-hmm
[00:18:49] Gary Lombardo: … the algorithm, right? our view at Athos, as I, you heard me say before, is that technology should, should make the merchandiser faster, and better informed and more accurate, right? With visible, with a visible audit trail and not quietly replace their judgment or actually what they’re doing, their job, for instance, completely. the moment a team can’t really explain why a product got to a, to a certain, got a certain treatment, they’ve crossed the line, without really realizing it.
[00:19:18] Greg Kihlström: Well, and, and I wanna talk a little bit about some, some present day, but, you know, moving to the future as well, where the buyer is an agent, right? And we’re, we’re seeing this already in, in small ways, but certainly, you know, Athos uses the term agentic commerce to, to describe this, this future. If the first evaluator of your product is an agent comparing structured attributes, you know, price, availability, specs, return policy, as opposed to, what humans may evaluate on, what … You know, where, where does brand preference live in this scenario? And where does, you know, differentiation actually live in, in this exchange?
[00:20:07] Gary Lombardo: Yeah, for sure. I … And that’s a, that’s a great question. You know, as, somebody who thinks about the brand, I’m always wondering [chuckles] this new world like-
[00:20:14] Greg Kihlström: [laughs]
[00:20:14] Gary Lombardo: … are we in control? Or is it just like we talked about, the funnel breaking, you know, brands to a certain degree breaks. But, but I don’t think brand preference doesn’t disappear in the new world. It really moves kinda upstream and downstream of that exchange. upstream it lives whether the agent is even considering you, as we talked about before. Are you being found, right? Like, that’s, that’s the GEO and structured data work we, we talked about earlier. You can’t be preferred if you’re not in the candidate set, for instance. So just that piece of your br- you know, concept of branding still needs to exist. downstream it lives, I think, in the moment the, the human, right? Because ultimately the shopper is still making a decision, looks at what the agent surfaced and decides whether, whether to trust it. Brand recognition, you know, price memory, you know, past experience still shape whether someone accepts the agent’s top pick or asks it to look again, right? So
[00:21:05] Gary Lombardo: eventually, you know, maybe we reach this agent-to-agent world and what happens there, different question, but I think largely it’s, you know, agent to human at the end of the day making the actual shopping decisions. So I think the, those two elements of a brand are important. But within this exchange, I think it’s differentiation shifts from, persuasion to precision. So if an agent isn’t, swayed by, you know, agents aren’t necessarily swayed by, like, big beautiful hero images or clever headlines it’s swayed by more accurate or complete structured attributes. And we’ve seen this precision over persuasion effect show up concretely with some of our customers. Like David Jones, one of our department store customers, saw, 66% performance uplift simply by tightening up how their product data was structured and represented with no creative changes, right, necessarily involved, no quote-unquote classic branding changes. So, so paradoxically, I think in the agentic world, being scrupulously honest and complete about your product data, becomes a competitive advantage in its own right, and the brands that win or the, the brands that are gonna win are the ones that, ones an agent can trust enough to recommend confidently.
[00:22:13] Greg Kihlström: Yeah. Yeah. Well, Gary, thanks so much for joining today. I’ve got one last question for you as we, as we wrap up here. What do you do to stay agile in your role, and how do you find a way to do it consistently?
[00:22:25] Gary Lombardo: Yeah, it’s a great question. I think, for all of us marketers out there, a lot of us are like, “Hey, what, [chuckles] how do we stay relevant in this, this AI world?” And I think marketing, in my opinion, is more important than ever in this AI world. And for me, you know, staying agile is really, making sure that I, I am talking to our customers, thinking like our customers, actually, shopping, right, through the AI platforms, you know, what products are showing up through social, not just the AI platforms necessarily, thinking like the, the, the end consumer, ’cause the market’s really moving faster than, than ever. So it’s acting like customers, talking to our customers, really hearing what they’re going through firsthand. And I think operationally too, in our team, team that I run, it’s like building out, a standing rhythm of, of small experiments rather than waiting for a big annual plan to be right. So testing a channel, test the results, adjust, et cetera. So agility is really something that I think is at the core of it. So, and of course, like everyone else too, I’m, I’m in my fingertips and, and trying to get into AI and figure out how we automate [chuckles] some of our marketing processes and do things smarter internally, which, you know, when you got a small team and limited resources, it’s, it’s critical to do that, and also a challenge to do it at the same time. So those are the things I’m constantly thinking about to stay, to stay agile in my role.
