Episode Summary
In the first episode of AI Commerce Innovators, host Anna Luo of DaVinci Commerce asks three leaders in agentic commerce how products get discovered and chosen when shoppers ask an AI instead of searching. Chip Knicker of Accenture describes what Anna calls the agentic shelf, where a shopper may be shown one or two recommended products rather than pages of results, walks through Accenture’s simulation of which industries stand to gain new demand from AI agents and which face more execution risk, and says every brand needs to be present in general-purpose agents before deciding whether to build its own. Emily Pfeiffer of Forrester and Diaz Nesamoney of DaVinci Commerce then explain why crawled product data falls short, why brands need to describe products the way shoppers describe their needs, and how awareness and consideration can now happen in answer engines while most purchases still happen on websites.
Key takeaways
Measure how your products show up in AI answers today. Chip Knicker says Accenture’s clients can measure how their products are showing up in LLMs, how they rank against competitors, and what the citations look like that are driving those recommendations
Plan for every product category, not only the obvious ones. Chip Knicker says every product and every industry is going to be impacted, and probably already is in some form; he expects repeatable grocery and consumables purchases to change first and considered purchases to be a pretty fast follow
Be present in general-purpose agents first, then decide whether to build your own. Chip Knicker calls becoming the choice of agents, being present in all the general-purpose tools, something every organization needs to do; becoming an agent of choice, a specialized agent on a surface the brand owns, is a decision, and he says it could be a competitive advantage
Describe products the way shoppers describe their needs. Diaz Nesamoney says shoppers ask LLMs about needs and occasions rather than simply asking for a product, so it becomes quite important for brands not just to publish their catalogs into the LLMs but to spend time enriching the content to match how consumers ask
Work out how AI will see your inventory and delivery information. Emily Pfeiffer says the LLMs that crawled websites for product data never could see inventory and delivery information, and that commerce requires operational data; figuring this out in a more intentional and planful way is going to be really important for digital commerce
Look for new-to-brand shoppers on the LLMs. Diaz Nesamoney says LLMs have a higher proportion of Gen Z users, tells brands that want new-to-brand customers that this is where they are, and says a brand that figures out how to get there and engage them is going to gain market share very quickly
Keep your own site evolving toward conversation. Emily Pfeiffer says she thinks it won’t take long before shoppers get used to natural-language queries and expect websites to handle them too, so keeping on top of the owned environment and making sure it evolves is really important
Start experimenting gently now. Emily Pfeiffer says to be wary of the investment because this is new and unproven, but she would not entirely wait
Treat LLMs as a new way customers engage with brands, not just another channel. Diaz Nesamoney says brands should not sit on the sidelines hoping to jump in when it is mature, because he thinks the market share issue is a real issue for brands that do not figure out how to play in the conversational model
About the guests
Chip Knicker
Chip Knicker is a managing director at Accenture Song, where he leads agentic commerce and his team helps clients with everything from how their products show up in AI answers to how agents could change merchandising and other operations. In this episode he discusses the industry simulation behind Accenture’s agentic commerce report, Make your brand unmissable.
Emily Pfeiffer
Emily Pfeiffer is a principal analyst at Forrester, where she advises digital business leaders and covers commerce technology, including agentic commerce, order management systems, B2C and B2B commerce, commerce search and drop shipping. Her research on the state of agentic commerce is discussed in this episode.
Diaz Nesamoney
Diaz Nesamoney is the founder and CEO of DaVinci Commerce, which describes itself as the agentic commerce experience platform for global brands and retailers. He was previously a co-founder, President and COO of Informatica, which went public in 1999, and the founder and CEO of Celequest, which Cognos acquired in 2007. He holds seven technology patents.
In this episode
| 00:36 | Welcome: more and more, shopping starts with a question to an AI |
| 01:40 | What agentic commerce is, and why Accenture sees a foundational shift |
| 03:08 | From the physical shelf to the digital shelf to the agentic shelf |
| 04:58 | Off-platform visibility, owned surfaces and agentic commerce operations |
| 07:16 | Accenture’s simulation: new demand and execution risk by industry |
| 09:27 | Every product and every industry will be affected |
| 10:47 | Horizontal agents and vertical agents |
| 11:37 | Becoming the choice of agents, and deciding whether to be an agent of choice |
| 13:01 | Brand stores inside ChatGPT |
| 15:55 | The panel: how products get discovered and chosen by answer engines |
| 16:27 | Generative AI, agentic AI and AI agents, defined |
| 19:32 | Owned and non-owned environments, and what exists today |
| 22:33 | Why crawled product data falls short |
| 24:26 | Matching products to needs and occasions |
| 25:01 | The inventory and delivery data crawling never saw |
| 26:25 | Brands wake up to the consumer base on LLMs; regulated categories and service products |
| 28:32 | It is early days, and everyone is still figuring it out |
| 29:56 | Answer engines earn more of discovery and consideration |
| 32:44 | High-consideration products first, then conversation across the whole funnel |
| 36:46 | Early adopters, Gen Z and new-to-brand shoppers |
| 38:02 | Three things brands can take away |
| 38:48 | Experiment now, but be wary of the investment |
| 39:14 | LLMs are not just another channel |
In their words
“It becomes quite important for brands not just to publish their catalogs into the LLMs, but actually to spend time enriching the content and describing their products in a way that match how consumers are a-asking for or expressing their needs.”
Diaz Nesamoney
“There’s something that every organization needs to do, which is become the choice of agents, and we call that a bit of a baseline.”
Chip Knicker
“Answer engines like Perplexity and ChatGPT and Google Gemini they are earning more of the product discovery and consideration processes in the shopping funnel.”
Emily Pfeiffer
“It’s okay, and it’s early days, and everyone is still figuring this out, and you’re doing the right thing because you’re educating yourself and you’re learning more, and that’s all you can do.”
Emily Pfeiffer
“It’s really a new way of consumers engaging with brands to buy.”
Diaz Nesamoney
Resources
The host and guests
Anna Luo on LinkedIn: Host of AI Commerce Innovators and Chief Marketing & Customer Innovation Officer at DaVinci Commerce
Chip Knicker on LinkedIn: Managing director at Accenture Song and its agentic commerce lead
Emily Pfeiffer on LinkedIn: Principal analyst at Forrester
Diaz Nesamoney on LinkedIn: Founder and CEO of DaVinci Commerce
DaVinci Commerce: The agentic commerce experience platform Diaz Nesamoney founded
Research discussed
Agentic commerce: Make your brand unmissable: Accenture’s report on staying visible, chosen and trusted as agentic commerce expands. It is the source of the industry chart on net-new demand and execution risk, the simulation run with Aaru, and the ideas of horizontal and vertical agents
The State Of Agentic Commerce In Mid-2026: Emily Pfeiffer’s post on Forrester’s blog summarizing her research on the state of agentic commerce, including chat-based product discovery and how brands weigh owned and non-owned conversational commerce experiences
Ideas and terms discussed
Agentic commerce: Emily Pfeiffer’s working definition: any generative AI agent or conversational experience that helps a shopper find, discover, learn about or buy a product. She notes that agentic technically means autonomous
Agentic shelf: Anna Luo’s term for what a shopper sees after asking an AI, following the physical shelf and the digital shelf. Chip Knicker describes it as maybe one or two recommended products, sometimes from the tail of an assortment
Off-platform and on-platform visibility: Accenture’s split between how products show up in LLMs and the surfaces a brand owns, such as its website and mobile app
Agentic commerce operations: Accenture’s third area: how agents could change the way a commerce business runs, with merchandising, supply chain, replenishment and recommendation as the examples Chip Knicker gives
Horizontal and vertical agents: General-purpose agents such as ChatGPT, and specialized agents on a surface the brand owns that can use first-party data
Choice of agents and agent of choice: Being present in general-purpose agents, which Chip Knicker calls a bit of a baseline for every organization, and building a specialized agent of a brand’s own, which he calls a decision
Owned and non-owned environments: The spaces a merchant controls, such as its website or app, and those it does not, such as a marketplace or an answer engine like ChatGPT
Also named on air
Aaru: The partner Accenture ran its simulation of 50,000 users with
ACP and JSON-LD: Among the structured ways of publishing product content to LLMs that Diaz Nesamoney names
Perplexity, ChatGPT and Google Gemini: The answer engines Emily Pfeiffer names as earning more of the product discovery and consideration in the shopping funnel
Related AI Realized episodes and events
AI Search Visibility: When AI Says Your Company Is Dead: Curtis Sparrer of Bospar on how AI answer engines decide what a company is, and how to correct the record when one gets it wrong.
Signal Integrity Is the Foundation for AI Go-to-Market Agents: Lara Shackelford of Hawksmoor.ai on the customer signals an AI go-to-market system runs on, and on what a large language model looks at when it represents a company.
AI for Go-To-Market: The New Revenue Team Playbook: Jonathan Kvarfordt on how revenue teams are being rebuilt around AI, and how he measured a generative engine optimization and public relations program.
Frequently Asked Questions
-
Agentic commerce is the use of AI agents or conversational AI experiences to help shoppers find, discover, learn about or buy a product. That is how Emily Pfeiffer, principal analyst at Forrester, says the term is used in practice, though she notes that agentic technically means autonomous: an AI agent doing something without a human expressly telling it to. She says that today the main form is conversational product discovery, on brands’ own websites and in answer engines, the second of which she calls assistive rather than agentic. Some answer engines also offer a little native checkout, which comes and goes.
-
Generative AI is any AI system that generates content, agentic AI technically means AI that acts autonomously, and an AI agent is a program trained for a very specific type of task. Emily Pfeiffer of Forrester says AI agents are easy to confuse with chatbots because people think of a customer service agent as someone to converse with, but one agent might handle a conversation with a shopper while another interprets a product photo to decide its attributes. She describes agents as the thing under the covers that end users do not know are there doing the work in the background.
-
The agentic shelf is the set of products an AI shows a shopper in answer to a question, following the physical store shelf and the digital shelf of e-commerce. Anna Luo, host of AI Commerce Innovators, uses the term, and Chip Knicker of Accenture describes what is on it: what the LLM recommends based on the search, its context and the other data it has found, rather than the results a brand has placed at the top of its own site. He says a shopper gets maybe one or two products recommended, sometimes from the tail of an assortment, instead of scrolling through a whole shelf.
-
Brands can help LLMs find and recommend their products by publishing product content through structured standards rather than relying on crawling, and by describing products the way shoppers express their needs. Diaz Nesamoney, founder and CEO of DaVinci Commerce, says LLMs were not designed for commerce and learned about products by crawling, so a brand often finds itself described with a hodgepodge of information. Standards such as ACP and JSON-LD are a great start, he says, but shoppers ask LLMs about needs and occasions, such as what to wear to a friend’s wedding on a hot weekend, so it becomes quite important for brands to spend time enriching their content to match.
-
Travel and hospitality, beauty and personal care, and health and wellness rank toward the top for new demand from AI shopping agents in Accenture’s simulation and automotive ranks higher on execution risk, but Accenture expects every industry to be affected, and each probably already is in some form. Chip Knicker of Accenture says the simulation was run with Aaru across 50,000 users in a number of countries, that grocery gets a lot of attention for commodity and repeat purchases, and that automotive faces execution risk because of complicated fitment, parts and service. He expects repeatable purchases to change first and considered purchases to be a pretty fast follow.
-
Whether to build its own AI shopping agent is a decision for each brand, while being present in general-purpose agents such as ChatGPT is something every organization needs to do. Chip Knicker of Accenture says being present in general-purpose agents is what Accenture calls becoming the choice of agents, a bit of a baseline; building a brand’s own, a vertical agent on a surface the brand owns that goes deeper in its domain and draws on first-party data, is becoming an agent of choice, which he says could be a competitive advantage. He describes a brand store inside ChatGPT as a great place to start for a lot of Accenture’s clients, because it puts the brand inside a widely used tool while giving it its own curated environment.
-
AI product discovery is harder for regulated industries because their content needs guardrails, such as age gating for alcohol, before it can be published, and harder for service businesses because the structured ways of publishing product content to LLMs do not yet cover services. Diaz Nesamoney of DaVinci Commerce says financial services firms and insurers that want to be in front of consumers looking for loans or insurance are trying to figure out how to get around those limitations. He says the first mechanisms were designed for commerce of physical products, replicating e-commerce sites, and the infrastructure is not quite there yet for everything else.
-
AI is shifting the upper parts of the shopping funnel, awareness and consideration, toward answer engines such as ChatGPT, Perplexity and Google Gemini, outside the spheres of control that merchants have. Emily Pfeiffer of Forrester says that in search, some awareness happened on the search engine and the consideration happened on a website, but it is now at least possible for a whole lot, maybe all, of awareness and all of consideration to happen in an answer engine, through a conversation. She says almost all conversion still happens on websites today, though that may shift too, and Diaz Nesamoney of DaVinci Commerce expects conversational experiences to pull the whole funnel into the LLMs.
-
[00:36] Anna Luo: Shopping used to start with a search bar. More and more, it starts with a question going to an AI. I’m Anna Luo, your host, and welcome to Nobody’s Browsing Anymore, They are Asking. In this episode, I’m pleased to introduce three seekers who are leading authorities on agentic commerce in their own right. Chip Knicker, managing director and agentic commerce lead at Accenture. Emily Pfeiffer, principal analyst of Forrester. Diaz Nesamoney, founder and CEO of DaVinci Commerce. I’m kicking off this episode with Chip, who’s here to share Accenture’s latest research on where agentic commerce is headed and how different industries respond to this tectonic shift. Welcome, Chip. The term agentic commerce has been thrown around in recent days. How does Accenture look at it, and what does the transformation of commerce look like?
[01:40] Chip Knicker: Yeah, we see it as a very foundational shift. We’ve been in the era of digital commerce for, the better part of 25 years, where we’ve all been familiar with buying products online and, the Accenture team has been helping, clients of ours set up the technology to to enable digital commerce. And been a great time in, in our professional lives, a lot of growth for our clients there. But as we move into the agentic, commerce world, we see this, this seismic shift or tectonic shift, like you said, where, we’re getting to a point where all of the, users and folks that are searching for a product are looking for things using intent-led searches. Saying they’re, looking for a pair of shoes for an event they’re going to and, starting very general without brands listed. And brands are rec- or the LLMs and are recommending products back to the customer based on their searches that don’t necessarily have a brand involved and may not list a product in the same way they used to. So there’s a lot of concern about that in terms of visibility, discoverability, and so that’s where we are now. And I think as, as this continues to evolve, we’re gonna see a lot more autonomous orchestration, so perhaps an agent buying something on your behalf. I think we’re also gonna end up seeing some agent-agent collaboration as we get into more support and change in the operation of a commerce business. So if you put all that together, we see this as just the, perhaps the biggest transformation since the onset of digital commerce.
[03:08] Anna Luo: When you talk about transformation, I’m really interested in the evolution, right? ’Cause before, commerce is just literally going to the shop. It’s on the shelf, physical shelf, and then after that you’re seeing e-commerce, and you have the digital shelf. And now what we have is what we call agentic shelf. Can you elaborate on that?
[03:29] Chip Knicker: Yeah, I think what you’re seeing now is customers you know, putting in a, an intent-led search, as I mentioned earlier, and, what shelf they see is really based upon what the LLM tool provides to them and based on the, understanding that tool has about not only the search and the context you provide, but also what other research and what other data that, that LLM has found for that type of product or that type of search. You’re not seeing recommended pages or recommended products, I should say, based on what our, clients have built or put at the top of their their search results, that you’re not seeing kind of what we have always provided in terms of product recommendations to folks. You’re seeing what the LLM recommends, and you may see something or a product that is on the kind of the tail of their assortment. It might be in in a classical way to think about it, the fourth or fifth page of a search result. And now you’re getting served that as the recommended product and really changes the whole nature of a digital shelf. You’re getting maybe one or two products recommended and you’re not getting to scroll through the whole shelf like you’ve been used to. Yeah, really good question. And it’s a-- puts a, I think, a sharp edge on kind of what’s different these days.
[04:33] Anna Luo: Yeah. Again, we’re gonna dive deeper into discoverability, right? Because if only one or two products gets recommended, it’s like, how do I get that onto, to the answer? But before we even get into that what are some of the solutions currently available?
[04:51] Chip Knicker: Yeah, there there’s a number of different things. We do spend a lot of time with clients what we call off-platform visibility. So that’s, that is discoverability GEO, kind of AEO, the terms that we are using these days. There’s a lot of ability for our customers and clients of ours to measure what their, how their products are showing up in LLMs today, how they rank against competitors, what the citations look like that are driving those recommendations. So we spend a lot of time in off-platform discussions. We’re also focused a lot on what we call on-platform, so all the surfaces that you own be it your .com site, your mobile app. There’s other, kind of clienteling applications that get used. That sits in the on-platform kind of area or offering for us, and that’s an important one to really think about how your sites function today. And then the third grouping, Anna that we think about quite a bit is what we call agentic commerce operation. So starting to think about all the ways the way your business runs today could change. And some good examples there are, like, your merchandising team. There’s some really eff- effective ways to be much more efficient in terms of how you merchandise your products, looking at loyalty, looking at service, looking at historical trends, data data that you have, and all those different elements. And then providing an agent-based recommendation is really gonna change the way, in that merchandising example, how merchandising works. And you could, play that across your supply chain. You could play it across replenishment and certainly, kind of recommendation engine. So we put a lot of kind of future focus on that agentic commerce operations offering.
[06:18] Anna Luo: Thank you, Chip. Let’s talk about Accenture’s recently published research paper called Make Your Brand Unmissable. In it is this interesting chart showing the impacts of different industries as AI agents get more and more involved in shopping. The simulated results show industries like travel and hospitality, beauty and personal care health and wellness, and groceries will more likely seeing net new demand with medium execution risk the results also show that home furnishings, automotive, apparel, shoes, accessories are skewing towards lower net new demand and higher execution risk due to AI. Please talk us through the simulations you ran and what it all means in terms of risk and reward for different industries as a result of AI.
[07:16] Chip Knicker: Yeah. Yeah. So we, the Make Your Brand Unmissable, content is something we’re proud to put in market. We actually used a, a simulation of with a partner called Aaru. So we simulated fifty thousand users across across a number of countries to really get a sense for what the shift is and what it means to different industries. And so what you see here is, we ranked what we think kind of new demand will show up based on industry. And again, this is our data across the simulation that we ran, and we see there’s certainly some, some execution risk and, think kind of complexity of the change across this. And so just interesting data points here. You can see on the right-hand side, obviously, in the automotive kind of world that there’s gonna be some challenge because you have complicated fitment, and you have complicated parts and service connection that you be really, thoughtful about and making sure all the LLMs are are ready for that. And you see that kind of rank on the higher end execution risk. Then you can see some other stuff towards the top, with travel and hospitality and beauty and personal care and health and wellness products. And, grocery gets talked about a lot from a commodity and repeat purchase perspective. So we see there’ll be a lot of demand in the next eighteen to twenty-four months from that perspective, and a little less on the execution risk. Still hard. All these clients we work with have a lot to do to transition and move into this new space. But just bit of a, ranking or stratification based on our research.
[08:34] Anna Luo: Yes. And what I find interesting is industries that make high consideration products like cars, luxury watches, and major home c- appliances, they are expensive or complex, and shoppers are likely to regret if they don’t research the products carefully and compare options. And then there are industries that make low consideration products, these impulse buys like toothpaste, general purpose vitamins. They’re inexpensive and very little personal risk if the shopper makes the wrong choice. Should these factors affect how brands or retailers make decisions on their agentic commerce strategy?
[09:18] Chip Knicker: Yeah, I think if, if you look at the, the analysis we did, it, it-- you’ll see some of the more considered products, on the financial service and utilities near the, the bottom ranking there in terms of net new demand. But what I’d say to that Anna, would be every product and every industry is gonna be impacted. It probably already is in some form or fashion, I myself am searching for and looking for things that are not that, considered, what’s the, what’s a, a new supplement to get. I actually did a search like that just the other night and got some good recommendations back. Kind of a, a simpler less kind of personal risk with that, but just made a simple recommendation. And then I’m doing stuff as complicated as, looking for, like just for fun, looking at some different automobile configurations with, just so many different options and engine type and, wheel type and interior and options, and that’s a really considered purchase. And, you can do a lot now in LLMs, and I think as that continues, we’re gonna see, both, the spectrum of, very kind of repeatable transactions that we’re comfortable in a grocery orientation or consumables orientation. I think that’s probably gonna be-- make sense that’s gonna be first in terms of change. Then I think the considered purchases are gonna have I think a pretty fast follow. So what’s interesting about that graph is you see all the industries on there, different rankings based on our current analysis, but, we think it’s gonna impact, every, every dot and every industry on this this chart.
[10:35] Anna Luo: How would you advise brands and retailers in these industries to build a strategy around agentic commerce so they will be unmissable? How should they use AI or AI agents?
[10:47] Chip Knicker: Yeah. The consideration and, we also put together in our research, there’s this concept of kind of a horizontal and a vertical agent. And I’ll answer your question kinda using some of that framing, Anna. Whether it’s travel and hospitality or kinda beauty and personal care, the, the first thing we’re recommending is really be be present with horizontal agents, and that’s kinda more general purpose systems. Think like a ChatGPT that are recommending products back to customers. You have to really be thoughtful about about how you show up in those horizontal agents. And an example or another part of that same kind of idea we have is if you have horizontal agents, you also have vertical agents. And we’re seeing, and we’ve done some work with some clients on, creating a, a, a vertical agent that’s owned, an owned surface that allows allows a, a, a brand to be really personal in their recommendations and be much more data-driven and specialized in how they’re offering their products. And so the thought here is, you create a vertical agent to, draw people in and create, really good connection based on that owned property. And so there’s kinda two different decisions to make. And, the way we like to talk about this and I’ll just kinda finish with this comment, is there’s something that every organization needs to do, which is become the choice of agents, and we call that a bit of a baseline. So that’s, being present in all the horizontal discussions and tools that are out there. And then there’s a decision for clients to make. In that beauty example I shared, if a client of ours would like to become an agent of choice. So that could be a competitive advantage if you wanna focus there. It’s much more deep in terms of domain. There’s first-party data in play. A lot of specialized recommendations. And just to close out your answer for your question is, kind of beauty brands, if they wanna, do that and become a vertical destination I think that could be a real choice of a real place of advantage.
[12:29] Anna Luo: Yeah. It’s interesting that you brought up the horizontal and vertical agents. I’d like to dig deeper into your agent of choice. This sounds quite similar to a brand having a brand store built in the ChatGPT app environment, for example, where the brand can curate all the content in it, speak to the shoppers in its own voice. The brand has full control over the shopper’s experience and creates a much more personalized interaction with the shopper.
[13:01] Chip Knicker: Yeah. I think it’s a good example of that, Anna. I think the brand store app Within ChatGPT is a great place to start for a lot of our, clients. It gives you a chance to, be present in a very used and useful tool, ChatGPT in this instance, and gives you a chance to create your own branded environment, with some curation in there. And I do think that’s, it’s a great example of a vertical kind of agent, but you can create that almost within the horizontal, type orientation. So yeah, we’ve started to work with, clients doing that, and partnered with DaVinci on some of that as well. And so it’s exciting to see, some of these kind of early manifestations, if you will, of the different ways to help our, clients and help brands be present, but then also be present in a way that creates a bespoke experience, much like the, the brand stores enables clients to do.
[13:50] Anna Luo: Speaking of the agentic commerce projects that Accenture and DaVinci Commerce work together with brands across so many industries, right? Like luxury, beauty, personal care, retail, telecommunications, CPG, finance, travel, automobile And interesting enough, these brands pretty much represent all the industries impacted by AI from your earlier chart. It’s impressive to see trailblazers in each of these industries. So broadly, can you comment on the patterns you’re seeing in how they approach agentic commerce?
[14:32] Chip Knicker: Yeah. I think it’s everyone started their journey with agentic commerce. I think all the clients we work with are-- have started, right? There’s some doing discoverability and GEO kind of analysis as starting point. They’re measuring, and they’re figuring out where to go next. And I think some of the, the folks that are a little more trailblazing are taking some other steps, and I think the partnership we have with DaVinci is a good example of a few of those customers that wanna have more of a curated and owned experience. They feel like they can differentiate, there and they wanna test and learn and see how that goes as well. So we’ve had a couple different clients we’ve worked with in different industries, so it’s interesting to see, how those-- the trailblazing nature kind of extends beyond industry too. Yeah I think it’s just a really great way to start to bring this, this change to life and do so, with a tool, ChatGPT, and within a tool, ChatGPT, that allows people to have a real branded and customized experience.
[15:24] Anna Luo: That was Chip Knicker, the agentic commerce expert from Accenture. Chip shared with us the evolution of commerce, the current state of agentic commerce, how AI impacts many industries, and why it’s important for products, both high and low consideration products across different industries, to be discoverable and chosen by AI answering engines. Here, I’m sitting down with Emily and Diaz to dig into the very question that everybody is asking. How does my brand, and more importantly, how do my products get discovered and chosen by answering engines and be experienced in a way unique to my brand? But first, Emily, let’s set a baseline to help everyone understand the state of agentic commerce. So in a few sentences, what are the key differences between generative AI, agentic AI, and AI agents? And also, what is the current state of the agentic commerce?
[16:27] Emily Pfeiffer: I’m so glad you’re asking. These are terms that we hear all the time right now, and sometimes not quite accurately. So generative AI or Gen AI means any AI system that’s generating content. AI has been around a really long time. Gen AI is available to the public only for the past few years or so. It might generate images or conversations or copy, right? So Gen AI is creating content. Agentic AI, agentic anything, technically means autonomous. So when you hear the term agentic, technically it means that there’s an AI agent that’s doing something without a human expressly telling it to do that. The term is used a lot of the time to mean there are AI agents, which I’ll explain in a second, or there’s something h- happening that’s conversational, right? Folks use it to mean a lot of different things, and as it becomes more a part of the vernacular, we’ll all probably have to accept that, and I’ll have to stop being such a stickler about it. But in the meantime, technically agentic means autonomous. Those AI agents that I mentioned are easy to confuse with a chatbot because we think about a customer service agent as a person that we’re conversing with. But that’s a limited way of thinking. They’re not people, these agents, which is hard to remember because we say that they have training and they learn, like we personify them a lot. An AI agent is just a, a system that... a program that is trained in a very specific type of task. So there might be an AI agent whose job is to interact and have a conversational experience with a human user. There might be an AI agent whose job is to interpret an image and decide what product attributes the product has based on the photo, right? There are AI agents that do lots of different things, and even the conversational ones, one might be trained in customer service, another is trained in product discovery. So the agents are really the thing under the covers that end users don’t know are there doing the work in the background.
[18:34] Anna Luo: Great. Thank you for that clarification. And what is the current state of agentic commerce?
[18:41] Emily Pfeiffer: Yeah. Again, agentic commerce is one of those terms that I could argue about the technical definition of, but it doesn’t really matter. Everybody is using it to mean some Gen AI agent or conversational experience that helps us find, discover, learn about, or buy a product.
[19:00] Anna Luo: So let me dig into your research report. It is called “The State of Agentic Commerce.” And you said, and I quote, “Today con- conversational commerce is almost prevalent for product discovery activities in both owned, meaning your own website, and non-owned environments, which could be answering engine or the marketplace.” This speaks to the maturity of how consumers use AI-powered commerce today. Can you share a few highlights?
[19:32] Emily Pfeiffer: Happy to. Yeah. So I wanna make sure that we understand, as you said, the difference between the owned and a non-owned environment. And for the folks who can see the video right now I’m gonna show a slide that explains the owned and non-owned experiences that we can have. So in an owned environment, this is the website or your app. It’s whatever the merchant can control. And non-owned is a marketplace or an answer engine, like ChatGPT. It’s anything that’s outside of the purview of the merchant, the spaces that they can’t control. And there are these sort of bands of types of experiences that we can have in both environments. So conversational product discovery is the first type of experience. On the owned website, the owned environment, that might look like a product discovery chatbot on the website. We’re seeing this migrate a little bit from a little chatbot bubble in the bottom right-hand corner to sometimes an AI assistant pop up with a button up next to the search bar. Either way, it’s conversational, and it’s a place where a customer can say, “Hey, I’m looking for a product with these attributes or something I wanna take on this type of trip.” So that’s-- you may see that as, as you’re shopping. That’s happening more and more. Non-owned environments, th- this is literally what an answer engine is. It’s a conversational experience that increasingly consumers are using to discover products that they might like to buy. So this is assistive. Technically, I would not call it agentic. It’s more assistive. But this is the most prevalent kind of flavor of agentic commerce, if we’re generous with ourselves and call it that today. There’s also commerce transactions, agents purchasing on behalf of customers, and then agents transacting autonomously with each other, agent to agent. And so these are I don’t wanna call it maturity levels, because it may not happen linearly, but these are the types of experiences that we can overlay over the owned and non-owned environments. Today, really all we have is conversational product discovery, as I described, in both owned and non-owned, and there’s a little bit of answer engines that can handle commerce transactions and offer native checkout. That comes and goes. There are some major answer engines that have actually pulled that off of their sites, at least for now. So that’s most of what’s there today, but it continues to evolve.
[21:51] Anna Luo: Yeah. I think your analysis of the conversational commerce resonates well with Diaz’ view on the importance of how individual consumers discover and experience products that th- they may or may not have thought about buying. A question for you, Diaz. You’ve talked to a wide range of retailers and brands spanning industries that includes alcohol beverage, health and wellness, beauty, luxury fashion, travel and hospitality, banking, and it-- the list goes on. What strikes you as their biggest challenge and concern in the discovery phase, and what does being discovered mean?
[22:33] Diaz Nesamoney: Yeah, sure. I have indeed spoken to many customers across, across the board, and it’s quite interesting because if you look at how LLMs work, and they were originally not designed for commerce they were just designed to aggregate all kinds of information, right? They primarily acquired knowledge about products and so on by crawling, right? So they just went to all these sites and pulled whatever they could. So you end up if you’re a brand looking at how the LLM is describing you, it’s a hodgepodge of information. Some of it could be good. You’re happy you got discovered, but you’re maybe not happy because of the way you’re being described or your products are being described. So that’s a big challenge, right? And for brands that really value all of the time and effort they put in building their brand, now all of the sudden they have this LLM trying to describe them. And then of course, when you talk about commerce, there’s a, a secondary issue, which is that commerce is about products, right? It’s about buying products, and the consumers want-- are looking to buy products, and brands wanna show them their products. So this whole area is the area of discoverability, right? Which is how do a brand’s products get found on LLMs. And so crawling clearly is not a very good way to do it. So the LLMs, as they got more into commerce, started coming up with more structured ways by which product content can be essentially published to the LLM. So you’ve got ACP, ECP, and a JSON-LD, a few different standards by which a brand can publish their content. Now that’s a great start, but then you run into the other problem, which is that the way consumers ask questions on LLMs is not that they go in there and just ask for a product. In fact, most consumers think of LLMs as advisors or friends, so they ask them questions like, “I’m going to my friend’s wedding, and it’s gonna be a hot weekend, and what should I wear?” And so the question is, how do you match that up to a product? And that’s a huge challenge, right? Because, in search it was easier. People know to-- knew to look for specific products. But here they’re describing needs, they’re describing occasions, and so on. That’s quite challenging. That’s where, it becomes quite important for brands not just to publish their catalogs into the LLMs, but actually to spend time enriching the content and describing their products in a way that match how consumers are a-asking for or expressing their needs.
[24:58] Anna Luo: Anything to add, Emily?
[25:01] Emily Pfeiffer: I completely agree. I think part of the problem with all of the, the scraping where the LLMs were saying “I want all this product data, I wanna be able to show it, so I’m going to crawl websites and scrape the product data off of them,” is that they never could see inventory and delivery information, right? To your point, Diaz, commerce is hard, and it requires operational data. It re- it requires promises that you can’t get by saying “I know this product was available on the website at the time that I scraped it.” Okay, how many? When will it arrive, right? So it was-- I agree, it was a really big miss, and starting to figure this out in a way that’s more intentional and planful is gonna be really important for digital commerce.
[25:42] Diaz Nesamoney: Sure. And I think the other interesting thing is that they-- and the LLMs are designed primarily for text, and so initially it was just descriptions of products, but nobody buys products just by reading about them, right? And even now they struggle with kind of, during the crawling, they’re not bringing up the rich experiences that are available on a brand’s website. And so it’s a subpar experience for consumers. But, consumers are there by the billions, literally the whole industry’s trying to figure out how to make that happen. Yeah.
[26:10] Anna Luo: Okay. My next question is to Diaz. Can you give us a few examples of challenges that brands are facing from different industries? What are some common threads you’re seeing and some differences that you can highlight for us?
[26:25] Diaz Nesamoney: Yeah, sure. First it’s great, and this is not-- that wasn’t true even six months ago, that brands are starting to recognize the importance of this very large consumer base on LLMs that are starting to look for their products. I do occasionally run into a few that are saying, “Oh, the numbers are small.” And I say “actually they’re not. The numbers you’re seeing on your website might be small,” and that again goes back to the fact that they’re not seeing you on these LLMs. But there are several other issues by industries. For example, some industries are highly regulated, right? The-there’s guardrails they need to have before their content can be published. Alcohol is a good example. We’ve had to build capabilities into our platform to make sure that the guardrails are there, like age gating and so on and so forth. There’s also, even right now as we speak even the structured ways by which product content can be published doesn’t cover service products, right? If you’re a financial services firm or if you don’t have physical goods to sell, oddly enough, you can’t publish that content. So if you’re a services business, like insurance, for example, how do you create visibility there? And so those industries are-- they want to be in front of consumers that are looking for loans or insurance products, but they’re trying to figure out how to get around some of these limitations. And there’s a lot of such challenges. So I think the mechanisms that Google and OpenAI came out with as a first cut, ACP used to be, were designed for basically commerce of products, right? They’re just replicating the e-commerce sites. But they weren’t designed for general discoverability of other services and other products. Now all these brands want to be there. They’re trying to figure it out, but the infrastructure is not quite there yet.
[28:10] Anna Luo: Yeah. Emily, you talk to a lot of your clients too, brand clients, retailers. How are they responding to the challenge, challenges that Diaz was just mentioning?
[28:20] Emily Pfeiffer: With sheer panic. Great way of putting it ... description. I would agree with that. Yeah, right? E- everyone feels like everybody else already has this figured out, and, I’m way behind. Everybody else knows what they’re doing, and I don’t. What’s a UCP, right? I have no idea what to do. The-- it’s-- I feel like I just, I wanna tell all of our watchers and listeners that it’s okay, and it’s early days, and everyone is still figuring this out, and you’re doing the right thing because you’re educating yourself and you’re learning more, and that’s all you can do. We- we’re gonna help you through it. But yeah, no, sheer panic is mostly what’s going on. And there are some folks who are experimenting and getting out ahead of it, but it’s, it is early days, and it’s okay not to have this perfectly figured out today.
[29:06] Anna Luo: Yeah, the people that we’re talking to, I think they’re all the trailblazers, so they tend to know quite a bit about what they want, the strategy, how to test. But there’s a whole bunch of other brands who are just, getting their feet wet and trying to figure out where they fit, whether they’re the brands that will, that can take better advantage of the LLMs. I think we’re gonna t- dive into a lot more of that in the next episode. But yeah, can you give us some sort of- highlights into, what’s to come for these brands.
[29:40] Emily Pfeiffer: Yeah. So things are changing in the way that shoppers consume information about products and services. I want to show you and talk you through what we’ve been used to previously and historically versus what’s coming now. So answer engines like Perplexity and ChatGPT and Google Gemini they are earning more of the product discovery and consideration processes in the shopping funnel, right? So if you think about the shopping funnel as awareness and then consideration and then conversion, where awareness is you have to know about the brand or the product before you even can think about buying it. Consideration is when you decide, do I really want it? I’m gonna learn more about it. I might figure out if the price works for me compared to other similar products. And then conversion is the purchase and your funnel narrows and folks fall out along the way, right? Traditionally, that’s the way we think about it. So previously, you had some awareness happening in search. Diaz was talking about more traditional search engines where consumers know, we all-- we’re all consumers. We all know what it means to put in a keyword or a key term and expect that we’re gonna get results that matched to the thing that we typed in, right? So some of that awareness was happening in search, and then we clicked a link, and we went to a website, and all of the consideration happened there. If it’s a completely digital journey, all of that consideration was there, and the conversion happened right on the site. But as things are changing, we’re seeing a shift. It’s now at least possible that a whole lot, maybe all of the awareness, is happening in answer engines, and all of the consideration can happen there as well, right? So in an answer engine, you aren’t just putting in a key term, you’re putting in a phrase. You’re having a conversation. You’re saying, like Diaz w- was saying, “I’m going to an event. The weather’s gonna be really warm. What should I wear?” You might get questions back or prompts or suggestions for the next thing that you might type in “How formal is this event?” What do you expect the weather might be? Is it outdoors or indoors, right? And as you continue that conversation, you’re going through more of that consideration process. You might say, “Ooh, I like this, but it’s too expensive. I like this, but does it come in another color? Show me more things like this.” And all of this consideration now is pulled out of the control of the owned environment we were talking about before, right? Most of the conversion is still today happening, almost all of it, is still happening on websites, but not all, and we might see that shift more as well. So when I think about the funnel in digital now, I’m just thinking about the upper parts of the funnel shifting outside of the spheres of control that merchants have.
[32:29] Anna Luo: Yeah. The control is, one of the things that Diaz talks about very often is how do you get back the control from these LLMs? Diaz, do you wanna just give a highlight on the experience from discovery to experience?
[32:44] Diaz Nesamoney: Yeah. Just just to add what Emily to add... Emily was saying, I think the obvious brands that are saying, “Oh, we need to be there,” is the high-consideration products that require a little bit more research and thought. Beauty products, which, is, y- what you wanna-- the beauty brands wanna be on the LLMs ’cause the text descriptions aren’t quite doing justice to them. But I think I just wanna pick up on something you said, Emily, is the difference is conversations, right? So I think eventually, so whether you’re up in the funnel or way at the bottom, even very low-consideration products, even think of an experience I had is I go to Amazon, I try to find a phone charger. I get 1,000 of them, right? And okay, which one? What... and it’s so hard. It sounds easy when you have a lot of choice. It’s actually hard ’cause you have to pick the right one, and you pick the one that’ll work with your phone, and you’re reading through all the descriptions. Really hard. I think of that same thing on the LLM where you say, “I have an iPhone 18. Can you find me a charger?” And the LLM’s doing all the work. So I do believe that conversational experiences are gonna push all of the funnel up all the way down into the LLMs. The urgency right now is more on high-consideration products and beauty products and so on, certain categories where they kinda have to. But, I do feel like it’s much broader than that. It’s conversational versus non-conversational versus where in the funnel your product sits.
[34:13] Anna Luo: Good discussion. Okay, I’m gonna move on to the next topic, and let’s go back to an earlier Accenture research that Chip shared. Let’s see. It’s the relative industry positioning on estimated net new demand and execution risk from AI agents. Here we go. The Accenture model finds that beauty and personal care, travel, and health and wellness are seeing the greatest acceleration in net new demand from AI agents. And that makes sense ’cause these categories represent frequent and complicated decisions that consumers need to make, and AI can make decisions easier for them to buy and often leading to incremental new spending. But at the same time, Accenture simulation finds that up to 60-- 86% of AI agent-mediated transactions are at risk of being abandoned or switched to a competitor when something goes wrong in the purchase pro- process. And the industries that are likely to be impacted the most, according to this simulation, are health and wellness, travel, and groceries. I’m gonna pause here. Based on what you’re seeing, what are your thoughts? Diaz, would you like to take that first?
[35:36] Diaz Nesamoney: Yeah. So a-a-a-again, just as I was saying be-before there are some obvious industries like the ones you described that’ll probably be impacted the most just because consumers have decided they just wanna use the LLMs for that. They’re not-- They don’t wanna, Or the current way of servicing or the experience is so bad that they have found a better way and they’re gonna move quickly. But I think ultimately the conversational model is what is more natural to consumers. And so I think regardless of what product it is or what category it is consumers will prefer to engage with them on LLMs. Right-- Even right now as we speak, I’m speaking with several banks, and these are very traditional banks. This is the last thing I thought would be a-- bank would be an early adopter of of building a presence in LLMs. But, their products are quite nuanced too. If you’re trying to apply for a loan, there’s like 20 different loan products with different, nuances to it. So I actually think it is, again, the conversational model. Now, it may be more the early adopters and the risk takers and the brands that sort of jump, jump in right away and the others might follow. But, and obviously the high propensity industries where they feel the pain a lot more and may-more immediately. But I do think after the early adopters go in there, jump in and start seeing the benefit. And market share is a big thing, right? LLMs have a lot of Gen Z a higher proportion of Gen Z. So I talk to brands where they’re like, “Oh, we want new to brand. We want new to brand.” It’s like what everybody wants. And I say “guess where new to brand is? It’s on the LLMs.” So if you can figure out how to get there and engage them, you’re gonna gain market share very quickly, right? So I think it’s more the brands and companies that are traditionally more risk-taking and willing to ta-make the bets, even though it is early. I know you said it’s very early, and I agree, and there’s a lot of unknowns and experiments and stuff that are happening. But I think the wisdom is for brands to get in there and be part of the experimentation soon, because those brands will ultimately be the leaders in their industry ’cause they jumped in quickly and learned.
[37:38] Anna Luo: Yeah. Speaking of Gen Zs, we actually have the next episode I think Emily’s gonna cover a lot more about those shopping behaviors and how they compare to the demographics that are using AI-powered selling. Let’s see. If there are three things brands can take away from this conversation, what would you say that they are? Emily?
[38:02] Emily Pfeiffer: I think the first is don’t panic. Just don’t panic. Everything’s going to be okay. It’s a great time to experiment. It’s a great time to learn and figure out where you’d like to start. Number two is to remember that there are some things you can still control, right? Diaz you’re talking about the conversational expectation that consumers more and more will expect to be able to just interact that way. That’s true in the owned environment too, right? So I think that it won’t take long before we all get accustomed to being able to use natural language, not just a search term, but a, a longer string, a query, right? And we’re gonna do it on websites too, and we are going to expect that they can handle it. So keeping on top of your owned environment and making sure that it too is evolving as all this changes, I think that’s really important. Number three is to experiment. This is a time to be wary of the investment. It’s new and it’s unproven, but I wouldn’t entirely wait. I would start gently experimenting as we move into this kind of new way that folks are going to be shopping.
[39:11] Anna Luo: Great. Thanks. And Diaz?
[39:14] Diaz Nesamoney: Yeah. I think one of the things that I think is really important to recognize is that it’s, LLMs are not just another channel. I think sometimes there’s a tendency that, oh, we have a new channel. We’ll get there eventually. But it’s really a new way of consumers engaging with brands to buy. And I think brands have to start thinking about-- And I agree that they shouldn’t panic necessarily, but at the same time, they shouldn’t sit on the sidelines hoping that they can jump in when it’s mature, because I think the market share issue is a real issue if you don’t figure out how to play in this new conversational model, whether it’s on your website. And I agree, Emily, it’s, doesn’t matter. It’s all surfaces. Consumers want that experience. They love it, and they wanna see it everywhere. So get used to the conversational model. Figure out how to get your content ready for conversations discoverability, all of that, and I think you’ll win. I- if you sit on the sidelines, you may find you’re losing market share and may be too late by the time you try to do something. Experimentation is the phase we’re in. You have to learn quickly, learn fast, and adapt to this new model as quickly as possible.
[40:49] Anna Luo: I hope you enjoyed today’s episode. I’m always interested in hearing topics that are top of mind for you, so please drop me a comment with your suggestions. Until next time, innovate responsibly.