Author transcript
Sean Nolan: Hi Saurabh, thank you so much for joining me on CX Today. To start things off, can you introduce yourself and tell us about your role as Director of Product Management, Agentic Commerce and Agentic AI?
Saurabh Pitkar: Yes, absolutely. Thanks for having me, Sean. My name is Saurabh Pitkar. I'm a Director of Product Management for Agentic Commerce and Agentic AI at Dell. In my role, I'm primarily responsible for all third party as well as external commerce integrations for Dell to sell Dell products on channels like LLMs, marketplaces, or social apps, or any new channels that you're seeing in recent times, like Meta's Muse AI or Instinct. My primary responsibility is to get the end to end product strategy and architecture delivered and figured out on those channels, so that Dell keeps on growing its revenue.
Prior to this, I worked on building a unified commerce platform that powers Dell's underlying transaction ecosystem across all global transactions, worth $150 billion plus. I have experience across supply chain, ecommerce, personalization, and product management. By education, I'm an MBA and an engineer.
Sean Nolan: Thank you so much. I wanted to start by looking at agentic commerce as a whole, and from two sides, because it affects consumers and it also affects businesses. To start off, from the consumer perspective, where do you see agentic commerce providing the most value as they go through their retail journey, their customer journey?
Saurabh Pitkar: I think for consumers it's become a very helpful tool overall. But primarily, the key benefits are coming from two main areas.
One is product discovery, which used to take a while. When we are interested in buying, let's say, a new DSLR camera for my hobby, I'm very much into going through ten or twenty different tabs, looking at different products, websites, and reviews, to finally make a judgment call and determine, "This is the one that I want to go with." That used to take hours of my time. But with the AI agent at my hand, literally those hours have been shrunk into minutes, and now minutes are also shrinking into seconds. AI agents can fire 20 different parallel searches, and they already have a great algorithm to map my intent to recommendations or specifications. So they can do things much faster. That's the power we can see with agentic commerce, as the first key benefit for consumers.
The second key benefit is where the intent is pretty clear and the products are pretty well understood. It's more like a repeat buy. Consider groceries or consumables, or seasonal things that we buy every season. For those, it is now much easier for me to give a specific instruction to an AI agent to go buy on my behalf. And I can trust the AI agent enough, as a consumer, to have it buy just the same product that I want, and not go anywhere else.
So the first benefit is more on the low clarity intent, where I have an intent but it's not yet super clear what I would prioritize. The second is the high clarity, repeat buy intent, where it's a chore for me to go through every single week or periodically and do the same thing again and again, which I can now see as pretty automated. Those are the two key benefits, in my opinion.
Sean Nolan: Thank you so much. I think that really demonstrates the convenience of agentic commerce, which is why people with their personal AI agents are increasingly looking to make their purchases through an AI agent. Now, to switch over to the business side and the retailer side of things, what foundations do you feel these organizations need to put in place to ensure they're not left behind in this new era of agentic commerce?
Saurabh Pitkar: I would say that these are going to be my personal views. From the business side, there will be a few different areas, or investment bets, as a business.
The first one is that businesses would need to ensure they have a connected supply chain, and have a supply chain flywheel effect going on. What I mean by that is, have the products that you're selling in a structured manner as product data. Then have your live inventory information from your distributors, factories, and warehouses connected in real time to that product data. Then have your demand shaping or pricing teams work in real time on top of that product feed, so that you have influence over how much profit margin you can get through these external channels, instead of asking users to visit your website, because that's the new reality. Having these three investments gives you that flywheel effect. That's investment bet number one.
The second bet, in my opinion, is for businesses to start thinking of exposing their commerce capabilities in a fully headless manner. Typically, you would have website traffic, humans coming to your website, and you build beautiful experiences for humans to engage with, and sometimes also have fun going through the website journey. But that's not true anymore. With AI agents replacing humans as users of your business or your website, we need to think more headless from a commerce perspective, so having headless capabilities exposed.
What I mean by headless is, let's say you have an add to cart operation that you offer as a call to action button on the UX. Have it exposed as a headless capability instead of a UX capability, so that the AI agent doesn't have to figure out, "This is the button I need to physically click while visiting the web browser." It becomes a way more efficient integration. Secondly, headless capabilities also help you scale your architecture for any such agents if you keep following the industry protocols. I know there are several protocols out there, like MCP, the Agentic Commerce Protocol from OpenAI, or Google's Universal Commerce Protocol, and so on. If you follow the protocols and make your commerce capabilities machine readable and headless, I think that would benefit.
The third one, once you have this setup, is how you monitor across omnichannel. Earlier, you had traffic coming in, so you would have a deep understanding of the user, their intent, the journey they took, and the time they spent on your website. But now you don't get those metrics anymore. So how you go for proxy metrics, or indirectly conclude what intent the user or the user's AI agent had, I think that would be key.
The fourth one is more from a B2B perspective, not for consumer businesses. In B2B, identity and guardrails will be a very important aspect. What I mean is, when B2B businesses have an agent to agent interaction, establishing the seller agent's and the buyer agent's identity is critical. Establishing what type of guardrails the B2B transactions need to operate on is also critical. You can imagine things like a negotiation process happening between a seller agent and a buyer agent for, let's say, a set of computers that the buyer is interested in buying, and the seller agent may go above the threshold for allowed discounts, which means the agent did not follow the guardrail. So how do we make sure there is a secure guardrail established around such agentic communication protocols? That would be critical.
So the supply chain flywheel, the fully headless commerce capabilities, omnichannel monitoring, and having your guardrails and identity established in a secure fashion: those are the four investment areas for businesses, I would say.
Sean Nolan: Thank you so much. I think that's really interesting. Touching on your last point, you can't have your sales agent making extreme promises to customers that you can't then fulfill. So having those guardrails is extremely important to having compliant sales journeys. I wanted to ask about the world of marketing. How do you expect marketing to change as AI becomes these shopping agents and leads the shopping journey? For businesses, will they still need to appeal to humans as much as they do to the AI agents that are doing the shopping? What's your perspective on that?
Saurabh Pitkar: That's an interesting question and something I have been thinking about, but again, it's an evolving space, so we don't have all the answers. Here's what I think.
In the near and medium term, we will have both humans and AI agents to serve from the business perspective. That means marketing dollars will have to be redistributed to understand how humans are behaving and how AI agents are behaving in this new world. We should expect that they are going to coexist.
What I mean by that is, from a persona side, if humans are visiting the website or trying to access your product catalog, humans are emotional beings. We sometimes make irrational decisions. We want to be influenced by looking at beautiful websites. Sometimes we see nudges on ecommerce websites like, "This product, only two items left in stock. Buy soon." As a human user, I would be very much influenced by such nudges.
But if you compare that with AI agents, AI agents are pretty much looking at facts and consistency. What I mean by consistency is, let's say there are APIs that AI agents are accessing, and certain transactions that AI agents are doing. They already know the trends, and they already know how businesses perform compared to other businesses in the same category. So AI agents rely more on such consistent information. And the fact is more about which product you have, and what the stock is. Is it one, or five, or ten? They don't care about nudges like, "This item is extremely low in stock," or "This item's price has fallen by 30%." AI agents pretty much don't get influenced by things like that.
So marketing has to figure out how to distribute money across these two personas and the traits that each brings. There will obviously be a distribution of funds across the channels where you are getting more traction.
To some extent, I would also say that first principles still continue to apply: your marketing spend should reflect where the performance is coming from, performance in terms of revenue, conversion, and so on. All the traditional metrics are still applicable down the funnel, when the transactions are happening. But up the funnel, the metrics around where the users are coming from and what intent they have for buying your products have gone out the window, because AI agents are typically not disclosing that information. But we have other ways of indirectly influencing those decisions via marketing dollars.
Sean Nolan: Thank you so much. I think that's really interesting. Your point about how you label different products, such as saying the price of this product has fallen by a certain percent, these are traditional marketing strategies and tactics that have been going on for decades. They will have to be rethought, because agents don't respond to those the same way as humans. It's a really fascinating new area. As you say, we're still exploring and figuring it out, but there's a big shift coming, for sure.
Finally, I wanted to ask a forward looking question. As we head into 2027, coming very soon, do you have any predictions about opportunities or challenges in the next year when it comes to agentic commerce?
Saurabh Pitkar: I think there will be several, for sure. This is an evolving field, rapidly evolving rather. I would really want the industry to converge on protocols at this point. That would be my first opportunity for the overall industry. We are at a point where we have ACP, the OpenAI Agentic Commerce Protocol, and Google's Universal Commerce Protocol. Meta and Sierra launched recently their personal agent protocol. Google already had AP2, and there are several companies, like Visa, that launched their protocol for agentic transactions and payments. So it's becoming a set of protocols. If you look back in history, the internet has been built on several protocols. It's not one single protocol. However, I would want some convergence for businesses to decide the right type of investment, and which protocols we are finally using together to engage in a partnership and a transaction securely. That would be the biggest opportunity for us as an industry.
The second one is that so far the discussions have been on the consumer. I see commercial as a very big opportunity going into the next year, primarily because the market size itself is three to five times that of consumer, at least. We are talking about several types of sales motions that could be revolutionized and see the efficiency in agentic commerce. Those are the opportunities, in my opinion.
You are asking about challenges as well. I think the challenges are going to be more toward businesses. Of course, challenges do exist on the consumer side as well, on how you build the trust to give an AI agent your card, your money, as well as your private data, like your email access. That issue aside, from a business lens, the challenges are going to be more around fraud. It's about understanding, "Is this a genuine transaction or is this a fraud transaction?" Such transactions can happen at scale now. You can fire up 20 different AI agents to do such transactions once you figure out a loophole. So that risk exponentially increases. Businesses will have to figure out how to critically monitor such channels if they expose their APIs and capabilities to them.
Secondly, it's about liability. When such agentic transactions happen, who is liable for ownership, any damages, any returns? Such things will be critical.
I think there are also going to be interesting use cases where speech gets involved. Let's say an AI agent can speak on your behalf, and you might send it to call a business's customer support team and say, "I want to cancel my subscription." That agent will basically stay on the call for 30 or 40 minutes at once, with no complaints. But it's not sustainable for us to stay on a call like that. So businesses may see a traffic spike because of it. Similarly, if I consider search, businesses may see agentic traffic doing rapid fire searches on their websites. Those are going to be some of the challenges in how businesses will need to respond to the agentic commerce phenomenon.
It goes back to my initial point on the investment opportunities around monitoring. If you're carefully monitoring where the traffic is coming from and how the transactions are happening, and your anomaly detection algorithms are super tight, you would have a very good handle on such behavioral patterns. So as a business, that's what I see.
Sean Nolan: Thank you so much. I think the points you made about liability, fraud, and trust are super critical questions that businesses need to be thinking about as we head into the new year. You shared some really insightful advice for business leaders today. Thank you for your time. It's been great speaking with you.
