Building Trust in the Age of Agentic AI: Sucharita Kodali on the Future of Retail Intelligence

Genai, I think, has empowered individual creators more than anyone. I think that to your point, trust is still an obstacle for large enterprises, but what that means is that these large enterprises end up facing that innovator's dilemma. And, you know, they may very well be disrupted by some small creators that are able to do what they do for cheaper, faster, and better than what large enterprises may, just by design, take longer to execute. Welcome to Bitex Live, the podcast where digital transformation gets a reality check. I am Santiago Naranjo from Bittex, and this is where we break the rules. We write the playbook and hear from leaders who are actually getting results. Let's get most respected voices in retail. But I would love to start by getting to know you a little bit more as a person who is this professional behind the VP of a principal analyst. Thank you for the kind words. I've been an analyst now for almost 20 years, and before that I was in the retail world as a practitioner, so worked at different kinds of retailers, big box retailers, small retailers, brand manufacturers, startups, all in retail. And so that is essentially how I got to be a retail analyst. And I've always been interested in business, even from a young age, so I guess that's what led me there. Retail is so particular because you are able to face challenges in logistics, brand financial, technology. I am also from retail, like more than 10 years, and I love the people that are around retail because you have the majority of the problems and less of the resources and then you need to have and pursue like creativity. Why you take the challenge to move from retail to analyst, how that arrived to you? Why do you decide to make this change? Well, I think that I have the easier job. It's always easier to tell people what they could be doing or should be doing. It's a lot easier to, you know, kind of be calling the shots from the sidelines than it is to be the people who are actually executing. I think that someday in the future maybe I'll go back to it. It suits me very well. It's a really flexible job and I get to learn every day. So that's, that's what I like about the work. The downside is I just tell people my thoughts and share research. You don't actually get your hands dirty and do the work that lead to change. But, you know, I think you became one of the more respected analysts on top of everything that you do, but also because you have the expertise and when you talk, when you share your knowledge, it's not just theoretical. You were There you were in retail, you know what you are talking. So it's good. I embrace all the analysts that please hire people from the industry that they know what they are talking about. So we can not start to talk about AI. You know, artificial intelligence arrived in any topic you will see artificial intelligence. I love it because it's like when our Internet arrived. It's just so foundational that everybody's talking about, but nobody knows how to digest this. From personal use to security, privacy, creation of NCP, new companies, SaaS, etc. Everything is happening and sadly sometimes the people that talk the most and then saying what should be done. So this is why I'm trying to be here with a respected person that have all these details. So I want to know if you have at this moment any big cases on AI, any big uses that start to get excited, to get excited about the use of AI in retail or any specific case that you are starting to see that look that could be a potential for AI on retail. So I think that AI has been in retail for decades. AI is not new and it has been used for many, many years. And most large retailers already have retail planning systems or pricing algorithms or labor planning modul or warehouse management optimization solutions which all rely on machine learning and in some cases even deep learning to execute what they execute. One of probably the oldest use cases even within neural networks is fraud detection. So I think that there are already many, many proven use cases of AI in retail. Are there new solutions? And I think that maybe the heart of your question is is there anything that is related to these new terms like generative AI or agentic AI which are being proven? And with generative AI there are some applications that are relatively mainstream now, whether it's content creation or workforce automation. I mean you see companies in retail, they may take a vendor data sheet and be able to use generative to populate a content management system for an E commerce platform. Like that's a very common use case. Chatbots now seemingly are ubiquitous and smarter generative AI driven chatbots. So there's definitely some use cases. Are they transformational? I think that they're incremental and I think that they're valuable, but I don't think that they single handedly are going to turn anybody's business around. And as far as AgentIQ, AgentIQ is still new and there are a lot of issues with being able to. I mean when you look at the definition of agentic, it has a very specific definition. It's a goal based approach where a system can act autonomously across many different touch points. And the example I always use is a driverless car is an agentic AI solution. What is the equivalent in retail? There may not be a lot that have been built out. Amazon's merchant solution is probably the closest to a truly agentic AI solution where they have a goal for selling a certain amount of inventory and then you know, kind of the system makes decisions on its own. But it's okay if we're not there quite at agentic in everything because there's a lot of value even with automation. And I think that we're going to be taking steps probably over the next five to 10 years to get closer to more agentic solutions. But even as we are able to automate, we'll see lots of gains and you know, the companies that do embrace it should hopefully see improvements in their processes. So Santhi, that was a very long winded answer to your question, but it's very powerful. Why I love it because I think we are not just in the peak of the wave of AI or generative AI, but we are starting to see incremental improvements on some ways that the technology start to use data to structure or generate some type of operations. Something I personally I am a builder by heart so my passion is to build open source on my weekends and we build together the for example, the support of Bitix. We used to have like 300 people across the world and now we were able to centralize in a chatbot that have some capability. We are not the one I think majority of the players in SaaS are doing the same and we have good things and also some trade offs. The good things always is to be able to centralize, to have clear data points, to act to improve the product or to improve the documentations. Trade off is always the companies or the brands want to speak with a human. Even though sometimes technology can be more precise technically what they should do, what they should not do. The buyers procurement want to have at least human interaction. What is your point of view on what will be the moment that or is going to be the moment, yes or no? I'm not clear about that the branch will accept a total agentic software, that they will accept an ERP that they will not talk with an account manager or E commerce solution or a PIM that will exist or not or we are going to need to have this agentic solution for maybe one, two decades. Then we are going to have a generational transformation that can adapt on this new way to operate. Because this question is a mix of generational change and technology. How do you see this process? That's a great question. And I think that you call out the single biggest obstacle, which is trust and being able to trust the system and to trust that the system isn't going to create a problem for your company or drive your company out of business. So I do think that where you'll start to see more of the trust is in some of the lower risk scenarios. So if there's the ability, marketing is an area that's relatively lower risk than some of the other parts of retail. And kind of if it's like you have a daily budget and you want to let the agent decide how it's going to allocate that budget across these four specific channel partners, that's a relatively low risk thing. So I think that we'll see it there. I think that in terms of buying inventory, it may be a little bit different. And even the Amazon example, Amazon has an agentic merchandising solution which is very different than what most retailers can do. It is a marketplace and kind of essentially all they're doing is the agent basically prices things for an endless aisle. Most retailers don't have that luxury of purchasing everything right. An agentic solution for most traditional retailers would involve a trade off of some sort. And it involves a trade off of what are you not going to buy in the limited shelf space that you have. That's actually a much tougher decision than what Amazon has to make. So where there's less risk, I think we'll see it happening. Where there's more risk, I think that we are going to just require more information and more data points. I mean, and that in fact is one of the things that holds, I think, any agentix solution. Because right now I think the agentix solutions that have been built have been shown they've gotten better and better, but for the most part they can take discrete one out attack that normally a human would take about an hour to do. In software, AgentIQ is pretty good at doing. If you extend beyond an hour, usually there are different systems and different decision trees and it makes it much, much harder. There's a reason that no agent yet is able to build you a house. There's a lot that is required in that, that we're not there yet. We may be there in, you know, 20 years, maybe even sooner, maybe, maybe not. It's going to take time, it's going to take data, it's going to take reinforcement where you have enough of a data set that, you know, the agent can learn if this was a good decision in this scenario or a bad decision in that scenario. And if you don't have enough of that data, we may never have an agentix solution. You know, I always use the example of I'm sure most of the listeners have probably come through an airport in the United States at some point where we have our TSA and you know, they have entirely agentic solutions. You could theoretically walk through an airport and image recognition will know exactly where everyone is and who they are. But yet we still have to go through a human being looking at everything and taking your ID and you have to go through a security, you know, kind of apparatus. There you go, like, you know, as a consumer, as a passenger, like I'm there, I'm happy. I would love for it to be a fully agentic solution. They are not ready for it to be a fully agentic solution. And you know, that's an example of like the technology is there but the trust in the system isn't yet there. I love that I'm fully identified with your point. Something that I want to change a little bit the years but something that I'm seeing is that everybody's trying to demonstrate how good the agents are but nobody is telling us in a very like normal or trustful way in what the agents are not good. Because it's like everybody is overselling the agents. It's like crypto. Exactly. And so I was building an A to a like protocol and solutions. Majority of the questions from the clients are like okay, but what if the agent have a problem or a mistake? How we are going to know and how we are going to realize that it's not too late. A challenge I'm trying to bring for developers or engineers is what if we stop overselling the agent e commerce or the agentic area and we launch some mission led agents that sometimes they are going to achieve the mission but sometimes they are not going to achieve the mission. If you are able to bring trust for a senior like procurement CMO or CFO to tell you, okay, you are going to have here 10 agents, some of them are going to achieve the mission and some of them they will not and they will let you know. I think trust will be on top because only to have, don't worry about it, your reconciliation on finance, don't worry, some agent will do it for you. That's just scary and it's not true. But maybe we can have three, four, six agents that say look for this specific task. We will try with these four agents to give this specific mission. If they complete it they will tell you if not they will be the first one to give you that they fail. So I'm trying to go this and I will the overhype of agents and why if we go for a mission led agent that it is okay to have 50 50, 50 sometimes they achieve 54 times they not because we will not lose the trust of the AI. Everybody wants that. Okay, let's move for another topic. I want to see your personal opinion and your personal use of AI. How you use in your daily operation, in your work, in your house AI. Are you using Siri, I don't know Alexa? Are you using ChatGPT which are an top analyst AI stack? That's always scary. So actually we are actually not allowed to use a lot of AI at work and that's actually something that I do think may be holding back. Some of the generative AI solutions is now so many companies have put restrictions around it that people can only use it on other alternative devices or they can use it for their personal needs and whatnot. I'll tell you, I feel like my children use it more. They seem to use it for anything and everything. So I like to use from I don't use Siri and Alexa all that much. I think I've been, you know, kind of habituated to just be cautious about where where you share a lot of super personal information. And even though you know, I know a lot of people and you know, people even in my family who have paid for subscriptions to a lot of the generative AI tools, I'm still paranoid about the data that gets into those tools. So I will still always use them in either a logged off form or use somebody else's login and passw for that. I like using ChatGPT for general questions, perplexity anytime I need like additional sources that Google may not have unearthed. I like Claude for any type of writing support or creative thinking of of how to phrase things or idea generation. So those from a gen AI standpoint are there but there are a lot of other point solutions like you know, there's a social media generator called Opus Clip which I like a lot that kind of you can take in a video, it will you know, spit out like you know, 25 video shorts that you can pick and choose for like Instagram or TikTok. So they're one one of my favorites of all time has been Descript, which is actually a podcast tool. And what's really interesting is you know, you want to talk about hype the founder of Descript was the guy who started Groupon and, you know, kind of. And I feel like this is actually like, you know, the more promising one. But Descript is awesome. For years it had the ability to transcribe content and it had basically deep faking technology to be able to like edit. If you misspoke something at this point in a podcast, it could literally, you know, kind of just take your voice and put in what you may have intended. Of course there's always risks with having that misappropriated, but, you know, if it's you doing it for your yourself and your own recording, I think that that's perfectly fine. And it's essentially like a very, very easy editing tool. So it uses a lot of natural langu and different forms of, you know, kind of automation to make things as easy as it could possibly be. So, you know, kind of somebody like me, who's not a podcast editor can actually edit piece of video content. So yeah, I mean, those are some of the tools that I've used. I think that what it has really done is Genai, I think has empowered individual creators more than anyone. I think that to your point, trust is still an obstacle for a large enterprises, but what that means is that these large enterprises end up facing that innovator's dilemma and, you know, they may very well be disrupted by some small creators that are able to do what they do for cheaper, faster and better than what large enterprises may just by design, take longer to execute. I love that and running. I'm very curious about that solution. I promise you I will try it. Yeah, definitely. Thank you so much for the advice. I will try all of them. I love wrapping up this episode by reminding everyone that this is part of our special series where we sit down with women who are rewriting the rules of business, technology and digital commerce. Sharita to conclude, what practical advice would you give to women leaders and entrepreneurs to boost their careers and businesses in this new area of commerce? So we talked a lot about trust and I think that in the era of mistrust, trust still is going to matter more than ever. And what that means is that you have to have genuine relationships with people and you still need to have relationships and you need to network and you need to actually meet humans like, you know, kind of in my network, you know, I don't respond to random strangers who kind of either I don't know them or they don't know me, but they just reach out to me randomly. There still needs to be an intermediary I will have needed to have met them directly or there needs to be an intermediary who I know and I trust that can broker that introduction. And I don't think that goes away in the next decade. In fact, I think it becomes more important because, you know, when there are no barriers to connecting with somebody, how in the limited 24 hours a day and seven days a week that you have are you going to choose where to spend that time? And it still is going to have to come through somebody vetting the opportunities that you have in front of you. What an amazing and powerful message because it's very easy for people to try to go in the opposite and obvious direction to okay, now I can do it alone, I can go going to my room. I don't need to engage with anybody. I don't need to create a network. And as you say, as in this moment, trust is even more important than technology because technology is there will be a commodity. But to be able to create a network and to grow is very important. It is a very, very powerful message for everybody that is out there. Thank you so much for sharing your professional insight, also your personal wisdom. Thank you so much for being here. This episode of BitX Live was brought to you by Bitex, the digital commerce platform for bold leaders. Head to bitex.com to see how we help businesses scale fast, flex smart and let the AI power future of commerce. If you like it, what you hear, follow the show on Apple, Spotify or wherever you listen. Because in this space, speed isn't a luxury, it is your edge.