S6E16 Diptendu Ray - AI Automation

[00:00:00] Teaser

[00:00:01] Casey Golden: ​ Interview 15 people about a process, and you're most likely to get 15 different versions. Giant Tiger's Diptendu Ray says AI automation starts with finding what they have in common

[00:00:14] Ricardo Belmar: And if AI saves your team a few hours a week, Diptendu says you won't see it on the bottom line, not unless you put those hours to work on something better

[00:00:22] Casey Golden: This episode of the Retail Razor Show recorded live from RetailClub AI Festival, Diptendu shares the side of AI automation that nobody's talking about. So stick around.

[00:00:34] Show Intro

[00:00:46] Ricardo Belmar: Welcome back to season six of the Retail Razor Show, the top ranked indie business management and marketing podcast on Goodpods, and the original show in the number one indie podcast network for retail. I'm Ricardo Belmar

[00:00:57] Casey Golden: And I'm Casey Golden. Welcome Retail [00:01:00] Razor fans to retail's favorite podcast where we cut through the clutter to give you sharp insights on what's happening in retail today and tomorrow. We get real about what's driving the future of commerce

[00:01:11] Ricardo Belmar: This is season six, episode 16, and it comes to you recorded live and in person from the RetailClub AI Festival in Huntington Beach, where we were joined by Diptendu Ray, head of product management and head of AI at Giant Tiger, the Canadian discount retailer with more than 260 stores.

[00:01:28] Casey Golden: Diptendu has been working with AI for more than 12 years, from personalization and machine learning to automation and conversational AI at companies like Yahoo, Humana, and Staples. And Rethink Retail named him one of its Top AI leaders for 2027, just like you, Ricardo

[00:01:49] Ricardo Belmar: That's right. And earlier that same day we recorded this, Dikshantu joined me on the PCH stage at RetailClub for a session called You Can't Automate What You Don't Understand. So we picked up [00:02:00] right where we left off with the hidden side of AI automation projects that nobody talks about upfront

[00:02:05] Casey Golden: Starting with the process itself, the version that's written down usually isn't the version your teams actually follow. Diptendu explains why every stakeholder builds their own version to hit their own KPIs and how to find the common ground when 15 people give you 15 different answers. Honestly, that one just gave me déjà vu

[00:02:27] Ricardo Belmar: Yep. Then we dig into data, your favorite topic. Diptendu breaks down which retail data is usually clean, which data is the real mess, and why a year-long cleanup project just for the sake of clean data is never really the answer

[00:02:41] Casey Golden: And we close on efficiency, which doesn't have to mean cutting costs. Diptendu's near term goal at Giant Tiger is moving inventory faster, and it has nothing to do with headcount. Plus why the hours AI saves your team won't show up in the bottom line unless you're[00:03:00] putting them back to work. That one's the mic drop

[00:03:03] Ricardo Belmar: It sure is. But before we dive in, I'd like to tell everyone about where the Retail Razor is showing up next, and that's Retail Agent Con. Next Tuesday, November 17th, from 11:00 AM to 3:00 PM Eastern, join me and more than 700 of your retail peers at Retail Agent Con, a live virtual event from our good friends at IHL Group with ClearSight AI. an honest look at what AI agents are really doing today across merchandising, order management, inventory, and store operations, which wins are real, and what has to change in your org chart to make them stick. You'll hear from retail leaders like Glenn Allison of Tractor Supply, Scot Sager of David's Bridal, and Justin Emig of Artisant Lane Furniture, plus Dr. Janet Sherlock on getting agents and people working together, IHL's Greg Buzek, Jerry Sheldon, and me as well. Registration is free for retailers at retailagentcon.com. That's four hours of real-world insight you can use heading into NRF's big show. Register today while [00:04:00]there's still time. Thank you to Retail Agent Con for making this episode of the Retail Razor Show possible.

[00:04:05] Casey Golden: Now, let's welcome Diptendu Ray of Giant Tiger to the Retail Razor Show.

[00:04:15] AI Automation's Hidden Traps (Giant Tiger)

[00:04:15] Ricardo Belmar: And we're back in the Retail Razor tent. This time we are talking with Diptendu Ray from Giant Tiger. And Diptendu and I had a really great session this morning over on the, the PCH stage from talking about, efficiencies and AI automation products.

[00:04:30] And I, I kind of think of it as, you know, in, in these kinds of projects, what are some of the hidden things that, as we said in the session, that people don't necessarily talk about, when, when they launch into an automation project with AI.

[00:04:41] So I thought we'd kick it off with one particular area. You know, one of the themes we had this morning is that when you start out on an AI project trying to automate an existing process, obviously one of the first things is you have to know what the process is, right? And, I think you had also brought up in this session, you oftentimes will discover that what's been documented as the [00:05:00] process isn't necessarily the process that teams are following.

[00:05:04]

[00:05:04] Your Documented Process Is Probably Fiction

[00:05:04] Ricardo Belmar: So I, I wonder if you could like speak a little bit to, you know, what do you do when that happens, right? Like how... Do, do you think this is a common, occurrence? And then how do you go about tackling that, from a, a, as a, challenge? How do you overcome the fact that you've got all these different people now that you've brought into the project, but the process we thought was the way it worked isn't the way it works?

[00:05:25] Diptendu Ray: Yeah. I mean, I think, we, we already mentioned that, you know, the, the, the process is sometimes not the process we imagine in our, in our, in our, brains. and I think the, the, the challenge is, you know, there are a lot of stakeholders to every process in, in an enterprise, and, for them, it has to align with certain KPIs or certain metrics or certain North Star of their own.

[00:05:46] Mm-hmm. and that's why I believe people kind of build their own versions of the process- Mm. -because, they'll be like, "Okay, I, I need to optimize for margins. This is how it should work." "I need to optimize-" Right. "-for inventory. This is how it should work." Right. "I need to in... I need to optimize [00:06:00] for labor at the stores.

[00:06:02] This is how it should work." And that's what creates all these different variations of the process, I believe. So I think first step is definitely to kind of, hear it from the horse's mouth, like I said, in the, in the, in that- Yeah ... our-our session as well. So include the right, the actual people who are kind of part of the process, and then, see what, what they are saying, and their version of the process could also differ.

[00:06:24] Like you could, you could in-interview fifteen people, and their versions of the process could have like fifteen different connotations to it. Right. but I think it's im-- it's, much more efficient to draw those, common denominators out of those fifteen versions rather than, relying on, a stakeholder of the process who has a, has a different version, completely different version

[00:06:43] Casey Golden: just gives me, like, heart palpitations. 15 versions of a process.

[00:06:48] Diptendu Ray: yeah I mean-

[00:06:49] Ricardo Belmar: Yeah.

[00:06:50] Diptendu Ray: it's, yeah, I mean, I think it's about, I don't know, I, I think also kind of embedded in the definition of process, right? Like, certain processes are about, things that you... [00:07:00] Like, let's say, let's say you cook a dish, right?

[00:07:01] Mm-hmm. Like, let's say you're cooking a, a pasta dish, and there are various different things you could do. I, I, I cook myself. I, I cook chicken curry a lot and, and, there are probably fifteen different ingredients in chicken curry, but the, how do you use them and how, how kind of, you kind of make them all work together, is very different. and, unless there is a documented version, and even if there is a documented version, I think, people tend to improvise as they become experts at the same process. Yeah. I do, I do the same thing all the time. Like I would, I would, I would listen to a, a recipe from YouTube, and I would be like, "Okay, I'll, I'll cook this," and the next day I kind of improvise on, on what I- Uh-huh. -What I heard on, on- Right ... on. So I think it's kind of the same thing, I, I see on the, on the process side as well, we-- there might be a time five years ago that when there was a process that was established and documented, but people, started using this. People started following the process, and they, they found s- their own optimizations, which like m- [00:08:00] minor optimizations here and there, and they kind of evolve the process over time. and sometimes they evolve it to their own benefit rather than the benefit of, the, the, the holistic benefit. and that- that's what creates all these different, uh, variations, I would

[00:08:14] Casey Golden: Retail's notorious for not having holistic solutions.

[00:08:19] Ricardo Belmar: Right.

[00:08:19] Diptendu Ray: True. So true.

[00:08:21] Why Retail's Data Is a Dumpster Fire (and the Fix)

[00:08:21] Casey Golden: another, you know, challenge with automation projects, we run into blockers around data. Everybody knows we have a dumpster fire, but, it's gonna take process. We have systems that get in the way, and people. We're set in our ways, right? We've been doing things a certain way. We're good at there is no time when you're on the... running retail.

[00:08:42] There's a fire drill, and, half the time you get to work after hours.

[00:08:46] Diptendu Ray: That is true.

[00:08:47] Casey Golden: what needs to happen in the prep for a project to avoid some of these problems? And one conversation that I've heard a lot lately is you can't [00:09:00] stop running the business and innovating and getting, clearing your roadmap and building that to fix the data.

[00:09:07] Are people reorganizing their roadmaps to fix the data in parallel? what is your real, you know, recommendation on data systems and people getting in the way progress

[00:09:18] Diptendu Ray: Right. yeah, and I mean, I think data, it's kind of notorious. but I would say, there are various different forms and versions of data.

[00:09:27] there's ob-obviously the transactional data which is coming from your POS or from your checkout, and those data is usually fairly clean, so the, the, the things that are kind of working on those is, is, is, it's good. the product data is, is the next second one, which is usually what people are talking about when they say our data is completely, rubbish or completely crappy because the product data is coming from different sources and there's no usual way, or there's no usual governance around it because it doesn't have to be governed. It's, it's more of, whatever comes in, you kind of show it to the customers. there's, there's [00:10:00] no real, I would say, there's no hard metrics to kind of measure what's the impact of having that as clean right. thing. Yes, same thing, right? I mean-

[00:10:09] Casey Golden: Redundant ... it,

[00:10:10] Diptendu Ray: things. Uh, So, my, my point is like, I think it's important to kind understand when you're building something, what, what type of data you're trying to use, and building your guardrails around them, right? it's definitely not feasible, it's, it's probably never feasible to do a data project for a data project's sake where you clean up the data. It's like a one-year project where you're just cleaning up data.

[00:10:34] Ricardo Belmar: Yeah.

[00:10:34] Diptendu Ray: So, I think the, the approach, in what we are taking is kind of, going one track, seeing what are the data we need, what are the gaps, and how do we fulfill those gaps. And, sometimes, the, the downside of that is, sometimes your cost of that particular project increases because there's more data to fix on, on a certain track, than others. but that's kind of the, I would [00:11:00] say the downside we, have to live with, right?

[00:11:02] Right. And, um- Right but other than that, I mean, I think other than clean, data, the other aspect of it is also privacy, security data, tho-those kind of things. and, there are various different ways and means to kind of handle that.

[00:11:16] I believe it's, one way to kind of do this is, have a, a, a more centralized version of it, where you could potentially, use that centralized version to build, something on top of it. And then, you know, when you are using that centralized version, you have built all those guardrails and all your, the, the fort around your most important, most, most, secured, data, like customer data or payments data. and then the other pieces of data you could potentially, have slightly different policies around it. but, but in, in, in, in, in, in, in general, I mean, I think, an approach around data warehouse, building a data lake, that usually is, is very helpful instead of, yeah ... kind of, working ad hoc where you [00:12:00] have data in five- All

[00:12:01] Ricardo Belmar: different disparate

[00:12:02] Diptendu Ray: five different

[00:12:03] Ricardo Belmar: tie in Yeah, yeah

[00:12:04] Diptendu Ray: to, tie it in, in, in, in, in, real time. But that's always a bigger challenge. So, I think that's, that's one part of it, right? Definitely, if it's not about clean data, at least to build- a cen-central repository of data- Yeah where we could-- if you could you could-- which could be leveraged, in the, in the, in in the future. Right.

[00:12:23] Ricardo Belmar: Right. Yeah

[00:12:24] The Shortcut Culture That Kills AI Projects

[00:12:24] Casey Golden: Any tips on, the people piece?

[00:12:27] Diptendu Ray: yeah, I mean, I think, I don't think it's just about data. people sometimes tend to take shortcuts with respect to fulfilling their own, goals. there are projects deadlines to be met. There are, things to be completed, which often, could, be detrimental to the quality data. I think it's definitely about the culture, right? And it's not ab-- just about data. It could be anything. It could be, a software we, we're building and the shortcuts we're taking with that, with the specific- Yeah short-- or it could be a, a, a vendor selection. if we do not take a [00:13:00] long-term view of, what it's gonna look like, there could be short-term,

[00:13:04] Casey Golden: gains and like, long-term.

[00:13:05] Diptendu Ray: Yeah, Yeah. So I think that's probably everywhere. even, even business models or, or even, business ideas could potentially have the same challenges, which is essentially, you know, the, the, the short-term, gains versus the long-term, strategic implications of it.

[00:13:21] And, and I think that definitely comes down to the culture and the leadership of the company and how- Yeah ... they could balance those two. So I would-- my, my, take on that is it's not just limited to data. That problem itself is not major data itself.

[00:13:35] Ricardo Belmar: Right. Right.

[00:13:36] Right.

[00:13:36] AI Efficiency Isn't Cost Cutting: The Mic Drop

[00:13:41] Ricardo Belmar: So one of the other things that we kind closed out our session this morning was talking about, how do you find efficiency out of these aI projects? And so many people when they hear efficiency they just automatically, their, their mind goes to cost reduction.

[00:13:50] But one of the interesting things we've mentioned that there's so many other ways to measure and look at efficiency. why don't you share some of thoughts around here. How do you look at efficiency if that's a goal or if the goal is [00:14:00] efficiency is it? It-- 'cause it's not always cost reduction reduction right?

[00:14:02] Diptendu Ray: No, that's it. yeah. So I mean, I think, one of the things which is, kind of my immediate, I would say goal is inventory efficiency, right?

[00:14:10] Which, which is about optimizing how much space we use. Yeah. How do we increase the inventory turns, and how do we make that most efficient? it has nothing to do with headcount. It's, it's all, all about how do, how do we move the inventory faster, and how do we provide-- A-as the inventory moves faster, we can provide more newness to the customers with in-- with new products.

[00:14:32] So that's one, one aspect of effici-efficiency. the other way I see it, is also how, efficiencies are, driven or redirected towards betterment of, the business, right? I mean, I think, I was talking to a friend, the other day, and it was about, you know, if, if there are productivity gains, let's say with some, some, software or ch- a chatbot that we launched, it's, it's getting, the team, let's say a team of 10 people, three [00:15:00] weeks or five, five, three or three to five hours, per week.

[00:15:03] Those five hours, are not gonna show up in the, on the bottom line unless we are able to channel those five hours into something productive,

[00:15:10] which which, which could- Yeah. Right ...Which could which could be anything about, it could be about building that relationship with the customers. Mm-hmm. It could be about brainstorming new ideas to grow the business.

[00:15:20] Mm-hmm. but it's all, all about how are we, redirecting the efficiencies into, betterment for the, for the, for the company. Mm-hmm. And, I think, NVIDIA CEO has also kind of touched upon this, better ideas, into, into, groundbreaking, inno-innovations across the company. Yeah ... otherwise, it's just, it, it's about, about writing emails and setting up new meetings.

[00:15:40] So then I think it's, it's not gonna, it's not gonna be a, a, a sustainable benefit in the long run

[00:15:45] Ricardo Belmar: Right.

[00:15:46] Diptendu Ray: Right.

[00:15:46] Casey Golden: I mean, I think ending on a mic drop moment.

[00:15:49] Ricardo Belmar: Yeah. That's exactly. Exactly

[00:15:51] Diptendu Ray: Yeah.

[00:15:52] Ricardo Belmar: That's it right there

[00:15:53] Casey Golden: This was really great. Thank you so much joining

[00:15:56] Diptendu Ray: much. you so much. Yeah.

[00:15:57] Ricardo Belmar: Thank you.

[00:15:58] Show Close

[00:16:03] Casey Golden: If you loved this episode, hit subscribe, drop us a five-star rating and review on Apple Podcasts, Spotify, or Goodpods. And if you're watching us on YouTube, like, subscribe, and drop a comment before you go. I'm Casey Golden

[00:16:17] Ricardo Belmar: Follow us on LinkedIn, Bluesky, Threads, and Instagram and subscribe to our Substack newsletter for highlights and bonus content. For transcripts and guest info, visit retailrazor.com. I'm Ricardo Belmar

[00:16:28] Casey Golden: Thanks for joining us on the Retail Razor Show, part of the Retail Razor podcast network

[00:16:34] Ricardo Belmar: Until next time, stay sharp, stay human, and stay ahead.

[00:16:37] This is the Retail Razor Show!

[00:16:39]