Dave Clark, Founder and CEO, Auger Sanjay Radhakrishnan, Senior Vice President and Chief Technology Officer, Sam's Club Moderator: Jason Del Rey, Tech Correspondent, Fortune, Co-chair, Fortune Brainstorm AI
Category
🤖
TechTranscript
00:00I like it.
00:01Who's that?
00:02Oh, yeah.
00:03I like it.
00:04We're live out here.
00:05When I hear supply chain, I do the same thing.
00:06Every morning.
00:07Every morning.
00:08Dave, it's great to be on stage together.
00:10Now that you're out of Amazon, you're allowed to do this.
00:14I'm totally unburdened by what it's been.
00:17I'm a founder.
00:18I can say any kind of crazy shit I feel like today.
00:21I'm excited.
00:22Perfect.
00:23And I said, I come in the jacket, and I have two great-looking gentlemen, but in a little
00:28more casual attire.
00:30So whoever told me to wear a jacket, you're fired.
00:33Sanjay, great to meet you as well and have you up here.
00:37I guess we should actually talk about the content, right?
00:39So Dave, I'm going to start with you.
00:42Over 20 years at Amazon, most recently, as we know, consumer CEO.
00:48You've now raised $100 million in, I think you're calling it the Series A round for a
00:52new startup focused on creating an operating system for supply chain software.
00:59Can you explain a little bit of what is going wrong in the supply chain space and how AI
01:06plays a role in solutions?
01:08Well, there's been a tremendous amount of investment in the space for decades, and especially
01:13through COVID, there was a tremendous amount of money that went into the space.
01:16And so you have a lot of interesting niche players that are there, right?
01:19So everybody runs an ERP software, and then they run a half dozen or a dozen other software
01:24that does demand planning and transportation management, what have you, but they don't
01:27talk to each other.
01:29And so you're left in this world where everything is sort of outputting an answer, and they
01:34have to post-process this in some form of Excel or Tableau or Smartsheets or something
01:39with the teams of analysts to try to come up with an execution plan.
01:43And our intent is to basically go in and allow them to be able to connect all these systems
01:48in an integrated way so that you can actually operate the end-to-end supply chain in a longitudinal
01:54continuum as opposed to what they're doing today.
01:57And supply chains are always, global supply chains at least, always difficult with circumstances
02:04arising out of your control, right?
02:07And sometimes political-related, other times maybe in a global health crisis.
02:16Now, we have with the Trump administration coming in some questions of what tariffs might
02:20look like.
02:21Are there solutions out...
02:24Are you hoping that part of the solution that you're building is something where you take
02:30a little bit of the uncertainty or at least planning in a more succinct way?
02:37Well, I can tell you, like, most every supply chain professional around right now is trying
02:43to answer the question for somebody, their board or the CEO or somebody who said, hey,
02:47if tariffs go to X, what happens to our supply chain?
02:50Like, do we need to change who our supplier is?
02:53What does our flow need to look like?
02:55What does our inventory look like?
02:56What are the ramifications of, you know, country X putting an alternative rule in place that
03:02blocks something else?
03:04And you know, everybody's using chat GPT or something and getting an answer really quick
03:09and they want to know why the hell their supply chain leader can't come back with this answer.
03:13Well, there's 50 people grinding out spreadsheets trying to come up with these answers.
03:17And what we want to do is try to...is we want to get all that data in one place so that
03:21you could actually ask those questions.
03:24And this is where the, you know, the larger models I think will really help us is getting
03:28the full data set together, which frankly most companies, you know, maybe Sam's Club
03:32excluded and Walmart, you know, outside of these big companies with large technical infrastructure
03:37really don't have the ability to get their data in one place to ask those questions and
03:42to run scenarios with lots of alternatives in them to get a better view of what might
03:47be in certain outcomes or what your decision choices could be given a set of potential outcomes.
03:52So as part of you wish that you're...I know you're building the technology right now,
03:56but this would be a perfect time to launch?
03:58Well, the one thing about supply chain is that there's always something going wrong
04:02somewhere.
04:03So I'm not particularly worried about it.
04:04Like I've been in supply chain for 20 some odd years and every day is just another cluster
04:09of some kind.
04:10You can finish that term if you want.
04:12I know.
04:13Okay.
04:14Yeah.
04:15Trying to be friendly.
04:16Yeah.
04:17But the...it is a...every day is some form of mess.
04:18It's just what is the mess and where is the mess and who created it and how do you figure
04:22out how to get out of it.
04:23And, you know, I've been in this a long time and know the pain that operators face every
04:29day and we just...we want to try to find a way to make that a simpler set of choices
04:35for people so they can really focus on the strategic outcomes as opposed to just answering
04:40questions for the boss.
04:42And Sanjay, I know you have some supply chain background, but I wanted to shift gears a
04:47little bit to your role at Sam's and I've always been fascinated with Sam's Club.
04:51I don't know how many people know, but Sam's Club is often the breeding ground for technology
04:56progress and innovations inside the greater Walmart company.
05:01And then often technologies go from testing in Sam's and then scaling to Walmart proper,
05:07as I call it, or the super centers network.
05:09We've had a ton of talk about gen AI capabilities, but Sam's Club is using different types of
05:16AI both in helping consumers sort of reduce some friction in the stores and also when
05:24your products do get through the supply chain challenges and get to your stores, making
05:29sure they stay there, they stay in stock.
05:32Can you just tell the audience a little bit about what innovations are working for you
05:38right now as it regards to AI-powered solutions?
05:41Yeah, it's great to be here with all of you.
05:45What I'll say is that we, as a company, we are a people-led, tech-powered, omni-channel
05:50retailer.
05:51So at Sam's Club, we have over 600 clubs spread across the US and Puerto Rico.
05:58And we have learned that to win over time, we really need to listen to our members and
06:03figure out how to use technology to anticipate and meet their needs.
06:08So we've been doing this for a long time, listening to our member feedback and figuring
06:13out how can we use that to make their shopping experience more frictionless in the clubs.
06:20And what's number one on that list?
06:22Yeah, so number one on that list is a few years ago, we got feedback from our members
06:27that waiting in the checkout lanes with the cash registers was a huge pain point.
06:32So we delivered this mobile capability called Scan and Go.
06:35It's available in the Sam's Club mobile app.
06:38Our members can actually walk into the club, open the Sam's Club mobile app, just start
06:41adding items to their cart, slide to pay, and head to the door.
06:46Now when they hit the door, they do have a friction point because you have an associate
06:50at the door waiting to check the purchase receipt.
06:53So we got feedback from our members that that's additional friction that they would like us
06:57to help solve.
06:59And so we have come up with this technology called Sam's Exit Tech.
07:02And it's essentially a combination of computer vision and artificial intelligence that kind
07:07of works seamlessly to validate our member purchases and then allows them to exit the
07:14club friction-free.
07:15So for people who haven't seen it, can you just describe, is it an arch with cameras
07:20overhead?
07:21That's correct.
07:22Yeah, so as you head to the exit of a Sam's Club, you'll see an archway.
07:27And the archway has cameras in it.
07:29As you walk through that archway, the cameras are taking images of the member's cart.
07:35Then our CV models kick in to identify items in the cart, which is done through feature
07:40extraction.
07:42So we are looking for things like size, shape, color, logos, brands, and cross-referencing
07:47it with our inventory catalog.
07:49And once we identify the items in the cart, then we are looking for the receipts from
07:54our point-of-sale system and trying to match the items on the cart to a receipt.
07:59Once we have that match...
08:00And this is all in how many seconds?
08:02It's all happening near real-time, right?
08:05So you can think about a club where you have the archway is right before the door.
08:10So you have a few seconds before the member can exit the club.
08:14So in a matter of a few seconds, we do that real-time validation and send a signal to
08:18the associate standing at the door that, hey, everything looks good, and the member walks
08:23friction-free.
08:24This technology is actually rolled out across all of our 600 clubs.
08:29And we are seeing that our members can exit at a rate of 75% completely friction-free.
08:36And it's a huge member satisfaction booster for us.
08:42Our exit experience NPS scores are close to a 90.
08:46And what it's also doing is it's actually giving time back to our associates in the
08:50front of the store so that they can use that time and spend it with the members and make
08:55sure that the shopping experience is more pleasant.
08:58I had one question about that, because I heard the 75% number, and that's more than half.
09:06That's a big number.
09:07I can do math up here, too.
09:10And at the same time, that means it's not 100, right?
09:13And so there's a staff member waiting.
09:16So I'm curious, when you think of productivity, is that person doing other things when they're
09:23not checking out?
09:25Or I'm assuming they're still standing by.
09:27So I'm just curious how you think about the labor dynamic there.
09:31Really, the role of that associate standing at the end of the club is transforming to
09:38actually greeting the member, thanking the member for their service, and really telling
09:42the member that we appreciate the business.
09:45So the tech that we are deploying actually helps us in that journey, because where previously
09:50they used to check every receipt, now they are actually spending 75% of time back, and
09:56they're using that time to actually greet the member and show them that we trust them
10:00and we are not judging them, and we are creating a shopping environment that is actually more
10:05pleasant and trustworthy.
10:07I'm going to come to the audience for questions.
10:09I don't know if we have any right now.
10:10Otherwise, I'll come back.
10:12Otherwise, I have no shame in calling on folks that I know in the room, so maybe right here.
10:19So on the labor point, and Dave, I'm curious for your take on this, too, obviously, when
10:27I'm here at these discussions or taking part in these discussions, productivity, that word,
10:33I even used it, surprised myself, productivity comes up and automation comes up, and there's
10:38the natural labor question.
10:40Dave, I'm curious with the development of the operating system you're talking about,
10:45how much as you start to talk to clients, is labor something they want to solve for,
10:53and just how do you think about the impact of what you're trying to build on workforces
11:00of your end clients?
11:02I think you get to look at companies that are always trying to take cost out of the
11:06system, right?
11:07You're trying to improve service levels at lower costs, and the big cost drivers for
11:11most companies, labor is a big chunk of that, but when you look at cogs, transport, there's
11:18a lot of other costs in there that tend to be bigger, frankly, than a lot of the unitized
11:23labor costs, particularly when you start talking about things that you can automate out.
11:28For what we're building, it's much more about reducing the overall cogs in the system, so
11:33taking working capital out by improving turns, getting cycle times faster, pulling inventory
11:40waste out of the system is a much greater driver of bottom-line impact to most organizations
11:46unless you're really talking the wholesale fundamental labor redeployments in some form.
11:53For us, it's much more about working capital, lead time adjustments, transportation optimization.
11:59We think there's a lot more interesting dollars in that space right now than there is in labor.
12:07Robotics is going to drive a ton of labor savings in that over time, but our focus is
12:12really about much more of the working capital utilization and transport.
12:16Is that also what you're hearing from Fortune 2000 companies, or maybe in public they say
12:25that, but when you're in a room just with the chief supply chain officer-
12:28I think they want both.
12:31Everybody wants everything.
12:33Everybody wants to give customers everything and extract maximum cost out of the system.
12:40Unfortunately, we've made productivity a bit of a dirty word, and we've had productivity
12:49things in place for a long time with humans.
12:53We keep finding new and better ways to engage people and people's brains and what people
12:59do for a living, and I think we're going to continue to do that.
13:04I think there's too much hand-wringing over that and worry about it, but I think when
13:09I look at supply chains, the labor savings is interesting, but the total end-to-end waste
13:17and overproduction, underutilization in transport, lack of forecasting capabilities, all those
13:26things, both planetary waste, human waste, and cost, dramatically outweighs what the
13:33opportunities are in front-line labor savings.
13:35Do we have any questions in the room?
13:38Otherwise, I'll keep going.
13:40Nothing yet?
13:41I'm going to call on someone.
13:42I will.
13:43Sanjay, another technology I know you all have been talking about that's AI-powered
13:49is you have what I think used to be called scrubbers, basically mobile robots that clean
13:59the clubs, and attach them now, I think, is it in some stores or all stores, are cameras
14:09also?
14:10Yes.
14:11These are helping with out-of-stock issues and just giving more knowledge to employees
14:17on inventories.
14:19Do I have that moderately correct?
14:21You have it right.
14:23Josh, inventory management in a physical retail environment is extremely complex because things
14:30move around all the time.
14:32A few years back, we fitted our autonomous scrubbers with cameras.
14:39These scrubbers traverse the aisles of our clubs, taking pictures of merchandise on the
14:45shelf.
14:46We take an average of 24 million images a day.
14:50We have CV and AI models that actually extract information out of these images and convert
14:56them into insights and inventory intelligence that we feed back into systems that our club
15:02associates use in their handhelds.
15:05We use it to drive the next best action for our club associates.
15:10For example, if our member's mark peanuts is running low on the shelf, we can actually
15:16generate a signal to our associates to drive a restocking action.
15:20Or if the price sign is incorrect or the price sign has fallen down, we can drive actions
15:25to our associates to go fix it.
15:27We also use it to actually update locations of items so that the next time the associate
15:32wants to find that item, it's much easier and we can send them to the right location.
15:37We have also added RFID readers to these scrubbers.
15:41As they traverse the aisles, they are reading, on average, 180 million tag reads a day.
15:48We can drive actions like if the small size of a T-shirt is running low on the apparel
15:53rack, but we find that inventory is sitting in a box below the rack, we can drive a restocking
15:59action and ask the associates to go restock the small size T-shirt back on the shelf.
16:05Using these solutions, these inventory management solutions with our AI models and some other
16:10innovations, we have eliminated in excess of 100 million tasks from our associates'
16:16lives in the clubs, and that's time back to our associates to spend with our members on
16:21the floor.
16:22I just want to make sure I heard you correctly.
16:23100 million.
16:24100 million tasks.
16:25Correct.
16:26I'm trying to think of how a role could have that many tasks, or this across different
16:31roles in the club?
16:32It's across different roles in the club.
16:34Got it.
16:36One more call for questions.
16:39No?
16:41You're my man.
16:42Can you tell us who you are and where you're from?
16:45Oh, yeah.
16:46Right here in the front.
16:47Do we have a microphone we can hand this gentleman?
16:50The mic runners gave up.
16:52Yeah.
16:53We can hear you.
16:54Do you mind standing up?
16:55No, not at all.
16:56Okay.
16:57I'm Arlan Deedy with Insight.
16:58Sanjay, what was the friction from your associates when you were employing this technology?
17:04Because there's always friction as new technology gets deployed, especially for lower-skilled
17:09workers.
17:10Yeah.
17:11What I'll say is we've done this enough that we know how to roll out and manage through
17:19the change management.
17:20We have different environments at the home office, kind of lab-like environments where
17:27they can simulate activities like this.
17:30And then we have smaller-sized clubs where there are members coming in and shopping,
17:36but it's a much smaller-scale club where we can actually start off and run the first pilots.
17:41We also have the option to actually run some pilots on a couple of aisles.
17:46And so we have a mechanism to kind of slowly bleed this in, test it out with our home office
17:54associates first, then take it to one club, a small-sized club, then take it to a larger
17:59club, and then test it and then roll it out to scale.
18:02So change management is an issue, but we have done this for a while that we have figured
18:08it out.
18:09What else?
18:12On that topic, I'm always curious how you go about – and Dave, maybe from Flexport
18:17or Amazon experience, you might have some thoughts as well – how you go about getting
18:24real candid answers from employees about how new technologies are impacting them for better
18:31or worse.
18:32And they might be more willing to give the open, the positive answers, right?
18:37But if there's a manager or several layers of management between them and whoever's
18:42receiving the answers…
18:44My experience has been the opposite of that, which is usually you can get the negative
18:47feedback really quickly.
18:49The positive feedback is harder to get.
18:51Completely agree.
18:52Yeah?
18:53Feedback is – I mean, our club associates know how to do their job, so when we give
18:59them tech that works, we hear about it immediately.
19:02When we give them tech that does not work, we hear about it really fast, too.
19:07So they're very open with feedback.
19:10Change is bad, even good change.
19:13And so any time you roll out something – I mean, this has been my experience – no matter
19:18how phenomenal you are at rolling it out, it is always bad for a little bit.
19:22And so you get the negative very quickly, I find.
19:25And the hard part, I've found, is working through, keep pushing through it to get to
19:30what is fundamentally broken versus what's just learning pains and being able to sort
19:38the difference.
19:39And one path is to execute it over – sort of do the Matryoshka, step it up over time,
19:44have something bigger.
19:45The other is burn the boats and throw it out there.
19:48You only have one choice, but to figure it out.
19:51I live in the latter.
19:53In the latter, burn the boats?
19:54Yeah.
19:55I think next conversation we're going to have is on burning boats.
19:58We're out of time now.
19:59Thank you, Dave.
20:00Thank you, Sanjay.