Challenges and Opportunities of AI in Logistics With James Sutton From Summit Advisory Team / Elevate- Ep. 01

Lori Boyer 00:00

Hi. Welcome everybody to Unboxing Logistics! My name is Lori Boyer, and I am going to be your host in this amazing new podcast. I'm super excited because today is our inaugural episode. Yay! In today's episode, we did a lot of chatting as a team to try to decide what are the most important topics that we're getting a lot of questions about, and who could come and talk on them.

And so today we came up with the topic of AI and AI in logistics. And of course, there was no one better. Immediately to my mind, James Sutton—the James Sutton—came to mind as our guest. He is amazing. He's held all kinds of titles, like chief analytics officer, director of enterprise data solutions, manager of supply chain, demand and analytics, and all those things that have kind of those nerdy number things in their title.

That makes me a little intimidated. But welcome, James!

James Sutton 01:05

Thanks for having me. 

Lori Boyer 01:06 

So excited to have you. Can you really quick before we get going, just share a little bit with our logistics community (we love them) tell us a little bit about yourself and your background. 

James Sutton 01:16

Yeah. My name is James Sutton. I've been with the EasyPost Family for about a year when our company, Summit Advisory Team / Elevate, got bought by EasyPost.

I oversee our supply chain analyze product offering, which is Elevate, as well as our analyze consulting practice. My background is more retail, more business-focused. So I actually started working at Best Buy when I was 16. I worked in stores for seven years and I still attribute a lot of my like, retail knowledge to those days of being the person who did that job.

And then I went to manufacturing for International Paper and learned everything about manufacturing and supply chain and supply planning. And then after that went to Finish Line JD Sports and helped them stand up a supply chain analytics practice, planning practice, so then enterprise data, and move their enterprise data strategy into GCP and Looker in the cloud.

And then three years ago we founded Elevate and the supply chain analytics practice and have been helping enterprise shippers and retailers improve their operations. So it's been quite a journey to get here. But I'm excited to talk more about AI 'cause our industry's changing really quickly and we're in the middle of another AI hype cycle.

Yeah. So it's the perfect time for this podcast. 

Lori Boyer 02:26

An AI hype cycle. I love how you said that. Yeah. I love AI. I'm a super AI geek-out nerd. I just think it's a super exciting time that we're living in. Where it's like we're seeing real-time changes, but we're gonna get to that in a minute. One thing I wanted to say that I love about James, and you guys in the community are gonna love about him, is that he actually has a little bit of a marketing and sales background.

And he and I were just talking about that, how he has a really unique approach to things because he's way into numbers, way into all of the analytics. But he also really can know how to speak to people, and I think that's such a cool skill. So shout out to James for that. But before we get started, I think it's awesome if we can get to know you a little bit better.

Awesome. So we're gonna do what I call our 30-second speed round of this or that. Most of 'em are just generic. Tell us what you prefer. A couple of them are actually even logistics related, so we'll get the real deep insight on you, so, okay. Coffee or tea? 

James Sutton 03:28

Coffee for sure. Every morning. Latte.

Lori Boyer 03:30

Latte. 

Summer or winter? 

James Sutton 03:32

Definitely summer. 

Lori Boyer 03:34 

Oh, you like to get out and have fun? 

James Sutton 03:36 

Yeah. I'm a golfer, so I want warm weather. 

Lori Boyer 03:40

Fiction or nonfiction? 

James Sutton 03:42 

Ooh, I'm probably nonfiction actually. 

Lori Boyer 03:44 

Yeah. Seem like a type. Always reading up on how to get smarter even. Okay. Mountains or beaches. 

James Sutton 03:49

I would say both 'cause I live in California, but I would pick mountains. 

Lori Boyer 03:52 

Yeah, me too. Dogs or cats? 

James Sutton 03:55

Oh, probably dogs, but they're, they're definitely a lot more work, so.

Lori Boyer 03:59 

Yeah, they're a little bit like having a kid, I'd say. 

James Sutton 04:01

Yeah. It's like high maintenance, but high reward. 

Lori Boyer 04:03 

Yeah, a hundred percent. City or country? 

James Sutton 04:06

I like city living. I'm a really big walker, so I love being able to walk to everything. 

Lori Boyer 04:10

James was just telling me a story about how he tried camping. Oh, yeah. But it didn't quite pan out, so... 

James Sutton 04:17

Yeah. Should we go into that? There's just a lot of ticks involved. 

Lori Boyer 04:20

We could have a whole other episode on that one. Yeah. Okay. Pancakes or waffles?

James Sutton 04:25 

Oh, waffles. Waffles. Definitely like the crisp waffles. 

Lori Boyer 04:27

Yeah. I love it. Okay. What do you think is more important, first mile or last mile? 

James Sutton 04:32

Oh, I mean, last mile for customer satisfaction. Yeah. 

Lori Boyer 04:35

Yeah, I agree with that. And just in time inventory? Or just in case inventory? 

James Sutton 04:41

Oh, it depends on the business. 

Lori Boyer 04:43

Oh, I knew he was gonna say that. He's such a politician. So you guys have to tell us in your comments what you think. But okay. That was super fun. But I do wanna get into our segment on AI. So, what I'm hoping that we all come away with is an understanding of some of the challenges in AI, the opportunities in AI, everyone's talking ChatGPT, everybody's talking about AI lately.

And so what can our listeners do (or our viewers) what can they do today to prepare? And kind of just what, where do we see AI going? So can you start maybe by telling me a little bit about your personal journey in AI? And you know, especially how that ties into logistics. 

James Sutton 05:27

Yeah. So I would say my experience is different than a lot of people in the data space.

I started more on the business side, as an analyst using data in order to make decisions. And then eventually through some roles where I didn't have data available to me, I became more of a data engineer. Where for the past five-plus years, I've pretty much been focusing on the infrastructure of data and cloud compute and streaming and transformations, and building good data models and not on AI. 

And I would say that like, you know, my Slack title is non-artificial intelligence because I've been almost anti-AI until recently. 

Lori Boyer 06:00 

Real intelligence. 

James Sutton 06:01

Yeah. Well, I just think that like, there's, you know, you go to a conference and every company is an AI company until you ask them what AI is and then, and then they're like, ah, you know, they don't have a good answer.

Right? And so, I think just recently things have kind, like, have flipped a little bit with ChatGPT. I mean, AI has been leveraged in applications, in production for years with like Netflix's recommendation engine and with Uber's driver selection. But I think ChatGPT is really the first time that AI is becoming like, more mainstream and people are seeing like, real use cases.

And so my experience now is, how, what are those biggest ROI events and opportunities where we can leverage better, larger data sets in order to make better decisions. And so I'm driving that forward with our company right now. 

Lori Boyer 06:46

I love that. I love the idea that AI has always kind of been around. I've heard for years that like even, you know, I work in marketing and so as a marketing professional, there have been studies that show that if you just put the word AI, people are like, oh.

Well, I should get that because it's AI, right? But we don't always really understand it. I would love to know from you, what kind of things here in logistics have already had AI sort of driving them for a long time? I know for instance at EasyPost we have different analytics to check the speed of delivery or things like that.

Where maybe would people already see AI that they weren't even aware was driving it? 

James Sutton 07:25

Yeah, so I think personalization was like the first like really big mainstream use case for AI. And that's more on the marketing side, but that's been like, leveraged by companies for like a decade now, and and that continues to get better as external data gets better.

And so, you know, everyone's being targeted based off of the the data that you're creating on your mobile devices and that that's, yeah. You know, I think everyone, it, behind the scenes, that's what's happening. And logistics, I think it's a little bit tougher. I think I already mentioned Uber, but dynamic driver selection based off of like geolocations of phones is is definitely a really cool use case.

But in more enterprise shippers you know, a lot of those things are not being leveraged today. It's all scan-based. Yeah. And it's more like legacy route planning. And so there's like, the supply chain industry I think is a little bit more lagging as far as leveraging some of these new technologies and, but that's also really cool that there's like a ton of opportunities in that space. 

Lori Boyer 08:18 

Yes. I love that you said opportunity, because I had a boss once upon a time and he never liked us to say, oh, this is a problem. He always said, this means there's an opportunity. Right? Yeah. So that means that everyone who's not doing it, that is an opportunity for growth.

So where do you see personally, Where in the industry is there opportunity? What kind of areas do we have the biggest opportunity in AI? 

James Sutton 08:40

Yeah. I mean, within logistics, I would say there's like three big ones on the top of my head right now. One, for shippers (EasyPost has been rolling out the smart rate service) selecting the right carrier service.

Lori Boyer 08:52

I'm gonna interrupt you. This smart rate service is like what I was talking about with where you check how, how quickly a package could get there just by actual real-time data, which cost, which carrier is gonna give you the best price. Things like that. Just for those of you who don't know, that's all AI-driven as well.

James Sutton 09:06

Yeah, and I think that's really cool. I think like TMSs and, and dynamic, I guess rule-based carrier selection has been in the industry for a long time, but what EasyPost is doing, trying to leverage the millions or billions of tracking events that are in there, in the system in order to detect a confidence level that a package will get delivered on a required delivery date, and then selecting the cheapest of those based on your desired confidence interval, is a huge change for shippers. 

And so I think leveraging, you know, large data sets in order to dynamically change who you're shipping with. And as companies are more open to onboarding more carriers, which they should with everything what with UPS, et cetera, right now, is definitely number one is shippers should use some AI in order to select the right carrier service mix. 

Lori Boyer 09:55 

Yeah. I was just talking to some of our customer support people and talking about different customers we have and whatnot. And they mentioned that most especially e-commerce, if you're in e-commerce tend to stick with one carrier.

So that kind of expansion into the multi-carrier—why do you think people avoid that, maybe is that just a little too scary? I can see how AI can be really powerful. 

James Sutton 10:17

It's actually more of an operational issue than a technical issue. So like actually being able to have truck pickups from multiple carriers from your distribution center is a different operational challenge that you have to plan for. So for big operations where you're doing, you know, hundreds of, or you know, hundreds of thousands of shipments a month, like it's, it's easier to fill truckloads and, and like have many lanes. But when you have smaller volume, it's like one carrier is easier to manage.

Lori Boyer 10:44

That makes sense. 

James Sutton 10:44

And then two from like, you know, your IT side, usually those teams are smaller. And so to like onboard more carriers is a time effort, to pay multiple carriers is a time effort. To have analytics on multiple carriers is a time effort. So I think there's like a few obstacles but you know, I think software companies like EasyPost are trying to build solutions to make that easier. But the operational challenges are still there that like, need to be overcome. 

Lori Boyer 11:09

Yeah, that completely makes sense. 

So where else do you see other opportunities? 

James Sutton 11:12

Yeah, so I think the, the biggest opportunity I see is within like, tracking of packages.

I think a lot of things, right, like the whole industry exists on this like tracking number ecosystem where tracking numbers get reused and they, they may be unique for 90 days, and then they get scanned into different buildings and then eventually they get delivered and hopefully they get scanned. And then that's how you, we track packages in the ecosystem today. 

Lori Boyer 11:34

And so there's not AI involved in that right now? 

James Sutton 11:36

No. It is just, it's all event-based, and then they publish events, and then like there's software providers that give, you know, you can track it on the company's website or you might get an email from a company.

But the whole tracking ecosystem is just, has opportunity, especially with IoT devices. So there's... 

Lori Boyer 11:52 

What's IoT? 

James Sutton 11:53 

Internet of Things. So it's like a very generic word, but there's like, you know, you can buy RFID tags from, you know, some pennies now. You can get different … there's like many Bluetooth devices. There's different ways of tracking locations of packages that are starting to, retailers are starting to use in their stores, but I think shippers to start to use on their packages, so that you can actually track where a physical package is throughout the life cycle of where it's in the building.

I mean there's a whole loss package ecosystem that companies lose a lot of packages, and then shippers like have to submit claims and it's this whole process for lost packages. Whereas if like we have better data on tracking where those are and then leverage artificial intelligence to call, call out those exceptions there's just a lot of opportunity there.

And then if you know better where your packages are and you have better data on that, then you can build more solutions to dynamically route plan based on different capacity constraints. 

Lori Boyer 12:50 Are you saying we can like hunt down the guy who like stole the package off my porch or not going to that level?

James Sutton 12:56 

We're probably not that level, but you know, with video AI you can probably detect who that is at some point, so...

Lori Boyer 13:02

And that's just crazy to think, but lost packages are a huge problem in the industry. And I think that when you, when you consider the fact that, I mean, the numbers are enormous, the people who are ramping up buying online, getting packages, but the customer expectations, like they kind of expect you to be almost perfect. 

Right? So it's like two-day shipping is free and that the package is gonna arrive on time. So being able to have that AI tracking that you really can reduce lost packages, I think that's, that would be really awesome.

So do you see that coming soon? 

James Sutton 13:39

No. No. There's a lot of infrastructure, there's a lot of hardware like, and investments that need to be made in order to, like, for the industry to change. I mean, what I just mentioned of like having physical hardware that like on the label, like, no one's doing that yet. And so... 

Lori Boyer 13:52

Hey, opportunity, you entrepreneurs out there? 

James Sutton 13:55

Yeah. The, the, the most advanced that I see right now in like last mile is like Amazon with like, they're basically leveraging driver phone location. Or to detect where a package is, so this last year you might have seen that they, you get notifications like, your driver's nine stops away from your delivery.

Right? So, but they're leveraging like, a tracking number is assigned to a truck, and the truck driver has a phone. It's not actually done in the package, so...

Lori Boyer 14:18

It's not in the package, it's the delivery person. 

James Sutton 14:20 

So, but it, it's still like really awesome to see how companies are innovating in that space. Yeah. And we'll see where it goes. 

Lori Boyer 14:27 

I love how you say there's opportunity. That again, our audience is awesome out there. Somebody go invent it. What else? Do you see any other opportunities specifically? 

James Sutton 14:36 

Yeah. It'd be hard for me to talk about AI without talking about forecasting. That's my background and like demand planning and so, so …

Lori Boyer 14:44 

I love forecasting, I think it's so cool. 

James Sutton 14:46

I mean, we just filmed a webinar about network optimization and we talked a lot about inventory locations and inventory planning. And I think like, even though that doesn't have to directly tie to like last-mile delivery, but like having the right inventory in the right location and then dynamically routing your orders to that right location, there's still opportunity in the ecosystem in order to create models and leverage models to plan better. Yeah. And so …

Lori Boyer 15:11 

Well, there's massive amounts of data. I mean, there's so much information out there. Yeah. I feel like forecasting; it's one of the things that's really, really interesting to me. I was just talking to someone today saying, I wanna do a whole topic on demand forecasting 'cause I think it's super cool, but I think it can get a lot better. So I agree with you that I think AI would be an awesome opportunity there. Anywhere else, I wanna talk about some of the challenges, and I think we've kind of touched on them a little bit, but I wanna make sure, are there any other opportunities you see before we move on?

James Sutton 15:42 

I think that's a good three to start with. 

Lori Boyer 15:43 

Okay. Awesome. Okay. Challenges, I mean, right off the bat, you mentioned the fact that we're a little lagging in the industry sometimes in keeping up with technology, so I'm guessing that's a challenge, but what challenges do you see as well? 

James Sutton 15:58 

Yeah, I mean, I think that most non-data professionals don't really tie these new AI models to what's required underneath it in order to under like, to build to build insights. And so what's cool about ChatGPT and these large language models is that they've spent, you know, a lot of time building up great data and then exploring that data and making it available in, in these models.

But in order to productionalize an AI model on top of internal data, you have to have everything like your data strategy in a good place. Like, you have to have all your data in the same spot. You have to have it defined as far as like what sales is or what shipments are or what inventory is.

You have to have everything in a good place from a data engineering ecosystem before you're ready for AI. And so I think that's one of the biggest challenges is that there's either the data doesn't exist on what you're trying to build, like, you know, having tracking data in more like detail level granularity, or it's like you haven't done the effort in order to build the data infrastructure to support your models. And so I think in general that's the biggest challenge. 

Lori Boyer 17:03 

So are you saying what, I mean, what does it look like if they have the data in place? If a company does are you saying most of 'em have data, but it's just not connected and it's not organized well or, you know, which is a bigger problem to you not having data...?

James Sutton 17:20 

So it's, I think it's actually the business side defining data. That's the biggest issue. 'Cause you, you, it's like AI is not just like magical where I take a Python package and I put on my data warehouse and just works.

Right. Like, you have to actually define and engineer what you, what model you're trying to train for, where those data elements exist. And do kind of that mapping exercise from like, here's my data to here's my model and then here's my like desired output. And so there's a lot of work there.

I was just at Manhattan's user conference recently and the CTO was talking in the key keynote about ChatGPT and how they're thinking about it. And, and it was like it, he did, his keynote was great. Mm. One of the ... 

Lori Boyer 18:00 

Shout out! 

James Sutton 18:00 

Yeah, shout out to him. Well, the use case he used was a DC supervisor asking like, what are my opportunities in picking? And the ChatGPT or the long, large language model, like spit back out and said, Hey, based off your current backlog and your current labor utilization and the net, your next shift planning, you need four more pickers in this area, right?

Like it gave a very prescriptive answer, which is like, what an operator wants, right? Yeah. But in order to answer that question is like, is very difficult from a data perspective because right like that in Manhattan's labor, like warehouse management system, that's like 12 tables where all that data lives.

You have to have a data strategy to tie all those data elements into the large language model in order to spit back the answer. Hmm. It's like the models don't just like know the data structure and figure it out on its own so there's, you know, there needs to be investment in that area in order to build out those functional use cases.

And then it's gonna be like the whole industry is gonna change as far as how users interact with software. 

Lori Boyer 19:07 

Yeah, so I think it's a great point. Sometimes, I mean, there's been so much hubbub and people freaking out about losing jobs and stuff due to AI, but what I'm hearing you say, and what I think and I've experienced as well, is we still need those people to be filling in the data, providing that information, asking the right queries to get us to the next point. 

One thing that you and I have talked about before in the past is you shared something that I thought was super cool that we should have, you know, kind of data people across different organizations. And I feel like that could come into play here. You know, we need to have advocates and more of a focus on data than we have in the past.

Would you agree with that? Do you think that we've got in place the personnel to handle the kind of data needs we'll need moving forward, or do you think companies should start looking at, you know, more, more data personnel? 

James Sutton 20:05 

Yeah, so I see companies invest in AI and data. Which is great, but you have to have a really focused approach as far as how you're gonna leverage that.

So you don't really want to build infrastructure for infrastructure or models for models, and then, like, nothing's changing with your business. So, yeah. You know, some people disagree with me, but I feel like data and AI is a little bit of a support function, so the real people who you're trying to influence are your operations teams, your analysts that live within marketing or live within each finance or any of these functional areas that are actually gonna be making different business decisions based off of your company's data.

And then your AI and your data functions are empowering those. And so I, I think that like, as the industry evolves, it's like how do we serve those people better? And then how do we build tools in order to make their experience better? And so the whole hype cycle around ChatGPT is interesting because it's changing the landscape on how users are gonna get insights from data.

And so, the whole idea of being able to change it into more of a natural conversation of being able, like ask, ask an analyst what the answer to my question is, and then having a large language model who knows my data, like answer it is like super intriguing and can be a lot faster. I definitely don't think that like, AI is gonna kill jobs.

It's just gonna change jobs. 

Lori Boyer 21:27 

Yeah, I agree. 

James Sutton 21:27 

So like, I agree, certain things don't need, just don't need to be done anymore. Like, we don't need as many like database admins that are just entering things in anymore because we like, you know, invented processes for scanning instead of just entering in things.

So like, the, the industry's just evolving and there's gonna be new jobs in order to make these things powerful. 

Lori Boyer 21:44 

I love how you said that. I was speaking with a videographer I worked with recently, and he was sharing a little bit about, you know, how in the industry when not AI, but computer animation came about decades ago. And how some of the artists were like, no, this is pure, we can't mess it up. And then some of the people were like, let's get it. Let's learn these new tools. And obviously it's those who adapt to the tools, who accept them and figure out how, like you said, to augment your job and how to make what you do faster and more efficient and better.

Those are the people who are gonna end up being successful long term. So I don't think we should be afraid, just like you said, but what can … so let's say that somebody's listening right now, they're a warehouse manager, they're a logistics executive. They're whatever your role is, those of you in the community. What should they start doing right now?

Obviously there's a lot of different systems that include AI. But outside of that, for them personally, what would be your recommendations? 

James Sutton 22:45

I mean, I think the, if you wanna get to know where we're at in the current state of AI, use ChatGPT, like, just get a hang of it. Start asking it questions.

Even ask questions that may be relevant to your job and that you're a subject matter expert on, and see what ChatGPT has to say about it because it, the reality is it's probably close to being right, and there may be some things you disagree with. And so I think just using it is number one.

And then two, I think like as a world, we kind of just need to like be skeptical about AI. So just because the computer says it doesn't mean it's right. 'Cause the underlying data and the interpretation of the underlying data, it's difficult. Like what we're developing, we're developing in real time and it's really fast.

And so like the outputs of these models are not gonna be right all the time. And so I think we, everyone just needs to operate under a certain level of skepticism and still trust your understanding of the business. And just use this as a way to make yourself better faster. Yeah.

I think like one of the most intriguing things I've seen in this space is how the developer workflow is changing with ChatGPT. So developers now can use ChatGPT to ask how to do certain tasks within development. Like whether, like, I'm trying to write this Python application, here's what I want it to do, and it'll write like first version for you and then you just have to edit it to like do what you want it to do.

And so, like how we are even developing applications is changing. It's no longer Google and Substack or whatever. It's now like asking these large language models to help me with your workflow, so yeah. 

Lori Boyer 24:19

I love, you said you had a whole bunch of nuggets there. I loved what you said about being skeptical.

Because I use ChatGPT all the time. I really love it. I am just wanting those kind of, I feel like James and I, I said this to him before, we're like the same person because while I work with people, I love numbers too, and he loves numbers and people. So we're the same, but one thing that I found is when I dive into ChatGPT, it's not always right, but I like to have discussions with it.

And ask it best practices and ask it to share some of the research. So I would think even if you're dealing with a difficult customer issue or if you're, you know, if you're playing with it personally, you're saying, I wanna bring up this topic to my boss. I wanna advocate for this. What are some best practices?

What's gonna help? There's a lot of ways to use it. But I love, love your suggestion of trying to figure out what it does know in your industry, what it doesn't in your specific job role, and just getting comfortable with it. I think to start. So that's really critical. Okay. Well there's so much that we could talk about, but I do want to ask you before we go, any predictions you have? This is like looking into our crystal ball.

And I also just want to know like, what excites you the most? What, what are you excited about for the future? What do you wanna see? What do you want your daughter someday to experience in the logistics world with AI? What do you have to say about that? 

James Sutton 25:53 

Oh, those, those are some big questions.

Big questions, big questions. I mean, I'm really excited on how the industry's changing, I think how people are using technology and interactign with software applications is just evolving in front of our eyes. So a lot of people are saying, this is like the second dot-com boom. And it really is.

I didn't believe it five years ago, but now, like, you're like starting to see it and starting to see the opportunities. So I'm just like, really excited to see how people change, how they leverage technology in order to run businesses or run like operations and run their life even. Right. Yeah, I think that's really interesting.

It's also really scary, right? There's just like so many concerns as far as like, what data do these large models have access to, and how do we protect ourselves and protect our company from ... 

Lori Boyer 26:41

Security! We're gonna have to have you back, James. 'cause that's a whole other topic we could get into, security.

James Sutton 26:46

It's a huge issue. How do keep, keep like, you know, bad players from impacting your models, right? Yeah. So I'm, I'm both equally excited and just like scared that I feel like we need to make sure that we're making the right investments in the right areas and being skeptical enough in order to not roll things into production that like could have adverse effects. So yeah.

Lori Boyer 27:07

I love that. Equal parts excited and skeptical. And embrace, but also kind of, you know, trust yourself and trust your knowledge that you have. That's awesome. I love it. Okay, James, I'm sure people may wanna reach out to you if, and if anyone has questions for James or for me, you're welcome to reach out.

But James, where can people reach you? Where can they connect with you? 

James Sutton 27:30

Yeah, I'm active on LinkedIn. I post about data. 

Lori Boyer 27:32 

He does post some cool stuff!

James Sutton 27:34

And analytics occasionally. So you can follow me on LinkedIn. You can follow my hiking adventures on Instagram. 

Lori Boyer 27:40

Nice. Nice. See, he said he wasn't outdoorsy, but there you go.

James Sutton 27:43 

Yeah, I do like hiking. I just like going back to it. 

Lori Boyer 27:45

He likes to go home and shower at the end. I like it. Okay, so follow him on LinkedIn, connect with him if you have any questions. It's been awesome having you for our inaugural episode. You really brought all kinds of insights, so I appreciate it, and everybody out there have a great day, and we'll see you next time.