Keith Moore (00:00):

A lot of people fear AI as it's coming for their jobs long-term. To some degree that may happen, but we're seeing a huge shortage of labor as it stands right now. So it's how can you leverage this technology to replicate a lot of the tribal knowledge that exists. And so when somebody retires, the operation doesn't absolutely sync with their retirement.

Voiceover (00:22):

Welcome to Supply Chain Now, the number one voice of supply chain. Join us as we share critical news, key insights, and real supply chain leadership from across the globe. One conversation at a time.

Scott W. Luton (00:34):

Hey, good morning, good afternoon, good evening, wherever you may be. Scott Luton and Supply chain Hall of Famer, Jake Barr here with you on Supply Chain Now. Welcome to today's live stream, Jake, AKA the John Wayne of Global Supply Chain. How you doing?

Jake Barr (00:48):

I'm doing wonderful and I am so pumped for today's session. This is incredible. I mean, I get so excited when I get the chance to talk to this guy because he's one of the brightest young minds in the AI space as it pertains how to use it pragmatically in the supply chain

Scott W. Luton (01:08):

That is high praise from Jake Barr. And I agree with you. We've enjoyed our chats thus far with our guests here today. And folks, we've got another great show teed up here. Today. We're going to be focusing in on quite the innovative topic, right? Ag agentic, AI supply chains. We're going to explore just what that is, how it works, the big time benefits that it all poses. We're going to gain some insights and advice, been there, done that advice for you, business leaders out there that want to capitalize on the massive agentic AI supply chain opportunity and all that. But we're also going to get a sneak peek of what lies ahead in terms of how things will evolve and how supply chain tech in particular may continue to evolve in this golden age. So Jake, to your point, it should be a great show, huh?

Jake Barr (01:54):

Absolutely. Let's get to it, baby.

Scott W. Luton (01:57):

So Jake, with all of that said, I want to get to work. Welcome in our esteemed rock and roll guest here today, Keith Moore, CEO, with Autoscheduler.ai. Hey Keith, how you doing?

Keith Moore (02:09):

Doing really well. Thank you guys for having me. And if I've heard correctly, I heard Jake refer to me as young, and so I am just absolutely honored to be referred to as young, my gray hair would say otherwise.

Scott W. Luton (02:21):

So Keith and Jake, before we get into all things Gentech AI supply chain, I want to start with a fun warmup question. Today's many things, it's International Sushi Day. It is International picnic day, and on a musical note, it is Paul McCartney's birthday. That's right, one of the most popular Beatles members of the Beatles. So I want to use that last one to ask you both. And Keith, I'm going to start with you. What is one of your favorite musical acts or experiences of all time?

Keith Moore (02:53):

Yeah, that's a great question. About a decade ago, I had the opportunity to go and actually see Journey in concert. I live here in Austin, so we're at the Austin Amphitheater down at Circuit of the Americas F1 track. And it wasn't the original lead singer, it was their new lead singer at the time. But that band puts on a heck of a performance. It was fantastic to see. Went with a bunch of great friends, had a great time, and it's something I want to have the opportunity to do given the age of the band at this point for much longer.

Scott W. Luton (03:23):

Well, hey, it sounds like an outstanding can't miss concert experience that you had. I love that. Love the band. And Jake, I'll tell you what comes to your mind.

Jake Barr (03:35):

I'm telling you this is tough because all music is good music because it stimulates the brain in my opinion. But I have seen the stones four times separately really. I think I may be the only one old enough to have seen him across multiple, like four different decades. But the Stones then a small trivia point for you, Scott, is I signed Jimmy Buffett to perform for $500 back in the early seventies.

Scott W. Luton (04:10):

You got to be kidding me. Are you serious,

Jake Barr (04:12):

Jake? No, no. When you're young and in college and you're doing a bunch of activities, responsibility for things on campus and sign buffet and the Coral reefer band to come to the campus and literally it was a great experience. I ended up getting to meet not only all band members, but to actually keep in touch with a couple.

Scott W. Luton (04:40):

Alright folks, Jake and Keith, thank you both for sharing and I want to start with level setting to learn a little bit more about you and what the company does. So let's start there. Tell us more about yourself, Keith.

Keith Moore (04:50):

Yeah, so about myself. So Keith Moore, CEO of Autoscheduler. I'm based in Austin, Texas. I've spent over a decade in machine learning and artificial intelligence. So I've been in this space since 2014, had built solutions across a wide variety of industries, but I actually grew up in logistics. So my dad has been a supply chain consultant my whole life. My last company was a pure play AI company here in Austin. I grew that to become a unicorn. So really proud of the work that I was able to do there and wanted to go do my own thing. So I actually got with my dad, we started Autoscheduler in 2020, and the whole intent of Autoscheduler was how can we take all of the technology that exists in AI to perform the function of warehouse orchestration, which means for plants and warehouses? Can we harmonize all of the different data sources that people are using to make decisions? Can we then automate and optimize the facility to prescribe what decisions need to be made at what time and where the bottlenecks are going to be? And can we capture that data over time to identify where there's margin and opportunity in every single facility to better drive service or reduce costs inside of supply chains?

Scott W. Luton (06:01):

Keith, I love that. And Jake, I'm going to get you two quick points there, folks. Keith is very passionate about making this next hour very educational and informative to whether you do business with auto schedule or not. You're going to learn a lot here today. But Jake, on the other side of the coin, he's also very humble. What would you add, Jake? Because you've seen auto schedule up close and in person.

Jake Barr (06:20):

Oh, actually, yeah. I firsthand touched it and actually worked through implementations of it because of the capabilities. But more importantly, I think what's kind of being under sold here is it actually is the first opportunity that's existed to actually eliminate the bull whip across multiple business processes that actually don't talk to each other. So you're talking about warehousing, transportation operations, the inbound flow of materials, all of that, which honestly are the most dysfunctional band that has ever played music on the face of the planet. And when you use the capability, it's like singing completely in tune and more importantly the cadence of the music and importantly, predictably what needs to happen next. Which if you've ever run large scale operations, and I've been a victim of that, that an average day is absolute chaos by default. I don't care how well you plan, the best sleep plan is crap the moment it's literally put together. So you've got to be able to pivot quickly, and that's one of the capabilities that I think Keith's going to get into here.

Scott W. Luton (07:40):

Love that, love that. The orchestration, massive orchestration opportunity to not only shape decisions that our humans are making, but in many cases automate those decisions. We're literal, don't even have to make the decisions. How about that? So we'll get into all that and much, much more. So Keith, I want to shift gears now that we know a little bit more about you, especially our new audience members. You recently wrote about your bold new vision in this recent paper, which we're going to make available here in a minute. The Agentic AI supply chain framework. So two part question, what inspired the framework? And what I'm even more intrigued by is why is now the time?

Keith Moore (08:17):

Okay, so I sat at the Gartner Supply Chain Symposium in early May, and I'm sitting there and I'm listening to every supply chain vendor on the planet. It's like, well, we're going to use AI and agentic AI as what we're doing. And you go and you ask the question like, okay, well how do you do this? What does this look like? Please explain this to me. And I found that there was a large mix of answers

(08:43):

On what ag agentic AI even represented. And then coupled with that, I actually sat in on a presentation that Art Mesher and Amber Sally gave that was all about decision engineering and how the supply chain that we have lived for the past 30 years has become more and more brittle as consumer speed picks up. And I thought there probably needs to be a definitive document on what agentic AI is, how people can actually use it, the power it brings. And it's almost like a roadmap slash playbook that companies can look at and say, it's not just working with vendor A versus vendor B. It's understanding the concept of what are we doing, why are we doing that, and how is that going to impact our supply chain moving forward? So I left Gartner, I think I left on a Wednesday. I spent 24 hours between Saturday and Sunday of that next weekend writing this thing to try and provide that playbook in my mind's eye. And it's obviously going to be open to interpretation, but for everybody to hopefully take and use, it's not about Autoscheduler at all.

Scott W. Luton (09:45):

So Keith, that might be the most consequential Gartner event ever. Given what it inspired you to do, it's going to help so many folks out there. Jake, what'd you hear there from Keith?

Jake Barr (09:55):

I've had the opportunity to read it, review it, and actually provide comment back on it. And I think it's best described as probably the most sane, clear, pragmatic description of how to understand how to use tech in a simple way that I've seen in years. I hate it when folks come out with the terminology that just confuses and then allows and obscene definition, which is always evolving, right? Right. You ask somebody, okay, well what does that mean? Well, and you can never get two definitions that are even close. So I think actually putting it in print and giving it a basis a people to actually compare against was a great first step. But to me, the most sensical thing that came out of it is it brought the true explanation of shaping decisions alive.

Keith Moore (11:00):

Maybe Scott, if you don't mind, to add a little bit more color on the please. That's why I wrote it, right? It was really a reactionary thing to this event where I saw this miscommunication. But from a timing perspective, everybody in the world knows what chat GPT is at this point, right? There's perplexity, there's Google, there's Google's Gemini, there's chat, GPT, there's Claude, there's all these different technologies that people are very familiar with using on a day-to-day basis. All of those technologies, so I'm going to use a term that's fairly technical, are dependent on what's called a large language model. And so large language models are going to be these massive models that are able to take huge amounts of text, distill that information down, and imperfect context answer questions based on billions of parameters, which is, sorry, another technical term. But the reality is anybody can open their computer or on their phone and they can ask questions to chat GPT and it will give them answers. People understand how to do this,

(12:02):

But that same technology, so just because chat GPT exists doesn't mean that you can magically automate everything in supply chain. And so I think that's the most people are just building that bridge and saying, well, I have all these supply chain things I do and I have chat GPT that magically does things better than Google did. I've accessed all this information, therefore LLMs and chat GPT can just help me run my supply chain. But the reality is there's quite a technical bridge that needs to be built because LLMs as they are designed, are not well structured to run a supply chain,

Scott W. Luton (12:36):

Right?

Keith Moore (12:37):

Correct. And so that was kind of the technical reason for why I wrote it, is to help people understand what are the gaps that still exist and how do we fill them.

Jake Barr (12:44):

So let's do this. Things

Scott W. Luton (12:45):

Change. We don't live in a vacuum, do we? Alright, so Jake and Keith, for those slower folks like myself that need to understand this and unpack this a little bit more, Keith, you were kind of segueing into explaining what ag gentech AI supply chain is. So let's go down further down that road. Tell us more what it is and how it differs from traditional supply chain execution.

Keith Moore (13:10):

Sure. So when you look at a supply chain today, it is a combination of systems that are used by people to make decisions, and I'm going to name some of them just to put it for everybody who's worked in this space. You have your ERP, your enterprise resource planning platform. This is your SAP, your Oracle, whatever it is. It has network level inventory management. It's generally where you're getting your orders inside of your systems. And then below that you're going to have all sorts of different systems that help you do your job more effectively. So you, your ERP that performs function number one, you then have an advanced planning system so that you can do supply and demand planning, potentially does system number two. You then have a transportation management system because when you're doing your supply and demand plans, you need to figure out what shipping where or when potentially.

(13:59):

And so your TMS actually tenders that F freight, and I could go on and on all the way down to your WMS is controlling what pick is happening at what second. But the key takeaway is as people leverage these systems all the way from SN OP to an individual pick in a warehouse or an individual delivery at last mile, there is a series of cascaded decisions that are made on faster and faster time loops. So you might do SNOP planning once a month and only update it on maybe on a weekly basis. Same with forecasting or demand planning. You're booking transportation every day. You're dropping appointment times into your WMS every hour you're making picks every second. And so the traditional supply chain that we operate, these decision points happen at different periods of time, which is why they create this bull whip that Jake referred to is that I may have made a decision last month, the whole world completely changed because somebody started firing rockets overseas.

(14:59):

And I'm not going to revisit that decision for another three weeks and that's going to significantly impact what I'm doing in a warehouse today. And so what a agentic AI is, can we take a intelligent decision-making system and wrap our existing execution system so you're not getting rid of your WMS or your TMS or your ERP, but can we look at the decisions that people are making on that periodic cadence inside of each system? And can we start to automate 80 to 90, maybe a hundred percent, but let's start with 80 to 90% of those decisions so that people can then focus on bigger problems inside of the supply chain. So ag agentic AI in supply chain is a function of wrapping existing execution systems with better decision making that is interoperable either with human guidance or without people altogether. I like it. Sign me up. Really long answer.

Scott W. Luton (15:52):

Sorry. No, no, it's good. It's important to set the table accurately. Going back to Jake's point about you get a thousand different definitions, Jake, out of what Keith just shared, how it works, how it differs, what's really important for folks to understand,

Jake Barr (16:05):

Keith hit on a bunch of 'em. The fundamental thing is that you've got something that's being able to look across almost like a neuro network of all the actions that are in flight and actually understand the consequence of one versus the other and actually then interrogate the information on all the activities. It's learning from it and actually being able to use whether you're actually setting some, Hey, when I find these kind of conditions, I want you to do X, Y, and Z or to tell me so that I can interact with it so that I can then intercede and take an action. So that's the piece that we have to understand is yes, it's grabbing all the information from all these individual processes. Yes, those individual processes because of their own silo based way of design had limitations for what window of time they could operate and look at. And therefore we created these bull whip of activities, buffers and time gaps and all that. Now we're being able to step back from that, interrogate it live, and actually as we identify, oh, that's a bad decision to make, bring it up and interrogate it and then either trigger an action from it. Because I've said, Hey, when you find this, that's a bad thing and this is what I want you to do about it. And predictively and prescriptively lay out the path for what to do about it all in a fan that are hardworking men and women in the supply chain can't do on their own today, they can.

Scott W. Luton (18:05):

Really important. You're saying pre-show Jake, we are insight rich, but to your point there, we need to know what to do, when to do it and automate it as often as we can. Alright, so Keith, you were saying earlier, you were mentioning WMST, M-S-E-R-P, all that stuff. I think I heard you say it doesn't replace all of those platforms, right? Speak to that and speak to how the AI agents interact with these existing systems.

Keith Moore (18:34):

So in the framework I proposed in most major manufacturing supply chain, so a retail supply chain is going to look a little bit different. Fulfillment supply chain, again, a little bit different. There are up to seven potentially more critical agents that are going to exist. And so the proposal here is if you look across all of your different systems, you can actually map what decisions are made in which systems at what time. And so each agent, so for example, I'll focus on warehouse where most of where I live there is a single agent that is responsible for warehouse and inventory management. That agent is integrated with A-Y-M-S-A-W-M-S. It might potentially also pull manufacturing data or if you're doing kit operations inside of your warehouse, you might be pulling that data as well. But that agent is doing a few things. The first thing is it's pulling all of that data into one place.

(19:30):

So you have a unique agent focused on a business process inside of my supply chain. It's pulling that data into one place. It is then not just exposing it via an LLM. So I saw there's a question in the chat about, well, what LLM does this? The answer is there is not a single LLM out there that can just pull in this smorgasbord of data and tell you what to do with it, which is where that agent and part of the critical piece of each particular agent, you're going to see more and more vendors start to emerge with this. The critical piece of that agent is there are all sorts of heuristics and optimizations that exist on that agent to understand how to interpret that data. If it's Manhattan versus blue yonder versus Kerber versus soft Dion versus SAP, right? It interprets that data, it understands the appropriate way and guidance to make decisions based on the workflows of reality in the supply chain. And what the LLM is then doing is it is simply taking all of that and providing great context to how to actually interrogate or make decisions on that data. So based on this data, here are the seven things we need to action right now. And that's both for human readability,

Scott W. Luton (20:39):

But

Keith Moore (20:39):

It's also for system readability. It can go and automatically action them in the WMS or potentially I can go and pass this information to my transport agent to tell them that I know in the warehouse this particular load is going to be seven hours late because we won't have the inventory for another six hours and it's going to take us an hour to load it. I went pretty deep there, but I

Jake Barr (20:58):

Like it. Oh, it's beautiful, Keith, because the reality is, and I've already seen witness, and this is why you said, Hey, coming out of Gartner, this is one of the things you felt compelled to do because I've literally had discussions with, I'll call it solution sets that are already out in the marketplace, that are in one of those individual verticals, those siloed verticals, and they've been dutifully running off working to build ag agentic capability in their siloed world. And I'm going, no, no, we're going to recreate what we currently have. We, so these things have to be able, as you aply describe, wait a minute, this is a big complex problem solve. And so I'm only as good as how much data I'm interacting with and interoperability of it. And so I've got these agents as I call it, that need to dance with each other across processes in order for us to really win in a major way across the entire supply chain. So unlimited possibilities and capability that this thing is going to unlock for us, but we need to make sure folks aren't doing it in a vacuum.

Scott W. Luton (22:18):

And we got to be able to, as you say, unlock it, unlock it to pick up where Jake left off the key technical enablers required to make this system, make this opportunity viable and make it happen today. Keith, what are they?

Keith Moore (22:34):

So key technical enablers. Thing number one is going to be a good foundation of data. I hate the trope of garbage in garbage out. It's so overused. But what we're talking about even with AI agents inside of supply chain is these are really complex systems that are operating on a component of the supply chain and they only know what the data is there to tell them, I have yet to meet a perfectly digital company that has a great uniform perspective on all of their data and it accurately represents reality. But the first foundation is you have to have some form of data and understanding of how decisions are made on that data in order to do this at all. So that's kind of foundation number one. Foundation number two. So once you feel good about the data, you need to then understand how is that data actually used to make decisions?

(23:28):

So what decisions can be made? And understand you can't automate decisions that you don't have data to make. If you go to a site and there's a smoke signal that they use that's not digitized for figuring out when they want to bring a trailer in, there's no way to actually do that programmatically. So having that process map is thing number two. And then thing number three, and this is something I see a ton of questions about, how do the LLMs do this or how do the LLMs work in different components in the chat, right? The key emphasis here, particularly with age agent AI and supply chain is we run a very deterministic business. You can't be wrong, you can't hallucinate. And so LLMs are really designed for that interoperability of the seams, as Jake mentioned in that how do we take all of this data and the decisions to be made simplified into the right thing for the right person at the right time as well as the right thing for other systems at the right time. But they are much more so part of the communication layer than doing a lot of the heavy lifting on thinking about how the operation should be run holistically because LLMs are not well designed to do that. So I just want to reemphasize that.

Scott W. Luton (24:38):

Yes. Yeah, so Jake, your quick comment, he just shared three technical enablers and including that last one, no hallucinations. We're dealing with critical,

Jake Barr (24:48):

They're great and the base foundation on the first one around data. And that's why when you say, Hey, how do you get started in doing this? Actually one of the best steps is actually as you bring in and do ingestion of the information and the surrounding data sources, you actually are running, I'll call it a quick litus test to actually tell you what the hell's wrong with it and where are the gaps before we get to the let's go use it to make decisions. Let's take the first step to actually go and do a sanity check on it to say, Hey, I've got some capabilities you may not have had 'em before to actually identify to you how really bad and where you're bad across your process areas for when data is available or the quality of it. And if you can then identify those, remember this capability is learning, it's constantly able to learn so it knows, hey, Keith screws up on that data flow about one out of every three times and here's the variability of what we see from what he actually ends up submitting. So now I have the ability to actually start to do smart interrogation of how I would act when I see certain data flows occurring. So I am ditto is three steps are great.

Scott W. Luton (26:24):

Alright, so what factor, Keith, the, so what factor kind of real world benefits can companies expect from a adopting the model and the framework that you're suggesting?

Keith Moore (26:35):

There are obviously going to be risks associated with, but as far as benefits go, the sky's the limit, right? From just a general technology adoption perspective, supply chain has historically been a laggard, which is fascinating when you look at overall spend in companies, supply chain's, about 1% of revenue for big CPG food and beverage, sorry, 10% of revenue spend for big CPG food and beverage companies. And so you're talking about faster decision making, you're talking about better decision making, which at the end of the day is going to increase service levels. So there's potential for actually increasing top line revenue and from a margin perspective increase margin significantly because part of what thes agentic AI should be doing is even if it's able to semi-automate or automate 50 to 60%, let's go on the low side of what some of the day-to-day functions look like inside of supply chain operations to keep things moving. You're talking about freeing up a huge amount of man hours, so a huge amount of people time to actually go and do other critical tasks. And from a just pure EBITDA perspective, that lift for a company is going to be massive, big enough to move the needle on the stock price. And so this is fundamentally going to change how supply chains operate, not over the next six months, but I think over the next six years pretty significantly.

Scott W. Luton (28:03):

Okay, and we're going to talk more, dive into your very accurate crystal ball in a minute. Jake, add to the benefits.

Jake Barr (28:12):

I'm going to use fundamentally one statement to kind of visualize in your brain what your outcome really is.

Scott W. Luton (28:21):

Okay.

Jake Barr (28:22):

It's creating calm from the chaos. Calm from the chaos. And I say that because fundamentally we are continually knee jerking our way through multiple iterations of how to execute the work not only today, this week, this month because of the level of volatility and process variability that we've been subjected to and to submit to you to say that that's actually dampening and it's getting better versus worse is absolute bull. Okay. I'll add some explicative on the end, but what we're missing perhaps in the discussion or getting clarity on is it's being able to bring calm to the chaos of how to determine and orchestrate your precious resources against real problems, not noise that's being generated from the level of process variation. Keith gave you examples, wait, I've got people in a warehouse in a DC or a plant that's bringing stuff from lines and trying to determine whether to put it away, pick it, get it, move it.

(29:43):

Oh God, I've got another set of people that are actually working with transportation suppliers and I'm sorry, but the latest thing I see is a normal day for any midday to large scale company is 50,000 kind of alerts or updates or transport updates or things that are behind or late or going to be affected. And you think about that in just an operational context. Holy crap. How do you bring those together and still say, Hey Keith, make sure you get the work done today out on the floor in the operation. So being able to redirecting to the most important task at the right moment based on the last set of cascading volatility is critical.

(30:43):

I don't have time for Jake to go off and figure out what I need the people to do. That's dead time, that's cost. As Keith said, we can't afford it. We just simply don't have the fundamental capability of being able to do it. The consumers definitely aren't willing to pay for it in the price of the product anymore. So you've got to find a way, how do I dampen the volatility, take the noise out, get my operationally fundamentally quickly focused on mission critical variations where I'm actually being led to what to do. I don't have to determine

Scott W. Luton (31:26):

That. Love that Jake and Keith, whether it's that or any sector. What are some of the most promising early use cases or pilot projects that you've seen thus far?

Keith Moore (31:37):

Sure. So I can speak to something that's public knowledge at this point, which is at my company. We've done a lot of work with PepsiCo and Frito-Lay in rolling this out for decision making enablement at a lot of their manufacturing facilities. And I can't give the hard and fast numbers that they're realizing return on investment has been fantastic. I will say that we're months, not years

(32:00):

And what that's netted out, I think the public data that could be shared is on an average A site's productivity, and these are going to be complex manufacturing facilities with lots of different inventory flows, all sorts of different task types that people are responsible for doing and coordinating in and around the manufacturer, the actual plant process. Those facilities on average see a bump in productivity and throughput of about 12% and the on time in full number skyrockets up to service levels are well above 99%. So when you're talking about what is the objective inside of most facilities in a supply chain, it is how do you maximize service and minimize cost? Both of those two things are being realized in spades. And we're also, I can speak to that one because we have a really great partner that's helped us with a lot of, and through a lot of the challenges that these types of projects run into.

Scott W. Luton (32:53):

Yep. Hey, Jake, really quick, what we're talking about is not theoretical, what Keith is just talking about. It's actually in place today.

Jake Barr (33:02):

It's in place,

Scott W. Luton (33:02):

Yeah,

Jake Barr (33:03):

Jake, it's in place. This is real. You can touch it, you can use it, you can execute against it. In fact, I would define many that are probably on the session today. I'll ask you a very basic question. Do you actually have the ability to look proactively 36 48 hours out across what your schedule should look like and what are the changes that are actually have happened over the last couple of hours that are influencing how you ac crewe your facility? How you think about the partner flows into the facility? My guess is most are going to answer with no, I can't

Scott W. Luton (33:42):

Probably so especially if they're keeping it real to themselves and to the discussion. Jake, Keith, now it's time to get some real practical advice from you here today because again, this is theoretical. You just were touching on some of the big returns from companies that have made the investment. So for business leaders out there that really want to capitalize on this opportunity that you've written about, what's one of the biggest hurdles that companies have got to be prepared for as they're trying to capitalize on this massive opportunity?

Keith Moore (34:12):

Oh man, this is my favorite question ever because normally it's like I'm an AI guy, his answer is going to be data. That is not the answer I'm going to give, which is the biggest hurdle you run into when you're rolling out AI systems is change management always, right? Supply chains like to operate the way they currently operate, they are generally going to be change resistant. And that doesn't mean people are change resistant or all people are change resistant, but thinking about how you roll this type of technology out in a way that enables people to be more successful at their job, particularly in a lot of functions, you can't hire the stat that you really need to run a good supply chain. And so there's no real risk of pulling a ton of people out of an operation. But a lot of people fear AI as it's coming for their jobs long-term to some degree that may happen, but we're seeing a huge shortage of labor as it stands right now.

(35:11):

So it's how can you leverage this technology to replicate a lot of the tribal knowledge that exists? And so when somebody retires, the operation doesn't absolutely sync with their retirement and that I speak very much so to plants and warehouses, the world I primarily live in, but this applies to any function in supply chain. And so change management is always the number one hurdle. You need to have a really good plan for how you're going to roll this out, how you're going to track metrics associated with it. So aligning on what those metrics are, how are you going to hold people accountable to use and feedback of the system and grow to make sure that it fits your business really well? There's not a one size fits all answer for change management. If I had it, I would probably be a billionaire living on an island. So I think everybody's lived through it before, but it needs to be front and center when you're taking on a project like this so that you think about how are we going to get ahead of this now so we don't run into potential failure of a project down the road.

Scott W. Luton (36:08):

Jake, your comments on that hurdle or any other hurdles you see?

Jake Barr (36:11):

Yeah, I want to echo the change management, but I also want to bring the human element to it because it really is about changing the behaviors of what you expect, those that are in the operation and the roles that they play because now we're talking about, wait a minute, we've basically glorified firefighting for so long and trying to pull or ram it out of the head, you now need to pay people to actually make a business decision because I've actually gone and determined all of the chaos and now I say, Hey, we have option A or option B of how we're going to execute the next couple of hours. So now I need them to put on their business leader hat, not just a firefighter who was sitting at the dock trying to say, well, there's 12 loads that we've got to somehow magically figure out how we're going to fit into two dock doors.

(37:12):

So it's changing again, the role of the folks, the capability is going to continue to progress further and further, and what we have to appreciate is that we're equipping our folks with better ability to deal with all the noise and then to come forward with how to make informed business decisions to help us drive the EBITDA impacts, as Keith was talking about, right, instead of, well, I thought I made the right choice of how to queue up those doors and how to use John and Sally and Bill for the next three hours. And you know what? We left four loads short this afternoon that we probably wouldn't have if I had just use the information.

Scott W. Luton (38:04):

Alright, so Keith, what would your advice be for those companies just starting out that want to take that first step towards building an AgTech AI supply chain?

Keith Moore (38:13):

Yeah, and I will echo that. If I were to talk to a company that was thinking about going down this journey, it's a few things. So step number one, and this is the basic must do, is map your data to your processes. You don't necessarily have to say, I'm going to get all of our data in order and it has to be perfect. Do not do that. Chances are the bigger your organization is, the harder that is going to be. So one at least understand what data do we have, how do we leverage that data to make decisions, have that ready? When you're looking at starting an AI project, the next thing you need to do is say, where is the biggest value that I can get with this data and what do I want to attack? And so I would start a pilot project there. How can we take something and leverage an agent to better make decisions or even just inform decisions out of the gate? Let's start that as an AI project and really start small, understand what are the metrics we're going to be tracking? How do we do that week over week? What does that system need to look like? Who are the right people to work with as a part of that go through vendor selection. The smaller the better, the faster the better.

(39:17):

So it doesn't need to be perfect. Your data doesn't need to be perfectly aligned. Just getting started and moving forward is going to be critical. The reality is you could plan for six months and you're going to fall flat on your face day two of that project for some reason or another. And so working in a very agile way is certainly encouraged as technology like this is getting adopted more and more and as it's developing naturally. So get started, do something, at least have that process map when you want to start.

Jake Barr (39:48):

The great thing about this is that, and Keith just gave you a couple of tips, there are so many problems and opportunities in a distribution operation that you can really peel it off in any way. So as an example, I deal with a number of companies who they're capital crunched at the moment as an example, and they simply, they need an additional building, but they can't afford it. They can't. So they're trying to say, how do I squeeze as much business through the doors that I a avail have based on the chaos of my order flow to maximize the absolute total amount of revenue that I can get squeezed out of those doors. That's an area I've got others who Keith said earlier, wait, God, we've got labor issues going on. It manifests itself in a lot of ways, and in many ways it tends to root itself in things like picking in warehouse operations or order fulfillment or whatever.

(40:50):

And so you get a lot of people that go, well, geez, I don't even know how to crew the pick area. I don't know what's coming at me based on the disruptions. How many do I need and how do I think about 'em in the next 2, 4, 8, 16, 24, 36 hour shifts, right? Where am I going to get 'em? Do I have enough schedule? Do I have an availability of enough? That's another one you can do probably. I mean, so you could peel many of these off as small incubation starting points for how to begin your journey and guess what? They're all viable

Scott W. Luton (41:27):

Target rich environment. Indeed. Jake and Keith. Keith, Jake was talking about the next 36, 48 hours ahead. Let's expand that horizon. Let's se over the next three to five years, how do you see all of this evolving and what excites you the most about what's to come?

Keith Moore (41:45):

Oh man, just it's a blue ocean right now. Everybody's talking about it. A few folks are really doing it and kind of moving quickly. There are going to be set vendors and set companies that mature very quickly over the next year in this space. And from a vendor perspective, what you're going to start to see is there are vendors that start to take dominance on the different domain specific agents that exist. So warehousing obviously being the one I'm going after, transportation planning, and it could be the incumbents in those space from a technology provider, but it could be completely new companies that come in and establish themselves as the 800 pound gorilla in this space in their particular sector. But that's going to be the next year to two years. Over the next three to five years, you're going to see more and more companies focused on inter agent communication that is still very immature. There is not a standardized protocol for it. MCP came out from OpenAI. That's not really for agent to agent communication. You're going to see a lot more of that where you're going to have a transportation agent and the warehousing agent communicating back and forth in an autonomous fashion. And reality is that's years away

(42:55):

Technology wise possible in months, but I don't think that's going to drive to broad adoption in years. And then five years from now, you're going to see there was this concept of control towers that has lived and died many times over the past decade. That control tower concept is going to be the one agent to rule them all the one ring, so to say that helps determine the framework for how different agents should interact within a network. And that's going to be, I'm not sure that's going to be a single vendor. I think that's going to be the most innovative supply chains are going to build their own to control what happens inside of their network. But that's where I see this going and there's a lot more detail than what I just covered.

Scott W. Luton (43:39):

There really is Jake, we're going to make Keith, we're going to make Keith talk about auto schedule because they're doing some really cool things, even though he is a humble, humble individual. So two part question, Keith, I'll start with the first one. How does auto schedule fit into this vision of the AG agentic AI supply chain and what role is it? Again, this is not theoretical folks. What role is it playing in enabling today?

Keith Moore (44:02):

We are the warehouse agent, right? I bet my company and our future on doing this successfully, which is how are we the agent that makes decisions, that control layer of a warehouse wrapped around existing execution systems to enable everybody to make the right decision at the right time to do that successfully? There was a massive, I mean, this is five years of infrastructure to be built before you even talk about putting LLMs on top of it. So I think we have a heck of a headstart in the space, but it's something, obviously this is very personal to me. I had love to say I'm a CEO, I'm going to do all of it all the time because we're a tech company. That's not the reality, right? You need to carve out what we are really good at, warehousing plants and distribution and moving inventory and finite in a finite space over a finite period of time and enabling some of the biggest, most successful supply chains in the world to make better decisions, drive higher service levels, and at the end of the day, run a more efficient operation.

Scott W. Luton (45:05):

Jake, what would you add? As someone that has seen it,

Jake Barr (45:08):

It's impossible to put into a few words the impact it can have on how it affects the agility of your business, your ability once this is in place. So I've witnessed it firsthand to allow you to slide shuffle changes that in the past would've been Herculean for you to undertake. So that's an under called piece that Keith and his team I've seen really haven't spoken to a lot, but it is an outcome that's incredible because you actually, you don't have to run and hide from change anymore. You can actually lean into it. We have such chaos out in the market, right? The level of volatility of your material suppliers and your production schedules and your transport partners, et cetera. Look, when you're dealing with your commercial partners, right, the places you sell product through, the last thing you do right now in many cases is actually lean into a desired change from one of those partners because you are so worried about your heightened risk of failure because you can't control, you're having to admit yourself. I'm not in control of my operations. Okay? This gives you the ability to actually go out and actually say, okay, well hey, would you want that a couple hours earlier if we could make it available to you? Oh, well, let's talk about that.

Scott W. Luton (46:43):

I like how you're keeping it real. As always, Jake Barr, you've got a great talent for doing just that. We're going to get Jake Barr's patented key takeaway in just a second. But before we get there, Keith, we were talking as I've been doing my homework on you and the whole auto schedule team, and I think a lot of the companies that you work with seem to value that accelerant, that ability to go from zero to 60 much faster and with less pain by leaning into third party experts. So to that end, how can folks get started working with Autoscheduler today? It's got to be pretty easy, huh?

Keith Moore (47:20):

It's certainly very easy, right? You fill out anything on our website, there is probably somebody that will get in touch with you in the next 24 hours. You're also welcome to add myself or anybody on our team on LinkedIn. Please reach out. Genuinely speaking, for everybody who's listening in, this is a subject my team and I am very passionate about. So even if it's just information and you're trying to discuss this more, want to understand a little bit more about the space and where it's going, please reach out. We're obviously very excited about bringing it to the world and some of the things we're doing with some of the world's largest supply chains. So would love to help people that are looking for that from a business and economic perspective, but also genuinely happy to contribute to the thought leadership component as well.

Scott W. Luton (48:06):

Well said Keith. And folks, I can tell you, having spent time in person with Keith at ProMat and other places, and now virtually several times, you'll enjoy a frank conversation with Keith, and he loves to make it about industry and make those powerful and agnostic driven learning conversations that you can tell he's passionate about helping folks find calm in all of this chaos. Okay, Jake Barr, we're going to share that resource again from Keith and the auto schedule team in just a second, but we've covered a lot of ground last hour. What is your patented key takeaway from today's conversation?

Jake Barr (48:43):

If your idea of AI is using a chatbot, boy, have you messed the boat? I'm saying that very bluntly because yes, that's a great way to simplify the distillation of some available information, but quite frankly, it's somewhat static. What we're talking about here today is when you're in operational chaos and you need to create a structured plan to maximize the efficiency of both the operation and your financial performance, then that's our focus. Because fundamentally, the to sail guys, AI and agentic AI are here. It's real, and it's something you can use.

Scott W. Luton (49:30):

Well said, don't miss the boat. Don't miss the last train to Clarksville get with the opportunity that exists here today and with folks that are doing it and have done it and shown the results. So let's do this. I want to make sure folks, be sure to check out the Ag agentic AI supply chain framework. In this, you're going to learn about the shift from sequential workflows to real time orchestration, which we've talked about here today. You're going to gain insights into a pragmatic roadmap to adoption. You're going to learn more about the challenges that companies got to address to realize the tremendous benefits, all that and much, much more. Get your own copy and give us your feedback on what you think of it. But Keith would welcome that. But all of that aside will be dropped the link to the white paper, the framework right here, one click away from downloading it. Big thanks, Keith Moore, CEO with Autoscheduler. Keith, I appreciate you and your team and how you're transforming, how supply chain is done here today. Thanks for joining us.

Keith Moore (50:32):

Thank you for having me.

Scott W. Luton (50:33):

Look forward to the next one already. Jake Barr, the John Wayne Global supply chain. I appreciated your commentary here as well. Always a pleasure to host these conversations with you.

Jake Barr (50:46):

That was great. Anytime I get to spend time with the brainiac here, I enjoy it.

Scott W. Luton (50:52):

Hey, that goes for two of us. Really have enjoyed you and Keith's perspective and very actionable perspective, folks, and that's where we're going to wrap here today. As we say, big thanks to our global audience for tuning in, but you got homework. The onus is on you to take one thing from this conversation, just do one. There's probably 27 things, but take one thing from Keith and Jake. Share it with the team, put it into practice all about D's, not words. And with that said, on behalf of the entire Supply Chain Now team here, Scott Luton challenging you to do good, give forward, be the change that's needed. We'll see next time right back here on Supply Chain Now. Thanks everybody.

Voiceover (51:30):

Join the Supply Chain Now community for more supply chain perspectives, news and innovation, check out supplychainnow.com. Subscribe to supply chain now on YouTube and follow and listen to Supply Chain Now, wherever you get your podcasts.