550: Discover The Blueprint For Your Autonomous Supply Chain, with FourKites

Speaker A: Awkward time to ask this, but. Hey, did you download the trail map? Yeah.

Speaker B: No, I don't need to.

Speaker A: I don't understand. You're trusting your signal out here.

Speaker B: I'm trusting T Mobile. They have the best network. And if we end up in bumtots nowhere, well, we've got T Satellite for backup.

Speaker C: Whoa.

Speaker A: I don't trust my carrier that much.

Speaker B: We'll just use your phone as a flashlight.

Speaker A: With America's best network and T Satellite, we're keeping you connected in places you never thought possible. And if you switch today, you get free phones for zero down and only 25 bucks a month per line for four. Find out more@t mobile.com or visit your local store. Best Mobile Network Based on analysis by Ooklev Speed Test Intelligence data to age 2025 with 24 monthly bill credits and 4 eligible port ins on essentials for well qualified customers with autopay plus taxes, fees and $35 connection charge per line credits and imbalance 2 if you pay off earlier, cancel contact US Finance Agreement example $299.99 Moto Edge 5G required T Satellite available with compatible device in most outdoor areas in the US where you can see the sky included with experience beyond or $10 a month. However, news monthly cancel anytime visit t

Speaker B: mobile.com let's talk supply Chain is not your average supply chain podcast. We feature not just the top of the industry but also diverse, uh, voices from within the community, new innovations and the disruptors making waves in the industry. Don't listen to the same old same old. Be sparked by new ideas and fresh perspectives. Only on let's Talk Supply Chain. Hello everyone and welcome back to let's Talk Supply Chain. Now before we begin, I always have a question for you. Today we're talking about AI agents. So with technology from today's guest, what percentage of exceptions resolve themselves? What do you think? Well, let me know. Your guess is over on social and keep listening because I'll let you know at the end of the show and I might even pull Jimmy in to see what he knows. Today I'm joined by a company that's helping brands turn real time visibility data into automated execution. Who is is that? Well, it's Forkites. Now Forkites is the global leader in AI driven supply chain transformation technologies, helping the world's biggest brands and businesses turn data into action like never before. And today I am joined by Jimmy Sebastian, VP of AI Products. With more than 25 years experience, Jimmy brings deep expertise in creating AI that's not just powerful but Explainable, secure and enterprise ready. And today we're going to be talking all about the topic of the day, which is AI. We're going to be exploring how visibility went from differentiator to table stakes and why agentic AI is the missing layer. We'll be talking about the problem with AI washing. Ooh, I'm excited about that. And why autonomy is a spectrum. And we'll be also sharing the simple steps you need to take to get started and how to measure scalable success. So welcome to the show, Jimmy.

Speaker D: Thanks Sarah. Uh, it's great to be here.

Speaker B: Yeah, I'm really excited to talk to you today. I mean, listen, we talk about AI all the time in supply chain, but today I think we're going to get to the nuts and bolts of it and I'm excited for everybody to hear that. Now Fork Heights has been on the show a couple of times before, so if you have not checked them out and you want to catch up, go and check out episodes 536 and 235 wherever you listen and watch the show. But I do watch want to send you to letstalksupplychain.com because we have our AI interactive podcast experience on the homepage and you can consume that content however you want. Plus it's interactive and we also have a couple more upcoming episodes over the next few months, so make sure to look out for those as well. But today we're taking on the industry's favorite topic and I can't wait to get your perspective. So to start us off, why don't you introduce yourself, your background and your role at Forkites.

Speaker D: Yeah, thanks Sarah. I'm Jimmy, uh, Sebastian, VP of AI products at For Kites. And uh, I've spent about 25 years building enterprise AI analytics and uh, machine learning platforms, uh, including what we call hybrid AI which is really bringing together all the different approaches of AI, machine learning, symbolic and knowledge based AI and generative AI, all of it together and of course now agentic AI. And um, I'd say all through my career I've worked on essentially making data more actionable. So starting with accessing, analyzing it and taking decisions based on of it and then finally increasingly acting on it. And that's exactly what we are doing at Forkites, um, where we have a digital workforce of AI agents that take on thousands of manual supply chain tasks and turn them into automated workflows. As you said, I care about AI that's really, uh, not just powerful, but explainable, secure and enterprise ready. And that's what we have seen, uh, that really serves the enterprise.

Speaker B: I can't wait because we talk about AI all the time, but I'm really curious as to the explainable part of it. But then also you mentioned hybrid AI. Now that's another term that I haven't used. It's not necessarily another acronym that I can add to my supply chain dictionary, but it is, you know, um, words, words that we're using around AI and they keep popping up every single day. So I'm glad that when you said hybrid grid AI, you really explained that and I'm excited to learn more. So give us a very brief reminder. Who is forkites? What do you do? Who are your customers that you serve?

Speaker D: Right. So Forkites, uh, we started off as a real time supply chain visibility company. Um, that's like the classic where's my truck? Uh, and that's where we began, right? And then we built the, one of the largest carrier networks, data networks in the world. I mean about 3.5 million shipments are flowing through our uh, network every day. Um, so that's, that's pretty powerful and you know, it serves our customers. But we have evolved a long way past beyond that. Today we are an intelligent control tower, um, especially execution uh, focused. And I think of it in three layers. When you think about the uh, ICT strategy that we have, we have the data network which is the backbone, uh, of it, and then the digital twins which layer on top of the data network and kind of give you a digital view of different entities like shipments and orders, inventory, et cetera. And then our agentic platform which is Loft, which actually makes all of uses the data that's available in the network, uh, filtered through the twins and then actually takes action. So you know, from going from uh, data to decision to done, uh, most recently, um, we are um, almost refounding ourselves as an AI native company, especially uh, for the supply chain front. And as I said, the huge focus is Loft, which is our new platform. It is home for our AI agents. So if you think about what is loft, right, that's where our AI agents live. It's a managed governed place where the agents can be deployed, um, and then they can be monitored and improved upon in terms of their workflows. Um, so that's where we are. So we are um, an AI first natively, um, AI driven intelligent control tower. That's the evolution. And who do we serve? Um, some of the biggest brands on the planet, Fortune finder, shippers, food, uh, and beverage cpg, grocery retail pharma manufacturing, you name it, um. Right. And plus all the carriers and brokers in our network.

Speaker C: Wow, I love that.

Speaker B: And it's always fascinating to me because we in supply chain, we name our technology and I think they do it in other industries as well. But you named it loft. And I'm always fascinated about the story behind naming the technology, um, because I've heard it all and it's interesting to me that you called it loft.

Speaker D: So it's a play on the kite, uh, four kites. Right. You store kites in a loft. And that's kind of where we, um, ended up with that name.

Speaker B: Love that. Thank you for indulging me for that moment. Now, visibility has been the talk of the industry for a few years now, but the dialogue is really changing. Can you talk about how visibility really went from differentiator to table stakes and what that means for businesses, but also how agentic, uh, AI is helping to bridge that gap?

Speaker D: Yeah, so it's a really big shift. Right. Um, so ten years ago, um, just knowing where your truck was, that was a great competitive advantage. But today everybody has it. It's kind of table stakes. A lot of companies are providing it and uh, uh, shippers and companies, they just expect to have that data. But, um, here's the problem that we continue to see as, uh, we're in this space. Companies are typically information rich, uh, with all the data that's coming at them, but action poor. Right. And so they've got all these dashboards lighting up, um, alerts are firing all day, but they're still having to fix a lot of the problems. Ah, with a lot of manual effort like emails, phone calls, spreadsheets. Um, and uh, you know, visibility tells you about a problem, right? It tells you that it exists, but it doesn't actually fix it. And so that's where the whole value of this has shifted to, um, you know, from seeing the problem to actually solving it. And that's the missing layer. So that, that, that's what, you know, uh, what agent AI does. And so it takes the signal and actually closes the loop. It, uh, reschedules the appointment, for example, it notifies the customer, it rebooks the load, all of that automatically.

Speaker B: It responds to emails.

Speaker D: Yes, yes.

Speaker B: I just turned it on for, um, my emails just now. And it was interesting. I was in there and the response was already drafted and I was like, how did that happen? And I didn't remember that I actually turned it on. I mean, I should have remembered that. Um, but it was really interesting because it did it without me even asking And I went into an email that I hadn't even opened yet, and the draft was already there.

Speaker D: Yeah. So that's. Again, many of these things are becoming more and more commonplace in what we do. But then, of course, uh, and we'll talk about this, uh, later, I guess, about how do you manage all of that control.

Speaker B: Yeah, it's so true. Now, before we dive any deeper, I do, just for the audience, let's indulge them for a minute. What is an AI agent and how is it different from a chatbot?

Speaker D: Yeah, it's a great question because, uh, you know, these get, These boundaries are blurring and people get confused all the time. So a chatbot, I would say, is primarily reactive. Right. So let's say you ask it something, it gives you an answer, and then it just stops. Um, it's more like information. But, uh, the fundamental difference of an agent is that actually it takes action. So you can give it a goal, and then it goes, figures out, utilizes the context that it has or asks you questions to gain more context, and then it looks, uh, at the data, reasons, and then it makes a decision. Um, that's the fundamental difference. Being able to take, uh, action, uh, based on the context, and then, uh, also bring the users into the loop in terms of sometimes it needs your input, and so then it takes that and then actions it. Um, so that's the basic, uh, difference. So I would say, like a chatbot, um, tells you information and agent helps you action it. And it's almost like a teammate, um, who actually gets the job done.

Speaker B: Well, and I also would say that it evolves with the data. The more data that you provide with it, the agent evolves, the chatbot does not.

Speaker D: Yeah, yeah, the agent, um, can learn from its past, um, experiences in terms of, like, you know, uh, what workflows it was able to execute correctly, uh, or the feedback you give it. Like saying, okay, I like this, I don't like that. Um, chatbot also has some of that capability. But fundamentally the difference is, uh, about action and closing the loop, which is, uh, a big difference, because otherwise companies are just awash in a lot of information and signals, but, uh, not really seeing the value, uh, in business terms.

Speaker C: Yeah.

Speaker B: Talking about feedback, I love giving an AI agent feedback over a human being. I mean, I'm just gonna put it out there because there's no judgment. You know, they're kind of like, oh, I see where you're coming from. Let me change this for you.

Speaker D: Right, right. Too polite. But then you have these, um, agents like or models like Grok, which. Which might talk back a little bit, but, uh, it's interesting. You know, the agents have different personal personalities, and models can have that too.

Speaker B: I won't sign up for that one. Let's just put it out there.

Speaker A: All right?

Speaker B: There's a lot of hype around AI right now. Like, I can't have any conversation, whether it's in person, whether it's on the podcast, whether it's my live show, without talking about AI right now. Like, a lot of organizations, they're shouting about it, but much of the data said that there's a real gap between the hype and the reality of it. What impact do you think. And talk to me about what AI washing is, because I don't even know what that is. What impact do you think AI washing is having on leaders and teams right now?

Speaker D: Yeah, yeah. Air washing is basically the skepticism and everything that comes along with everything being labeled as AI. So it's a real problem. So people are just slapping AI onto everything. And lots, uh, of times it could be just a chatbot or a dashboard, which is a shiny new label, right? And you call it like, AI powered this and AI powered that. Um, and the issue is that there's a lot of damage that it does to the, to, uh, people having belief in, uh, what the actual capabilities of AI are. For example, it breeds skepticism. Teams get burned by, uh, pilots, uh, that may have been labeled so and over promised. And then they're kind of cynical about all of it, right? And so then they're like, okay, now, once weren't, uh, twice shy, right? Um, so the budget could get wasted on a pilot that never made it into production. And sometimes in the, uh, industry, we call it pilot purgatory. Right? It just kind of stays there and it's kind of cooking, Cooking. But, uh, nothing really goes into production. And the worst part is the inaction that this leads. Leaders, uh, get burned once and then they just freeze. Um, or then there's like, uh, a lot of rounds of AI governance reviews and things like that, um, which basically hurts the company. So how do you cut through it? Honestly, it's basically, uh, by asking one question. So what action does the AI take? And show me the outcome, um, and show me the metrics that measure the outcome and not the demo. Real AI can answer that, but AI watched stuff cannot.

Speaker B: Right? That's how you know the difference. And it's interesting that that one question can really help allev a lot of this, right? Action, outcome metrics. But I think another really Interesting thing that you talked about was, um, you know, decision burnout, but I almost call it AI burnout.

Speaker D: Right.

Speaker B: Everybody's hearing about it, everybody's being pulled in different directions about it. There's different understandings around it. Um, content is saying different things. Right? Like the news and the content that you're consuming on a day to day basis, it's saying different things. It's almost like the fake news concept.

Speaker D: Right.

Speaker B: How do you muddle through all of that to make sure that you're making the right decision, you're making the right investments, et cetera, et cetera. And if there is AI washing out there, which you've clearly defined that there is, how do you wade through that noise? And I love the fact that it comes down to that one question. Is there anything else that you would suggest that they do to make sure that they're cutting through the noise and really getting the understanding of what it means for them and their teams?

Speaker D: Yeah. Uh, ultimately it's kind of going back to the same principle. You, um, look at what is the value? Like, you know, no matter what the, you know, the collateral or the marketing material may be telling you or the sales pitch is, look at the value that it actually brings. Um, and then just stay focused on it. Do not be confused by, okay, it's using this model or this new heuristic or whatever that is. Right. Um, so cutting away all that clutter and just focusing on, okay, I implemented this AI agent or AI agentic workflow, and then what did it deliver to me in hard business terms, um, if we keep that, then we can cut through a lot of the AI washing hype as well as a lot of slop. So, you know, it's easy for people to be creating a lot of, um, documents and material and demos and things like that using AI. But ultimately the question is like, you know, is it delivering value? And that's what, uh, you know, business, uh, leaders should stay focused on. And that's how you are able to still continue down the path of innovation and your business goals. Right. Otherwise you might be like put off by, uh, you know, let's say, like I said, a pilot which did not go well. Uh, but don't be discouraged. Just again, look for ones that are delivering value and then double down on those.

Speaker B: I have one other question for you because you mentioned governance. Should governance come first? Because I'll tell you, in thinking about AI and talking to some organizations that are talking about AI, a lot of companies are pulling back and going, we need to lay the foundation for the governance. How are we going to use it? What are we going to use it for? You know, what kind of LLMs do our, um, you know, the professionals, the team members, what can they have access to? What can they not use our data for? How do we keep the data all in one place so that it is usable and we can go into a, uh, relationship with somebody like for kites, with the basis of foundation to be able to create the value. What do you say to that? Does the government governance come first?

Speaker D: Um, in some ways it does, but there's a subtle or a clear way to do it, the uh, right way to do it, and a wrong way or a long way, uh, things that takes a lot of time. So the right way to think about governance is put in some very clear boundaries on what is absolutely off limits. Like for example, AI should not be, let's say automatically updating financial transactions or mission critical systems. Right. So keep certain things off limits and, or like control those. But in the other areas, right. I think if we put too much governance then it is going to kill innovation and slow down, um, you know, your business, uh, objectives. And the other, the thing to remember is that everybody's in a race. There are companies like, you know, if a shipper or a company is taking time in doing the innovation or utilizing AI, their competitors are going to definitely not wait. Right. So there are. So the key is, again, I could go on, but the key is to look at certain areas which have to be managed very tightly. But for others, put in your boundaries and let the teams, uh, innovate and do not be afraid of trying AI when it's not. Uh, the benefits are huge, uh, but you don't want to put too much governance on it.

Speaker B: Hmm, interesting point. I think we're going to talk about balance in just a little bit, so I'll leave it there for now. Now you've defined AI agents. Um, talk to us about what Persona based AI agents are. How are they different? Why are they a good fit for supply chain?

Speaker D: Right. So, uh, this was a deliberate design choice on our end. So our AI agents, they are all named and Persona based. For example, we have Tracy for track and trace, Sam for supplier collaboration, um, Alan for appointment management and so on. Right. So it was again, like I said, a uh, design, uh, choice. And so some other companies have gone the route of building like one giant monolithic AI brain. But we built a team, right, and each one is modeled on a real operational role, uh, with its own slightly different Persona. And uh, so we kind of give them names. Um, and why does that matter? Well, a few reasons. One is customers just are able to relate to them so much better. You know, what an agent does, and it's easy to recall. Okay, I know Tracy. For Track and Trace questions, I go to Tracy or Track and Trace. Uh, collaboration. Increasingly we're at a stage where we have different agents, um, are able to specialize better. Right. And just as we have different models, it's also increasingly that we need to put in more specialization. So then again, having different names and different Personas, uh, helps to, uh, kind of group or categorize those, um, capabilities better. And each one has a narrow job, a clear context, tight guardrails, making it easier to trust, to measure and to adopt. Right. And again, it builds familiarity. People get to know their agents, the, their, um, the nature, or what's it called, the Personas, how they work. And ultimately that we have seen drives adoption. And at the end of the day, the mental model is really that of a digital coworker. And that's the idea. So, uh, these are the digital co workers on your team, uh, such as an expert, um, at their thing. And then, you know, together they orchestrate and get the job done.

Speaker B: So do they actually look like people, I have to ask?

Speaker D: Yeah, we have our, um, agents. Um, you know, we create an avatar which is like, yeah, looks like a real person.

Speaker B: So this is really interesting. And I don't know if you know this, but recently on Thoughts and Coffee, I added, um, some team members and they are AI avatars. And so I have Camille, who introduces the guest on our show at the top of the show. And then I have Diane and Julian who are at the news desk and they actually give us a two minute rundown of the top news stories of the week. And so I have, without even realizing it, I've created a team of AI avatars that I'm utilizing within content. And, um, I think it's interesting that you've done the same with names and avatars. And I think it's crucial, I think, about, you know, change management being 80% of that digital transformation.

Speaker D: Exactly.

Speaker B: And how do we get, you know, our teams used to working with an AI agent? And I think you're right, I think, making them relatable. Yes, right. Making them, um, into a way that they could be a coworker, a digital coworker. And now that I've had time to work with our AI avatars on the show, it's almost like second nature. I don't really think about it anymore. And I'm like, I'm going to throw it to you. Diane and Julian over at the news desk and they take over for two minutes and uh, it just makes it more dynamic, more relatable and it brings AI in a way into the conversation in a very, very different way.

Speaker D: Mhm. Fully agree.

Speaker B: Love that. I love that. So autonomy is a big part of this conversation. What do we mean when we talk about AI agent autonomy? I think there's some fear behind this because are we expecting AI to act completely independently or what does that look like?

Speaker D: Yeah, yeah, it's important to get that right. And uh, we need agents to be autonomous, as you were saying, because that's ultimately where the value is. But autonomy cannot be sacrificed, um, or it cannot go, um, it has to be under control conditions and it's not all or nothing. Right. And it's really like a dialogue and not a switch. Um, and uh, it's about maintaining control and managing the agent while you let it do certain tasks that uh, you've kind of enabled it to execute. So the first thing is, um, define the scope and the boundaries and define what the agent can do and more importantly what it cannot do. Right. Because some companies, they take the view of, um, okay, here's the problem, just go figure it out autonomously. But the problem is that there are a lot of unknowns and uh, you hear about agents still making some hallucinations or taking certain decisions and if that is all not clear, uh, and then it just does things on its own without proper oversight, um, that's where things can go wrong. And so we want to make sure that humans, um, the managers of AI fully are in control. It's about managing policies, it's working within your, the governance limits, you know, making sure that there is limited liability to the company in terms of like, you know, what the agents are doing, um, and fully adhering to, like I said, the policies and managed. So when the agent does act autonomously, it is doing in a very bounded way with full explainability. And so you can kind of always see what it did and why it did what it did. Right. And be able to then provide it feedback like you're saying, um, so that it improves next time.

Speaker B: Well, and I want to ask you of, uh, both sides of the coin, are there best practices or a playbook around developing the correct level of autonomy or the boundaries that you're setting? And then on the flip side, are there red flags to look out for if organizations aren't seeing the right results with AI agent Autonomy. I know that's a big question because I'm asking you both sides, but I think it's important because, um, you talked about guardrails, you talked about boundaries. What are the best practices around that? I mean, we know human to human, what that looks like, but what about AI agents?

Speaker D: Yeah, there's definitely a playbook. And I would call it graduated autonomy. So you start with human in the loop. You measure the accuracy, um, the agent owns the trust, and then you widen the authority. Um, so bound the scope tightly like it's a clear task with a clear definition of success. Uh, building the guardrails and the escalation path from day one. And then close the loop, Tie every action the agent takes back to the real business outcome. So that's what I would call as a graduated autonomy. Now, uh, you mentioned red flags, like what to watch out for. Right? So a decision that the system cannot explain. I think that's something. If an agent takes an action and it's not able to explain, uh, satisfactorily that it has adhered to the policies, or, um, just simply there is no explanation given, then there's a problem. Right. And that's where you need to go in and tweak these settings that we have. Um, and also if there is no escalation path, so if you implement systems where, for mission critical items or at specific junctures, if there is no human escalation path, um, or if an agent is operating outside of its boundaries, uh, or the other thing could be that nobody is measuring the outcomes. These are some of the things where if we are not careful or uh, if those things are not being done in an AI implementation, then those are red flags that I would look for. Uh, point out, hey, you know, there's. This needs to be, uh, better managed and we are in danger of, uh, things getting out of hand.

Speaker B: So in, in an autonomous AI agent scenario, you just talked about operating outside the boundaries being a red flag. Now this might be a silly question, but does the AI agent tell you they've done that? No.

Speaker D: No.

Speaker B: Or you have to go and figure out whether they've done that or not and then sort of give them a, uh, you know, slap on the wrist type thing, because autonomously they should be able to figure out that they did something wrong and alert you. Right. Isn't that kind of the point?

Speaker D: That's true, but you know, it's not always the case and we have to put in guardrails. Right. Ah. So think of AI as a. Something that needs to be Actively managed. Um, right. And, uh, there are like, you would have heard cases in, um, the news where it said the AI told you that, hey, I didn't do this, uh, or the code is fine and I have not deleted anything, but it actually had deleted the code in production. So things like that can, um, happen. And if you ask the AI, it can give you a confident response that, no, everything is fine. So you need to have, um, in AI engineering parlance, it's called harnesses. So you have like, frameworks that check the work of the AI, sometimes even multiple. Like, you know, depending upon the mission criticality of the application, you may need to do multiple levels of checks to ensure that the AI is, uh, the work of the AI is validated and is per policy. So you cannot like trust the AI that the executioner of an action to fully say that, uh, okay, you know, let's say 90% or 99% of the cases, it's fine, but that one person could be a problem. And so you need other ways to check if that's really happening.

Speaker B: Could you just ask it and just

Speaker D: say, yeah, like I said in 99%, if you ask it, it will tell you, okay, here's what I have done, right? And 99% of the time that's fine, but that 1% could be a problem. And so you need to have like, other ways to ensure and cross check its work.

Speaker B: I get that. Okay. Sorry, I was just sort of thinking, because I'm thinking about how I'm sort of playing around with AI. Um, so let's get back to what we were talking about around balance. Right, we mentioned balance a little bit earlier. Um, how should organizations find the balance between people and technology? And what should some of that change management and AI trust building look like? We already talked about it in relation to AI avatars and how relatable they are and how that can really help with some change management. But what are the other things that we should be thinking about?

Speaker D: Yeah. So, um, in balance, and then how you deploy the line that I always come back to is replace the heroics, but not the humans.

Speaker B: Okay.

Speaker D: Like the agents, uh, take away the 2am firefighting and the repetitive action exception handling the grind. And it frees people up to work on things that they really care about. Right. Or they really want to work on, like the relationships, uh, the work that genuinely requires team thinking and team collaboration and the complex tasks. So those are the ones that people enjoy and want to do. And, uh, the tedium is what is best relegated to AI. So think of AI as digital coworkers and not replacements and honestly naming them Tracy, Sam, it really helps and people relate to it. Um, so that's the first part. And then on change management I'd say that one of the one painful visible. Uh, I'd say start with a uh, workflow where you have a real pain. Just try to get a quick win on it, which is obvious. Everybody agrees that this is something, it's a problem, um, and it needs to be controlled. So get a quick win on that and then make um, sure that the agent um, is able to explain what it did. And we have good metrics measuring it and that's how you show success and build trust. And then as you build trust, this gives you um, that flywheel effect to build on these small successes and then take it to uh, the organization rather than just saying, okay, we're going to change everything. Um, and it's like a huge change management thing. Plus there's a lot of if it's not proven and we have not measured the outcomes, then uh, there is more resistance later.

Speaker B: Well yeah, I mean somebody's going to be like, you want me to implement this but what is it going to do for me? And how can you actually prove that that's what it's going to do? And so you're right. I mean you do have to implement in steps, stages, um, and that's generally where you're going to see the success. So I really like those pieces of advice and how to find that balance now going to, you know, how to get started. Right. Because a lot of people are hesitant, a lot of organizations, teams, leaders are hesitant because they don't have some of this governance in place that we were talking about. But you said, you know, as long as it's not too tightly controlled because you want to get innovation. But how do you get started? Like what is something that I would just take action on today to get started with AI agents and maybe just to play around to really understand the capability of the technology.

Speaker D: Yeah. Again, this tried and tested method, just start small, uh, is what I would say. Pick one workflow, um, that is high volume, well defined and we know that there's a pain involved in keeping that or this human. Um, it's a repetitive task that needs to be improved upon. So uh, those kind of workflows are the best can candidates. Um, and then you measure it and then prove, and then you know, scale, rinse and repeat. Right. So that's, that's the typical, um, it's been tried in other areas And I think it holds true for AI as well. But then one other thing that people maybe holding back people is um, a myth, uh, that I really want to kill. So you do not need perfect data to start. Like you know, back in, when we were trying machine learning models, you needed a lot of good data, quality data, um, and training data to start. But in the case of uh, agentic AI, you do not need that. Agents are especially um, the generative AI based agents are really good with the messy data, you know, unstructured emails, documents, attachments, etc. And so you don't have to wait for perfect data. Um, and that's a big benefit.

Speaker B: Well, and I think one thing, I go back to what you said about that one question. Action outcome metrics. Maybe one of the places to really get started is to understand who you're working with or what potential is out there as far as the technology is concerned. And asking that one question, right, of those vendors and of those partners that you're talking to. Exactly as well. Okay, so talk to me about what success should look like if I'm thinking about implementing this. Um, you know, how do we measure it? How do we measure it for success? How do we measure it for the short term and the long term? I think that's one of the biggest questions I don't know about you, but that we get is what does success look like around this?

Speaker D: Yeah, it's a great question and a very important one. So I think of it in terms of three horizons. Um, so the short term it's operational. What percentage of exceptions resolve themselves, um, or by the agents? How many hours did you get back, Response times, deflection rate. So those are some of the things that you can set up and measure. So for example, uh, we see around 80% of customer service inquiries handled by the agents. And honestly, uh, I'm proud of the fact that we're able to autonomously handle about 85% of our shipment exceptions. Uh, that's a big um, uh, number that we have been able to achieve. So that's not, not it's a dashboard just lighting up, but it's an actual problem solved. Um, so those are the operational metrics that we look at in the short term and medium term. Um, we would look at bigger business KPIs, like for example on time delivery moving up by a couple percentage points, detention spent down by let's say 20%. Lower expedite costs, lower inventory carrying costs. So we are able to tie the agentic actions and workflows to business Outcomes. Right. And then measure those and then those take a little bit of time. You know, they don't happen like overnight, but after a few cycles, few weeks, few months, then you start seeing that trend. Um, and that, that's the, I uh, would say medium term measure, um, for agentic success. Yeah.

Speaker B: M. I really appreciate you sharing that because measuring success I feel like is the hardest KPI that everybody is trying to achieve today. Right. And rightfully so. I mean organizations need to make sure that the investments that they're making are really successful on the short term. But also the longer term, what does the future look like? How do we continuously improve on this and what does that mean for us as far as success? And I think getting clear on that vision is also really helpful as well in communicating that to partners so that they really understand what they're helping you to achieve. Now talking about looking to the future, what do you see as the next development in supply chain AI? Where are we going with this?

Speaker D: Yeah, so that's where uh, it starts to get really exciting. Um, and the arc, uh, kind of is right now we are looking at specialized um, agents which do different tasks. Right. But then we are m very rapidly moving towards multi agent orchestration where Alan and Tracy and Sam are all working together to orchestrate your business. Uh, and the future is towards a genuinely more and more autonomous and in many ways self healing supply chains. Right. Picture the agents being able to uh, so disruptions are a given. Disruptions are increasingly happening due to various factors around the world. Um, and companies can do little to control that. But how quickly you're able to respond to that with changes in strategies. And that's where the agentic uh, collaboration really comes in. Um, in this again we have the digital twins. And so the network data, it's reflected in the network data and then abstracted in the digital twins. So you'll see digital twins and agents kind of converge um, to simulate, to decide and to act. So the whole thing gets more and more proactive and predictive. So we've been talking about proactive and predictive, but um, uh, then how do you take action on top of it? And that's where the loop is closed with agents. Right. And um, the human role increasingly as I'm seeing it's shifting to being that of an orchestrator, uh, to that of being an exception strategist and focused on the high judgment activities and really managing a group of agents. So that's where I see us uh, heading towards, towards supply chain where parts of itself are increasingly self Running self managing, self healing. And the companies and the brands that lean in now, they're going to be competing, um, and coming out on top, um, and the others are going to feel the heat.

Speaker B: Yeah, feel the heat. And you said self healing supply chains, the first thing that came to mind is that you're going to alleviate the stress for supply chain leaders and supply chain professionals across the globe. Right. Like that was the first thing that came to mind. I almost feel like there's a collective sigh of relief.

Speaker D: You know, it is like we, we have been waiting for, um, you know, tools like what is possible today with AI because it's just the pace of change and the constant disruptions, they are not letting up. It's just only becoming more and more chaotic. And we need, uh, these kinds of tools and these kind of digital coworkers, uh, to be able to navigate this future.

Speaker B: We need some breathing room and we need to reduce that stress and burnout. This is great. What a great conversation. So organizations really want to make their AI investments. Like nobody can afford to lose time and money to AI washing. But plenty of leaders just aren't sure what type of AI might work for them, where to use it versus people, what success looks like, how to measure it. So I think we all need to be having more of these types of conversations, especially with our partners, if we really want AI to transform our supply chains from reactive cost centers to proactive competitive differentiators. Now, did you have a guest at today's Big Question? Well, at the top of the show, I asked you with Forkite's AI agents, what percentage of exceptions resolve themselves? And Jimmy talked about it in the actual episode. So I'm not going to ask you because you already said it. And what he said was that It's a huge 85%. So just imagine the amount of time that you could save so you can forget the routine and focus on the strategy. Well, if you need a partner to deliver AI agents that autonomously execute supply chain operations, like that 85% that I just mentioned, you want to head over to forkites.com now? I want to give a massive thanks to the team at Forkites for making this episode happen. Remember, everybody watching and listening, uh, to subscribe, rate and review the show, check out more content from the best and brightest over on letstalksupplychain.com we've got incredible guests. Innovative strategies, transformative solutions you cannot afford to miss. Now. Jimmy, thank you so much for joining me on the show today.

Speaker D: Thank you so much, Sarah.

Speaker B: Well, until next time, My name is Sarah Barnes Humphrey and this is let's Talk Supply Chain.

Speaker C: Did you know that the average cost of losing an hourly supply chain worker has reached $19,607 and that recent research shows that 77% of hourly supply chain workers are considering a job change in the next three months? This could have a huge impact on your productivity, bottom line and culture. Workstep is helping supply chain companies to better engage their distributed hourly workforce at scale, understand the true reasons behind their workforce turnover, and take actions to make positive changes and reduce attrition. Workstep has successfully helped many companies reduce their frontline worker turnover by up to 36%. Visit workstep.com to learn more. We have plenty more content for you featuring the best and the brightest in the industry. Head over to letstalksupplychain.com to check out the latest. And if you are looking for a solution to a supply chain challenge that you and your teams are struggling with right now, well, we have most likely had that solution on our show.

Speaker B: And you can listen to the episodes,

Speaker C: find out whether you're the ideal client, whether they're the right fit for you, and whether their solution is what you are looking for without even getting into their sales funnel. Head over to letstalksupplychain.com put that keyword into the search bar and all of that content will come up. And remember to come back next week for another unmissable episode of let's Talk Supply Chain. It will be packed with expertise, insider secrets, real industry stories, 2025 trend updates, and so, so much more. And remember, if you enjoy the show, there's a few ways you can support us. You can follow us on LinkedIn, Instagram, Facebook. We're also over on TikTok. Subscribe to our YouTube channel, let's talk Supply Chain, and you can you can subscribe to one of our three newsletters. Sarah Barnes Humphrey, um, myself has my own LinkedIn newsletter called the Monthly Pop. So head over to my page and subscribe to my newsletter. Let's Talk Supply chain has a LinkedIn newsletter as well, so you can stay on top of all of the live events that we have and all of the trending news as well. Plus, we have a let's Talk Supply Chain newsletter on our website at, uh, let's talk letstalksupplychain.com as well. And if you're looking for some really cool merch, we have a shop for you on letstalksupplychain.com full of hoodies. And tote bags and so much more. Plus, if you're looking for a community of like minded professionals, head over to secretsocietyofsupplychain.com to sign up for free. You are going to get exclusive access to some content that you have never seen before.

Speaker B: Because we haven't put it out in

Speaker C: public, we're also putting together virtual and in person networking events. We have our Woman in Supply Chain Connections group to bring women from the industry together. And then we also have our Supply Chain Marketing Master's program for marketing professionals in supply chain as well.

Speaker B: So something for everyone. Head over to Secret society of supply

Speaker C: chain.com and I can't wait to see you there.

Speaker B: A great week everyone. Thanks for listening. And remember, ship happens.