In this Florida Supply Chain Summit webinar, supply chain and technology leaders from Century Supply Chain Solutions, JAXPORT, Perry Ellis International, and Laneway.io discuss how artificial intelligence is transforming global logistics and supply chain management.
This session explores how organizations are using AI-driven supply chain technology to improve real-time visibility, strengthen operational resilience, optimize transportation planning, reduce costs, and enhance customer experience across increasingly complex global supply chains.
Featured speakers share practical insights into predictive decision-making, integrated supply chain data platforms, logistics automation, and the growing role of AI in improving agility and operational efficiency. The webinar also includes a live demonstration showcasing how AI tools are evolving to support faster, smarter supply chain operations.
As a technology-focused global 3PL, Century Supply Chain Solutions combines operational expertise with its proprietary VIZIV supply chain visibility platform to help organizations build more connected, resilient, and data-driven supply chains.
Featured Speakers
- Jeff Price — Marketing Director, JAXPORT
- Andrew Petrisin — CEO & Founder, Laneway.io
- Sandeep Baghel — SVP of Information Systems, Perry Ellis International
- Dave Oldam — VP of Technology, Century Supply Chain Solutions
- Jim McCullen — CTO, Century Supply Chain Solutions
Learn more about Century and VIZIV:
Transcript:
[00:00]
I’d like to introduce Jim McCullen.
He is from Century Supply Chain Solutions, and he is their Chief Technology Officer.
Welcome, Jim, and welcome panelists.
Thank you, Joanne.
So it’s great to be here again.
This is really a bit unique.
So Joanne approached us and asked if we could come back last year, not even a full year ago, eight months ago, we came together to talk about what each of us were seeing in our companies around AI.
And we were talking about a lot of starting components that were taking place in the logistics space, in the supply chain space.
And so today we’re back to talk to you about what we’ve seen happen in the last eight months and what’s so much different.
And I’m sure as all of the speed of change in AI right now is insane.
It’s, there’s a, constantly new components coming out, new pieces coming out, new
just new levels of capabilities.
And so the group that I’ve got here today, our focus is going to be to share with you what we’re seeing at our companies, what we’re doing to really
take advantage of this, of the advances in AI, specifically in the logistics and supply chain space.
And just quickly on Century.
So Century is a global logistics provider.
We’re focused on origin cargo management.
And if you’re not familiar with what our origin cargo management is, I just did a podcast a few weeks ago in the Logistics of Logistics podcast with Joe Lynch.
So if you go find that and Google it, we had a great discussion all about OCM.
So make sure you check that out.
So what I’d like to do first is just introduce the folks to the panel again, and then we’re going to get into a dialogue and share what’s happening within our company.
So, Jeff, do you want to just give us a quick intro?
Absolutely.
Thanks, Jim.
Thanks, Joanne, for organizing.
I’m Jeff Price, director of marketing at Jaxport and the Jacksonville Port Authority, of course, in Jacksonville, Florida, International Trade Seaport.
Good to be here.
Great.
Thanks, Jeff.
And Sandeep, and you’re on mute.
Hello, everyone.
Sandeep here from Perry Elllis International.
We are a global fashion house with multitude of brands.
I am the head of IT applications here, and I’m customers of Century.
Great.
And Andrew?
Good to be here.
Good to see everyone.
And thanks for everyone for joining the call.
I’m the founder and CEO of a startup called Laneway, focused on solving the allocation management problem.
Great.
And Gabe?
Hey everyone, Dave Oldham, Vice President Technology here at Century.
First off, thank you all for joining.
Yeah, I’m responsible for a lot of the AI initiatives here at Century, and the later part of this summit, I’m gonna run you through some real-world solutions, so thank you.
Yeah, so make sure to stick around.
Last year, we kept Dave as a secret till the end of the webinar.
This year, we’re bringing him up full steam, but he’s gonna share his screen and show you some of the stuff that he’s doing in AI today, so you definitely wanna stick around for that.
So, Jeff, I’m going to jump in and start with you.
So, last year we talked a bit and you were you were getting into building an AI research assistant.
You were creating some automated reports with AI.
So how’s it going?
what’s the world look like for you today at Jaxport?
Absolutely.
Well, last year, AI was very reactive for us.
It was a chat assistant.
It was something that I would go to, I would ask a question, I would get a response, and then do something with it, probably context switch to a different application to use that information.
And the switch for us has really been making AI more proactive in servicing insights.
So specifically, I’ve stood up a tiny AI stack on a Raspberry Pi, $300 with two $20 a month subscriptions for OpenAI and Anthropic.
And also threw in a little bit of software that I’ve spun up myself and open sourced.
And essentially what it does is you can just load it up with tasks and schedule tasks throughout the day and it’ll go step by step and accomplish those tasks and maybe get you 60, 70% of the way done.
And from there, then you can take that information, that structure or research or what have you, deliver that to maybe a more robust agent like Codex or Claude Code, finish the job and have all this work going on in parallel.
So it’s really, in my opinion, true multitasking at this point.
So Jeff, just a couple of quick, like why Raspberry Pi?
Like, what is it first off and why did you choose to do that?
Absolutely.
So Raspberry Pi is a very small computer.
It’s very cost effective.
And here at JAXPORT, we’re a government agency, so we’re always limited by budget, by headcount.
And so this was a very low cost, easy solution to stand up just on my own desk.
It’s isolated from our network, so it’s completely separate.
[00:05]
And from a cybersecurity perspective, it’s very safe.
So that was some of the thinking behind standing up that particular device.
And it’s also very capable.
So it has enough memory not to get all too technical for the crowd today, but it has enough memory to be able to do a variety of things and certainly fits our use cases there.
Excellent.
Well, and so this concept of having an agent perform functions, right?
I think that’s a big area that we’re hearing a lot about in the space, right?
So over the last year, we’ve seen aggressive use of chat tools, right, like ChatGPT and Copilot and
and Claude for chat purposes, right?
And now we’re really seeing this opportunity to deploy agents to perform work for you.
And so it sounds like, so that’s saving, would you say at this point it’s saving you time as an individual?
Have you seen it expand outside of your scope at JackSport?
Yeah, absolutely.
So our IT team has stood up in AI server as well.
And so they’re looking at AI applications on the IT front.
And, as head of marketing, we’re looking at marketing types of applications as well.
I might add, Jim, that, my first agent application just on my own time was actually, I spun that up in 2023 and they couldn’t do anything except thank each other for the work that they did not do.
And so with the advancement of models over time, the advancement of software and introduction of things like Aider and
and then eventually Claude Code and now Codex and Gemini CLI and all the others, you can really accomplish a lot.
And I’ll give a couple of concrete examples.
So if you’re working in supply chain data and trade flows and trying to understand trade flows into and out of the United States, you might be familiar with a product called Peers by S&P.
And so one of the things about Peers, very powerful, but the business name data is just not very clean.
And so we’ve always been hampered by
by that effort.
We don’t have a data scientist on staff.
We don’t have the budget to go out and hire one.
And so this digital worker, for instance, that’s sitting on my desk has allowed us to normalize business record names out of a couple of million business records across Jaxport and a few out-of-state ports so that we could finally compare on a company-by-company basis.
Are they growing their volumes in certain ways?
Are they declining in certain ways?
And then that changes the conversation that we have with representatives at the company.
Excellent.
Excellent.
And then from a training perspective, Jeff, you guys are a large company, so I’ve been talking to a lot of CTOs about how they’re going at training teams for AI.
And there’s companies setting up academies like AI academies.
I’ve heard others.
where AI has become part of an employee review process.
You’ve got to either spend a certain number of hours on AI or a week or month, or I’ve had somewhere, they’re challenged to, you’ve got to deliver a solution that makes your life easier, in the next 30 days.
And then others talking about the concept of builders, right?
And so that, in essence, everybody at that, that end user standpoint is in a position to become a builder, to create solutions to solve
problems for themselves.
Right.
And so what are you seeing?
Well, I first started with my own our own marketing team a few years ago, actually, and showed them a video on YouTube by an entrepreneur who had challenged his own company to replace themselves with AI.
And so it’s a very provocative video.
And of course, it taps into everybody’s fears about, job loss with AI.
But nobody was replaced in that video.
And that was certainly something that I mentioned upfront saying, look, your job is safe.
We just want to do things better and have an open experiment to be able to try to improve our productivity, improve our efforts so that we can work on behalf of the people of Jacksonville and get, more freight coming through our facilities.
So that was very inspiring.
And what we found is that as people started to use AI a bit more, become more comfortable with it, they started to share the successes that they had.
And that then incentivized additional momentum behind that.
So that was on the small scale.
Now, going across the enterprise, I’m part of a group with our HR director, IT director, and chief of regulatory compliance.
who’s put together a policy for JAXPORT that gives those guardrails and identifies those things that AI should be used for here at this organization that works for us and our values.
So it lays out, the do’s and do nots.
We’re not giving vital systems to AI.
We’re not making any personnel decisions based on AI.
So those sensitive types of things remain human-centric.
And AI can help us with those types of tasks that we have in our job just to be more effective and more productive.
[00:10]
Excellent.
That’s great, Jeff.
And Sandeep, I want to jump over you.
And by the way, the rest of you guys too, if you have questions as we’re going through this, please chime in.
I’ve got one or two questions up on the screen already that I’ll bring into the mix too.
So Sandeep, last time we were talking to you, you were
rolling out Gemini across the organization.
You were, I know you were talking about creating some videos, some marketing videos using AI.
Want to understand where you’re, where you’re at today, what you’re seeing.
But also one of the questions that came up was, how do you decide what technology
you’re going to use, right?
And so I’d love to understand, Sandeep, why you went with Gemini just before that from Century’s perspective.
We’re mixed right now.
We’re using Copilot, ChatGPT, and Claude for different reasons right now.
So Copilot is super powerful for us for language translation.
So as a global company, that works really well for us across the world.
Challenges in some of the countries like China are limits that you can’t use some of these tools like ChatGPT and Claude.
So we have to manage from that standpoint.
And then Claude, we’re finding extremely powerful from a developer perspective.
And then ChatGPT has just worked really well from a marketing, sales, customer service side of things.
So for us, it’s been a mix.
But Sandeep, maybe you could start with that.
You’re a Gemini shop, maybe just talk a little bit about that and then talk about some of the things you’re doing.
Definitely.
So last time when we met, it was I think eight months ago.
But eight months ago in AI world is like an eons.
Like we have so many development happening all over the world from the AI perspective and things are changing so fast.
Sometimes it’s difficult to keep pace with how every day you’re getting new stuff in AI.
For us, choosing Gemini as our enterprise level AI was very easy because we have been a Google workspace for the last 10 plus years.
And Gemini, if you look at Arena leaderboard in the AI like where there’s a measure of all the major LLMs.
Gemini, Claude, ChatGPT, they’re all, they are sometimes one guy will
go on top of another, sometimes another one.
And yes, there are going to be specialization slowly.
Right now, all these agents like the big LLMs are still more generic.
And for our use case, yes, people, a lot of software engineers love Claude Code.
But what we found is for the applications we develop in-house,
Gemini is good enough for us, and then it makes it easier for us to maintain our contractual work in one place.
And not only that, for our employees to learn, we are all talking in the same language.
So we continue to use Gemini, and then we deployed Gemini to all of our employees.
12 months ago, we started with our top level leaders to educate them on AI and what AI
brings to the table.
However, we wanted to make sure that the rest of the folks in the employee world also get into the bandwagon.
So what we did was we started an AI task force with the leaders from the business and my IT leaders in the together.
We developed little courses for our team.
We created a news portal and then we had twice every year, twice every
week we are sending email newsletters about what is coming up in the AI world, how we are using it.
We did gamification on our end so that to make people enthusiastic.
Sometimes when we just talk AI, it becomes like an abstract layer.
And not only that, people get scared that their job will be impacted.
So by continuously communicating, not only through
emails, but directly meeting different teams which are global in our organization, physically going to the different places.
We allay that fear is not about jobs, it’s about everybody learning.
And if all the employees can be more productive, we’ll have better revenue, better cost management.
So selling like that actually helped.
And not only that, when we are not talking just in abstract, we are providing a use cases.
Okay, so you are a marketing design person, like you want to do marketing campaign, look how you can do it using Gemini and some other third party tool we purchased.
So when people saw that and saw that that actually is making them better, their work life better, people really started embracing very fast.
[00:15]
And we are very happy with the progress we have made social.
And Sandeep, you mentioned, the communication.
I think that’s so critically important, right?
Like to be able to share these successes across the organization, find a channel, find a way to do that, but also to document.
And one of the questions came up, too, like, how do you determine the ROI on an AI agent trade?
and so, are you doing any measures of KPIs or ROI around that side of it?
But before you answer that, I think, for everybody listening today, right, however you’re using or taking advantage of these AI tools, it’s so important that you that you start to think about how can I communicate the value of this tool to the organization, right?
Because whether it’s your board of directors or your direct manager or your customers, I just finished working on a deck for a customer presentation today.
Everybody just wants to know what do we.
Yep, Jim.
So yes.
So from going from like how do we maintain the ROI in the discussion?
In our organization, what we are doing is, so let’s say Gemini is our corporate AI tool, right?
So we open it for all the employees deliberately and wanted them to use it.
Costing or adding value.
Yeah, so instead of looking at how it will go through, the employees started experimenting.
And in last couple of weeks, what we found was few of the employees are hitting our quotas.
Like we have quotas like each employee get this many number of credits in Gemini.
So that’s when we are seeing that more and more people are using and then we are okay asking what you’re doing, does it really impact?
How is the ROI on that generic use cases?
Then any other application where we are doing beyond Gemini, also, we buy different AI applications, like for design or marketing.
We are going through the full process.
If it’s making my employees more efficient, it’s giving them, let’s say, 20 hours or eight hours a week,
what exactly they can do in that eight hours.
And that is monetized.
I know that’s the most difficult thing right now.
It’s like, how do you, when you’re spending this much money, how do you justify that cost?
Because you can’t keep spending money.
We are not like tech companies like Google, OpenAIs of the world who are getting hundred billions of dollars, right?
So we are going through that ROI benchmarking.
We are asking business now that we have taught them and they understand how it all works.
Think how much it will save and what that savings will entail.
Will it increase my revenue?
Will it decrease my cost?
And if that tool’s cost is less than that, then we are moving ahead.
Otherwise, we keep looking for opportunities.
Excellent.
That makes a lot of sense.
Thanks.
For sure.
Hey, Dave, I got a question for you that’s pretty strong here that I want to send out to you, and then I’ll come back over to Andrew next.
So here you go, Dave.
You ready for this one?
Hello.
I don’t consider these LLMs AI.
If this is an AI session discussion on logistics, what are we supposed to learn here?
Anyone in any company of any industry in size can use Claude code.
My 15-year-old kid can use these tools.
What do you got to say to that, Dave?
Yeah, I’m excited to get started here.
I think I’m at the the end of discussion.
But there is a there is a major difference in using an LLM.
To rewrite an email for you.
So you sound professional to your boss.
I do it every day, right?
But what we’re doing here is centuries.
We’re basically saying, how can I use an LLM.
to replace the work that I do, so I manage the LLM.
So a perfect use case example for us, we have what we call our AI-driven routing engine.
So we at Century have the responsibility of taking a vendor booking and turning it into a confirmed carrier booking with a ship order release.
There is so much work and institutional knowledge and standard operating procedures and routing guides and failed, service strings and blank sailings and equipment shortages.
How do I, as an operator, take a vendor booking and get it to a carrier booking to follow my customer’s SOP?
Well, in previous years, there was a lot of institutional knowledge to make that happen.
It would take me the entire day.
With Century’s AI routing engine, in about five seconds, I can take about 200 bookings and follow every rule to the finest degree and come up with a perfect plan.
So what does it mean for our customers when I take it off of the person’s desk and I put it into the agent, the agent builds a plan, gives it to the final person.
They either prove or decline it and move forward.
[00:20]
One, that you’re controlling cost, you’re getting on the best vessel per your port pair, destination, whatever type of mix, and you’re beating out your competitors because you’re first to the carrier and are becoming a shipper of choice.
So just an example of how, yes, there is an LLM involved, but there’s tons of what we call skills.
So I give the LLM a pre-trained skill set with very, very good data and I’m executing at speed.
So that’s the difference of again, reformatting or rewording an email for your boss to actually integrating it into an operational process.
Yeah, and I want to build on that a little bit more, too.
And I think the the skills documents are are critical, right?
When we think about across logistics, right?
Like the
I know Eric Johnson from the JOC was talking about this, right?
Like, the AI technologies are finally catching up to what we need in logistics as a whole, right?
And so much of logistics is, yeah, you have these bulk processes where it’s like, okay, we have to book and route, tens of thousands of trucks or whatever, right?
But then you have all this really custom work for customer supply chains, especially the things that we see at Century Two.
where each one is unique, right?
And a lot of that intelligence exists in people’s heads.
today, right?
And so these skill documents are allowing us to to document those processes.
So agents can perform them right and to your point, Dave, perform them at speed where, in the past you might be able to do a routing once a week.
You just didn’t have enough time, whereas now you can do it on demand as often as you need to.
And you can come back later and analyze everything that you did and run scenarios against all of that data, right?
All of that exhaust data from, from the AI process.
So, so yeah, certainly changing.
And, another tie into that question came back to asking around another question around,
.
What about specific logistics, AI companies?
And are you all getting bombarded with companies, with AI logistics solutions.
I’m gonna open that up to the table.
But from my perspective.
what I’m seeing is very targeted solutions that are out there.
It might be something related to tariffs, for example, right?
Like obviously you can layer a lot of AI around tariff decision processes.
It’s massive amounts of data that has to be analyzed and decisions made, right?
But then I’m also seeing a lot of workflow applications out there where maybe a year ago or so, those were pretty powerful.
But now as you look at them, you’re thinking, okay, yeah, I could use Claude Coworks or something else to do exactly that, right?
have to factor in like, okay, is the intelligence that that company is bringing of logistics and, valuable enough to you, or do you feel like you can do it yourself?
I don’t know.
Any other comments on that from the group?
Yeah, from the logistics.
So someone has correctly said, LLMs are not necessarily just an AI.
So the way we started using three, four years when ChatGPT come into the picture, right?
It’s usually chat bots.
Then what happened last year, we are creating on top of those chatbots agents.
Now where people are going is agentic workflow.
So let’s say if we are just talking about logistics, I have shipping.
So we are a wholesaler.
We are shipping it to multiple retailers.
And each retailer had their unique SOPs for how I want to ship my shipments, basically, products, right?
So for those SOPs, we need to make sure that our warehouses are following that.
So we put those SOPs in Notebook LM.
So earlier they have to go through a document and figure it out what exactly for my specific retailer I have to put my labels as, right?
Right now, and they used to wait for somebody to tell them or read through the long document.
If I have that in Notebook LM, they can cushion, again, it’s still a chat, but the idea, the next step is,
How can I have my warehouse management system integrate with live, looking at those documents, and accordingly making sure that my label is getting printed correctly?
So these are some of the use cases we can use.
And like from Century’s world, when we are getting products from our overseas Far East factories.
So we have different factories, and we want to make sure that they are all following same quality control.
And we were providing our quality documents, different specifications.
How do we measure that?
So that’s where we are using artificial intelligence.
at the core of the brain, yes, these are the few LLMs we are using, but how do you orchestrate that to get a workflow to work for you?
That’s where the bread is, butter is basically burning.
[00:25]
Yeah, the LLM is a tool and it’s how you apply that tool is really what it’s coming down to now, right?
So Andrew, the question came up again, specific logistics, AI companies, logistics, right?
So I think good opportunity for you to give us an update.
So last year when we talked to you, you had the concept together.
You were out talking about Laneway.
Maybe again, you can just re-explain Laneway just quickly for everybody on the call today.
you were looking for investors.
You had a mocked up solution.
I think you built in Lovable or something like that.
Can you give us an update of where you are today and how AI is impacting all of that?
Yeah, glad to.
I’m going to talk about the pilot here in a second.
We’re launching this summer.
But I want to address a couple of the questions and I hear from the chat.
I will say, for the attendees who are chatting,
We’re definitely not an LLM at Laneway.
And I think we actually have a really unique angle in this AI world within logistics.
And I’ll give you a bit of a hint at it and then walk through it a bit.
as that manifest this February, and I had a big forwarder come to me and say, there’s, we had talked to what Laneway was doing and was like, there’s, I hear so many different, startups doing agentic workflows.
We need more people who are changing business models like you are, right?
So really, I think we’re carving out a really unique space as an enabler, right, for AI adoption within the industry.
And, I’ll talk specifically what that looks like.
So what we’re building actually is a marketplace for allocation.
So the way I see the world is
you can build these agentic workflows end to end, adjusting field data, automated bookings, and the carriers are adopting more and more APIs.
But at the end of the day, securing capacity is still done over the counter, right?
You have an NPC or 52 that I think everyone knows is imperfect.
And, I’ve talked to, major, AI companies and the startups working in logistics and say, we’ll do all this work.
We send in the booking and we hope we get a confirmation, right?
We’re stuck, right?
And I think, as much as you want to like build out and automate all the workflows, before that and after that, you have this point in logistics where at the end of the day, we’re matching supply and demand, right?
And that crucial point is still, in my estimation, like done manually, right?
And the way we see the problem is because allocation is not a unit, right, that you can buy and sell.
And until it becomes a unit you can buy and sell, you can’t actually automate the process end to end.
So we’ve coined what we call allocation equivalent units or AEUs, which is a bit of a fun play on the TEU.
And what we’re doing is essentially having carriers sell allocation, right, to customers.
So if you think about your service contract today,
The way we see it is actually, there’s actually two costs in that contract, right?
There’s the operational cost for them to move your TU, that’s pretty fixed.
And there’s actually an allocation cost, right, which is a scarcity cost on a given week.
It’s really a variable cost of supply and demand.
So what we’re doing with this pilot this summer is actually breaking those two costs out.
So you as a shipper or a contract holder basically have a lower base rate and you pay that when you ship.
And then the carrier will just sell allocation.
So you’re not bound by NPC-52.
You can buy flexibly when you want, and you can resell it back to someone if you don’t need it anymore, and they’ll attach it to their contract.
So what we’re doing is taking what I think ultimately is still the manual linchpin of a lot of ocean logistics in particular.
But I’ve talked to folks in domestic trucking.
I’ve talked to folks everywhere who have this same problem of the contract breaking up and down.
and turn it into something that I think actually can be used by by AI agents.
And I think that’s something that’s really exciting for us.
And, maybe a an analogy, for folks who have used it sounds like a lot of folks on the call have used, cloud, Claude Code, Claude Code, or other tools is, everything becomes so much more useful when you have like a standardized plugin, like an MCP, right?
And
We see that in some ways is like when you can standardize unit of allocation, then you can manage it much better.
You’re essentially structuring data to allow it to be used in processes.
So a little over-explained potentially, but there’s a bunch of questions on what we do.
So I wanted to give it a thorough opening, if you will.
Yeah, and I think, too, the concept that MCP servers allowing Agentic AI agents to be able to talk to each other, like we all know.
[00:30]
in in the logistics space.
We’re still buried in EDI interfaces and, lots of old formats and structures and everything else.
Right?
So so, certainly that is something that opens up new opportunities in the logistics space that, that we haven’t really had in the past.
Yeah.
And I and I’ll say, too, we like, I think I see us as
really an enabling technology for AI.
Like we obviously use our own AI tools internally, like folks in the chat have said, one thing I’ve built personally is our own plugin called Hive Mind that basically pulls from MCP servers from our notes, from our CRM, from our product roadmap, from our emails, from our Google calendars, and essentially is a a personal assistant that is like a dynamic wiki for the company, right?
That
Helps me as a founder on a small team to keep track of everything and to make sure we’re following up with everything.
So we use these tools too in the same way that I think folks in the chat were saying is, I think that’s what people are getting at when they say AI native company, right?
But we really, I see us both ways as both using the tools to build out our own products, but also really as a fundamental enabling layer for agentic workflows in a logistics space.
And David, David, Jeff, you guys have any other comments on that? Jim, hearing all the incredible use cases from our panelists today, for me, it’s stepping back for a moment.
It’s all about mindset. what’s your biggest challenge today?
Right?
And then you have to ask, why?
Why is that still a challenge?
we’ve had the world’s knowledge with Google for two decades.
We now have super intelligence in our pockets.
It doesn’t take a lot of money to solve it.
It just takes a decision to solve that.
So whatever your biggest challenge is, you can solve that.
So that’s what I would offer any logistics director that’s looking to use an AI solution, think about your biggest challenge and then go solve it.
Yeah, and it’s a valid point, you’re right.
And I think there was another question up there, I can’t see it all the way down on my screen, but it’s just saying like, okay, if Century’s using LL,
How does that differentiate you from anybody else that’s using LLMs, right?
And, I really think, again, it’s a tool and it’s how you apply the tool and how you use it, right?
So, for example, if we’re using, anybody can use Claude code, right?
But if we’re using Claude code more efficiently to be able to build customer integrate, to, evaluate design and build customer integrations and get customers onboarded faster and, flow data through systems easier, well, that’s an advantage to us, right?
And could other people do that?
Sure, absolutely, right?
But it takes that combined human effort of expertise and everything else to bring all of that together.
And then when you talk about the general productivity, okay, yeah, there are simple productivity tools for individuals, but if we can collectively make our teams more productive, then we can be more competitive from a pricing perspective, right?
And as Dave said earlier on the routing piece, right?
So if we can route shipments faster than somebody else can, because we’ve figured out a unique way to use these tools, then
We’re putting our customer at an advantage, right?
Because when you’re in a peak period and everybody’s trying to get capacity at the same time, if I can book shipments faster than a competitor can, I’m giving a true advantage to that customer, to our customers, right?
So yeah, it’s a set of tools.
It’s how you deploy them.
I think certainly in the logistics space because of the many, many hundreds and thousands of providers that we have to interact with in so many different formats and everything else that AI really just helps
helps there.
So maybe I’ll turn it over to Dave.
Dave, do you want to chat a bit and show us some stuff?
So first of all, Jim, a bit of a thumbs up if you can see my screen, the digital dashboard.
What is the adoption?
Oh, so I’m going to get to a couple of these questions for sure, I think.
Go for it.
So yeah, so you can see my screen here.
Yep.
Okay.
So while that’s in the background, let’s talk about some of these questions.
And originally, I was tasked with coming to this discussion to show something flashy in AI.
And when you start looking at the participants, you really have a mixed bag, right?
You have people who
who are just getting started, and then you’re getting some pretty in-depth conversations about adoption rate of OpenClaw and LLM as being a table stake and so on.
So I would say I’m going to try and go somewhere in the middle, not get too advanced and lose a few people, but also make it interesting.
You can find me at LinkedIn, David Oldham.
You can reach out to me, and I would love to continue conversations with anybody on this call.
So everybody,
Is everybody asking the same question?
an LLM is not AI.
That’s a fair statement.
It is AI, but it’s been beat to death and people are using it for different purposes.
[00:35]
Other questions around, token usage and how do you control that?
And just, how is this going to change the space within the logistics industry?
And we talked about, our our tool, the AI driven routing engine.
we we had just onboarded a customer where their entire logistics group would sit in a room every Friday and manually go over carrier allocation, space, routing, optimization and opportunities.
Those people are no longer in that room anymore.
They now get a report, they review it for 30 minutes, and they move on, right?
There is an LLL behind it, but there’s also tons of skills documents, there’s tons of industry data, there’s disruption algorithms and everything else that feed into an engine to route freight way better than a human would ever do it, okay?
But let’s go back to the beginning of our journey.
So when we met eight months ago, we talked about a chatbot.
We have a chatbot.
Everybody has a chatbot now.
The industry does not need another chatbot.
But I’m going to start there just to talk about the evolution of it.
There’s a lot of conversation out there as build versus buy.
Are you going to build your AI or are you going to connect with a Palantir or someone like that and buy the ability to build agents?
What you’re looking at here is the next iteration of a chatbot.
So we have, a hybrid approach of build versus buy.
So one of the buys we have is snowflake.
Snowflake is our cloud-based data warehouse, and it is the foundational layer for the data that feeds into our agentic AI processes.
Snowflake has what’s called snowflake intelligence.
Snowflake Intelligence is now going to take a chatbot to the next level.
So you can see I asked a very simple question such as, tell me about my carriers.
And before I would just get back a grid of my carriers.
But now there is a thought process and you can see how the agent is actually working.
And it does more than just bring back a grid.
Yes, it brings back a grid of all my carriers, what they’ve shipped.
That’s helpful.
I now have the ability to dynamically generate some charts, but I start bringing in key insights.
Hapac Lloyd is your largest carrier.
dominating value despite the 194 containers indicating high container realization.
Right?
I’ve asked a question such as, is this purchase order arriving on time?
And previous years, you’d get back a yes or no purchase orders arriving on time or not.
But now I can tell you the vessel is not arriving on time, but the purchase order is arriving on time.
Here is when it’s due to arrive.
And then, business impacts and recommendations that may be coming out of a customer’s SOP.
Analyze year to date.
Go ahead.
Just also to expand a bit.
So like, okay, yes, this is using an LLM to give that intelligence back, but the importance of the underlying data, right?
Both on the routing engine and vessel schedules and what’s happening here, right?
Like, I think that’s another differentiator.
If your logistics provider doesn’t have the level and quality and completeness of data, it doesn’t matter what you do in the LLM in some cases.
That’s fair.
and when we talk about Snowflake, Snowflake was our investment years ago, knowing that this was coming.
We need to get our data and stage it so that it is actually useful, right?
And then you layer the intelligence on it, and you get an intelligent chatbot with, key findings, recommendations, and so on.
Okay, so that’s the next iteration of the chatbot, which I think most people are comfortable with.
Then we said, okay,
We’re now using Claude code like many others, right?
What are we doing with Claude code?
Well, our most extreme example is that now our engineering group or our development group no longer starts from ground zero.
So as our product team puts a request into our ticketing or tracking solution, we use Jira.
So product team goes out and says, Hey, we got a new customer.
We need to enhance the product in this way.
We actually have…
an autonomous agent that’s running Claude code that sits on a machine.
It actually picks up that JIRA task.
It looks at all the skills that it has.
It looks at all the code base.
It connects with the data model and it actually writes the code for the change request, puts it into a pull request and gives it to a developer.
Now when a developer opens up the request for a change, the Claude code agent has actually already generated the change.
They review it.
Maybe they tweak a little bit because it’s not ever going to get to 100% in today’s world, but maybe got them to 90%.
They make a quick change.
They deploy that code to production.
We’re doing that for tickets.
We’re doing it for tasks.
We’re doing it globally.
You no longer just get a request from a project manager.
You get a request from a project manager where Claude code has made its best effort to actually solve that problem.
So I talked about clawed code.
Some of the people on this call may not know what clawed code is.
So let’s actually go in and do something with it.
So I’m looking at, this is our Viza product.
I have the ability to go into the supply chain map.
And for this customer, what we’re looking at are all of the vessels that they are currently on.
Basically, this is in real time.
As of an hour ago, this customer has one container on this vessel, and it is going to get to Los Angeles more than 10 days late.
[00:40]
So I want to view the data.
I now have all the data, right?
So what I’ve done is I’ve taken that data, I’ve exported it to Excel.
So this is all the data behind that screen, right?
I want to build some dashboards or some insights off this data.
So what I did is I went into Claude Code.
I went into Claude Code, and let me get my little prompt here.
Give me one second.
So I said,
I am a logistics expert, and I am providing the file containers.csv, which contain all the ocean containers on the water.
Can you review this data and give me some insights?
So I didn’t give it very much.
But what I did give it behind the scenes was I gave it some skills.
So inside Claude Code, you have the ability to create skills.
You want to spend a lot of time here.
This is where you take your institutional knowledge and you bake it into the process so that when the LLM runs, it runs the way you want it to.
You can connect to your apps.
And you can create skills, right?
So maybe you want to connect to Office 365, pull down all your emails, and then do something against those emails.
But I have an Excel document, and I have some skills.
So if I go back, my first prompt I gave it was, again, I’m the logistics expert.
You have this CSV file.
Can you generate some insights?
And it says, oh, you got 3,800 rows.
What can I do with it?
It’ll give you some insights such as, you had 346 containers, 600 TUs, $42 million worth of goods on the water, and 68% on time.
Now, it generated this workbook for me, okay?
So if I go back in, again, I gave it containers, I now have container insights.
So it took, in about 10 seconds, it took this document and it converted it to this.
So multiple tabs.
with different things that it knows I’m going to want based on my skills of a logistics expert.
So the first tab.
And it was just from that one prompt you gave it, Dave.
2 sentences.
That’s all I gave it.
So I gave it an Excel document with 38,000 rows, and it came back and it says, okay, first off, executive summary.
You got 348, 46 containers in transit.
$42 million worth of goods, 67% on time, average three days late.
I’ve now got some different findings and things to watch out for, such as you’re heavily loaded on one carrier, they’re habitually late, and so on.
But it then goes further to break it down.
Carrier performance, it generated all this for me, right?
So how are my carriers performing at the carrier level?
How are they performing at the vessel level?
I now have insights on origin ports, discharge ports,
trade lanes by status and size, and then container details.
So with just two sentences, I’ve taken that Excel document and I’ve generated this, okay?
So maybe that’s super impressive for some people.
Maybe you’re already doing that.
So then I said, well, what?
I don’t want to work in Excel.
Excel doesn’t feel professional for me.
I need to send a report to somebody.
So I went back in the Claude and I said, hey, this looks great, but I’m more of a visual person.
Using the same CSV file, which contains all the information, I want to build a modern looking website.
And I want to have drill down.
so I can show it to my boss.
So I hit that, about 20 or 30 seconds later, I got this, oops, this one.
So utilizing that Excel document, Claude Code built this for me.
It’s just an HTML file and I can email it to anybody in the organization, right?
Maybe the buying group says, I need to know where my container is.
I’m like, Hey, I got you one better.
Here’s every container that’s currently in transit.
So this file says, Hey, you got 346 containers.
Currently active, 68% on time, 109 late vessels.
Two of them are critically late, and you got 87 vessels on time.
How many are at origin?
How many are at feeder port?
274 containers on the final leg, 13 are currently arriving.
Scroll down on time versus late.
How late is very late?
A mixture by discharge port.
mixture, well, the ability to drill down.
And then I have all of the carriers.
So I can say, well, let’s for a second imagine someone says, where is this container?
I come up here.
It built in a search.
I put in that container.
Well, it looks like it’s delayed between one to 3 days.
It’s going to Long Beach.
it’s on rim Logistics it’s on this Indian Ocean vessel I can click on it and there’s one PO and item inside of it and I have a bit of more of a detailed page so you can see you didn’t build any of this this was Claude built all of this from that second request all I did was told Claude you have a CSV file I’m a logistics manager I want you to build a nice looking dashboard and it literally took 5 seconds and did exactly that oh that’s
And somebody asked a question, and they asked, Viziv is Century’s platform for global supply chain operations, customer visibility, vendor management.
[00:45]
But they also asked, is Claude code populating and creating everything in Viziv today?
Which is no.
No.
We treat the AI like our superpower, right?
Like, I know how to write code.
I’ve been writing code in Viziv for 30 years here or 25 years here at Century.
But I now have an army of engineers behind me.
I’ve taken all this knowledge that I have as my skill.
I’ve put it into Claude and I said, Run.
So now when tickets come in, they just come to me.
And it’s like, Hey, someone wants to update this data.
Fine, here’s the update statement for the database.
Someone wants to put a button on the screen.
To me, I’m thinking it’s the death of the dashboard.
So you probably have teams of people and analysts who do this with two sentences, I’ve built this.
Yeah, so, like, Dave, and showing snowflake and what’s possible there and this and what’s possible here, I envision in the future that we still need this .
transactional source of truth from an operational standpoint, right?
Which, is our visit platform as well.
But the more that Century and other companies can enable their users to be able to use these tools to create what they want on the fly.
You demonstrated that earlier in Snowflake.
I don’t think this is honestly that far away where like a customer could ask something in the platform.
The platform can go behind a cloud and say, produce this dashboard for the customer interactive dashboard, right, in seconds.
which is, something that would, to your point, really just do away with any of the static dashboard needs that we’ve seen in the past.
But not only that, if I go back to what Sanzeep and Jeff were talking about, that’s a reactive world, right?
With the intelligence, we’re consistently scanning the data and saying, I think you have a problem, generate the response and push it to the user so that when you log in, like when you log into Viziv, you start off at what’s called my day.
And the AI pushes problems to your day.
It’s not you going out and trying to find them.
I have a shipment that’s missing a document, right?
I have 94 POs that have not yet been booked.
So historically, those would be developers who are writing those codes to push that to a user.
The AI starts looking at pattern recognition and disruptions and it’ll push it to the end user.
Hey, Dave, another question came up, AI tools often hallucinate, how do you manage
keeping that from happening in these, dashboards and Snowflake and things like that?
It’s a very fair question.
Quite honestly, it’s a challenge.
different models have different levels of success to Sonnet versus Opus.
Your skills come into play, your checks and balances come into play of, comparing it to real data sets.
But then we always have what we call the human in the loop, right?
we’re not at the point where we’re moving our customers freight autonomously.
Like I said before, we’re giving our operators the superpower and we’re allowing them to execute at speed by taking away all the noise that they once had to do.
Yeah.
So again, it’s a human in the loop.
We’re not building code directly.
Sorry, go ahead.
To that like hallucination point of view, the more detailed and questions you will ask with more guardrails, the
there will be less chances of hallucination.
It hallucinates when you ask a generic question because it fills in for that generic thing.
Yeah, that’s a good point.
The skills are going to be key.
Correct.
We’ve invested more time and skills than we do on anything else here.
Thanks, Dave.
That was great.
Good stuff sharing.
Hey, Andrew, there was another question for you.
Do you want to just look at it there?
I, yeah, I’ve been, I’ve been trying to answer them in the chat as you pulled them up, but happy to speak to it as well.
Yeah, do you want to just talk about it a little bit?
Yeah, yeah.
So there’s a couple of questions about who’s the ICP, the problem you’re solving for, as well as binding contracts.
So, and again, who’s participating.
So we do have
major carrier and multiple shippers and forwarders who are working with us on a pilot this summer.
the top line is, I think, actually, I sometimes, I’d say the tongue in cheek, I think shippers pay too much and I think carriers don’t make enough, right?
And that’s because there’s just such an excessive amount of efficiency in the system because of how allocation is managed.
So, you
to think about, you’ve signed this contract, let’s say for $1,700 at the base rate, right?
What you’re doing implicitly is essentially pricing in the cost of the allocation.
So you’re buying basically MQCR52 as a product.
So let’s say the cost of the operation, the operational cost, the cost around the service is 1100, the cost of the allocation, basically you’re fixing at 600 a week, or like 600 essentially when you go and book every time, every week.
[00:50]
independent of whether the cost of the space actually is $600.
So I think often you’re probably overpaying, and then when you’re underpaying, you basically get bumped.
So that’s like the MP352 problem.
I think, I think there’s a question about like how much is it really a problem?
I’ve had shippers call it the bane of their existence.
I’ve had shippers say they hate it.
I’ve had carriers tell me,
It would be a miracle if we could solve MP2 or 52.
So that’s the feedback I’ve gotten.
Not that it’s universal experience of everyone, but, it’s not that irregular that I’ll get like a forwarded email from one of our product customers where they’re, emailing back and forth about trying to acquire allocation.
So, my experience has been that the capacity management side is,
still remains a challenge.
And I think most people think about it only when the market’s tight, if that makes sense, ’cause you’re trying to find it from the shipper side.
But I think often when the market’s soft, you’re basically overpaying for allocation if you think about it as something that’s explicitly priced.
So, much less to talk about the general inefficiencies that happen with overbooking and fall down and everything, right, that goes into the system to essentially utilizing the vessel.
If you can thin a lot of that inefficiency out, I think you basically drop costs down across the entire system.
That makes sense.
Hopefully that’s helpful.
Yeah, and certainly, yeah, and we see it from the logistics side of when you’re in a crisis period, our teams are having to do a heck of a lot more work to try to find that capacity for customers, right?
So if that can be leveled out, it helps us, it helps the carriers, it helps the importers for sure.
Yeah, and I, one other quick thing, ’cause I think there’s a comment about right out.
So what I’m talking about too is specifically contract, right?
So I’m not talking about spot market rate, I’m talking about you’re shipping on contract on time year round and you basically manage the allocation, right?
Separately from the carriage terms that then you essentially keep your carriage terms independent of essentially the state of the market.
So throw that in there too, just to make those and clear.
Thanks, Andrew, appreciate that.
There’s another question in here, legitimate concern about AI replacing staff on a larger scale.
How do you manage that with your team and the companies?
And, quite honestly, and I think in the logistics space, in the retail import space, in the provider space, we’re all trying to do a lot with with less people, right?
The, we’re trying to reduce costs everywhere across supply chains.
And so people are,
called on to do a heck of a lot more.
And so for us, at least at Century, empowering our teams to be able to work more efficiently, to do less of these repeatable tasks, opens them up to do other things for our customers, opens us up to grow the business as we’re onboarding new customers without having to add more and more staff to offset that.
So, yeah, so, for us, it certainly hasn’t been an issue of, okay, we, we just don’t need people anymore.
We need them.
We’re just reapplying how, how we utilize them.
I do think it’s really important for, for tech folks out there, the more that about the business, the more valuable you’re going to be, right?
Because we are seeing these tools enable business people
people to be able to become builders, to be able to deliver a lot more of this work.
So when you think about the multi-step process in the past of, okay, I want to build something, I’ve got to go to a project team or product team to put together specs and somebody’s got to draw up screens and
A lot of that goes away, right?
And it streamlines that entire process.
We get there much, much faster.
But yeah, so will jobs shift and change?
Absolutely.
And as individuals, you should all be thinking about that and thinking about like, okay, what does my role look like in the future?
What am I doing to learn AI to be able to utilize these tools personally, professionally?
professionally, because that’s going to benefit you.
So so, yeah, I think, it definitely has an impact.
But, I think it’s positive for folks that embrace it.
And, Jim, I’ll add just a note on that.
I really like that you highlighted that many people become builders the way I think about it in the future.
You’re, if you focus on building, solving problems.
or developing talent.
These are things that AI may not necessarily be able to accomplish.
And I think about the history of technology, it’s opened up more opportunities than it’s taken away.
So I think the future is bright.
I am an optimist and definitely encourage everybody to continue learning.
Never stop learning.
Thanks.
[00:55]
Thanks, Jeff.
And I think I’m going to stop on that point.
Comments are closed.
