Jeff (00:00)
Today, supply chains are being reshaped faster than leaders can react. And the tools we relied on for decades can't keep up.

Welcome to Written in Reflections, episode 26. This is a forum for exploring the dynamic, complex, and essential nature of cross-border trade, and a space to reflect on the deeper questions that shape how we live, lead, and move through uncertainty. And today, this complexity is showing up in a new way.

I'm sure you have noticed, but AI and supply chain, including trade, have become two of the hottest and most misunderstood topics in global business today. Every week, new tools appear promising automation, prediction, and optimization.

Because trade sits at the heart of every supply chain, the stakes feel even higher. Tariffs, export controls, routing decisions, compliance exposure, these aren't abstract concepts. They shape cost, risk, and resilience in real time. Over the past year or so, I have held conversations with many people.

those creating AI tools, those exploring them, and those utilizing AI features wherever they can. In the midst of all this noise, I have found that teams are feeling the same two things, skepticism and fear. Skepticism about whether AI outputs can be trusted in domains where accuracy and compliance are non-negotiable, and fear that automation means displacement.

that expertise built over years that could be replaced by a model that can't explain how it reached its conclusions. This is why I wanted to have this conversation. So today, we're going to cut through the hype. We're going to talk about what trustworthy AI actually looks like.

and explore how system-level intelligence can help leaders navigate a world where supply chains are both more fragile and more consequential than ever.

Today, I am joined by Baz Kuti, the CEO of Sustain360, a company taking a very different approach to AI in supply chain. Instead of black box automation, they're building agentic AI, systems that can reason across an entire supply chain, run scenarios and recommend actions. And instead of relying on hyperscale models, they're building sovereign AI, intelligence that is transparent,

explainable and fully controlled by the enterprise. Baz, welcome. I am so thrilled to have you join me today for this very important conversation.

Boz (03:17)
Jeff,

thank you very much for the opportunity and great to be here.

Jeff (03:20)
Great. To start, Boz, can you tell our listeners a bit about yourself, your journey into AI and supply chain, and what ultimately led you to develop and now bring Sustain360 to market?

Boz (03:36)
Well, background is in tech all my career, so about long decades of it. Two parts to it. I really worked in enterprise clients like GE, Emerson, Schneider, all industrials, manufacturing sectors, compliance, regulatory needs are key in what we do. And engineers need accurate information. And if you look at what that's taught me is that there's no lack of data.

The data is there, but it's fragmented. all over the place in different formats, in historical events, in compliance documents, in procedures, and so on, ERP systems and the like. But how can we bring that fragmented data sets into an area where we can make faster decisions, which are empirical, and engineers can trust the outputs from them? And so what I did was look back and said, how can I build something like that for the engineer? ⁓

thumbprint effectively is the engineer in mind, because we work with so many. And so we're building AI, effectively, or Gen.AI, which allows us to look at not just the biggest and fastest models out there, but more which are smaller, domain specific, private, and can be governed and auditable. And that's what led us to sustain 360.

Jeff (04:58)
Wow, that's fantastic. So let's talk about SysSane360 itself. ⁓ As you were building it, what gap did you see in the market or in the way companies were approaching AI that convinced you that a different kind of intelligence was needed?

Boz (05:15)
That's a great question. If you look at the shift of AI, AI has been with us a long time. My first algorithm was when I worked in a company in the UK, and we built an algorithm for 24-hour store opening to identify which stores were open and which stores were not open.

And that led to significant business change. And this was in the early 90s. And so we wrote an algorithm which allows us to do that. Now, so AI has been with a long time. What's happened is that the speed of change of AI with this new paradigm called GenAI allows us to bring huge, huge amounts of data sets together for the first time and put them into language models.

And these language models are general knowledge effective. So those general knowledge have been trained on massive data sets, basically the internet, and allows us to ask questions against it in prompts. Now in a regulated industrial environment when you're dealing with engineers, is that going to work?

can you actually trust what's coming out of that model? Can you trace back like an MRI on that data set? That's going to be tough to do. Because these models what I call black box. They're opaque data sets which have been used and the weights which are created are proprietary to the vendor which created them. So why don't we have a paradigm shift? Why don't we move to a world where these models are smaller?

The training data sets used, both from the base plus proprietary data sets plus client data sets, are completely transparent. You can put a trace through that data set and exactly know how the outcome was produced and trust can be built with the model.

And so can we deploy that now not in another environment, which is not in a public environment, but deploy that in a private environment, where the organization's intelligence is secured. And so because intelligence is proprietary to you. That's kind of the genesis of my journey.

Jeff (07:35)
Okay, that's great. So you're focusing on supply chain. And so for those who are hearing about this tool that you're developing for the first time, can you kind of just tell us what does it actually do? And maybe more importantly, what does it not do?

Boz (07:48)
Great question. So if you look at most enterprises, large enterprises, they've invested in supply chain infrastructures today. Other logistical networks which are in place. And so let's take an example of where I used to be. And we built a fantastic system which had algorithms in it which optimized the movement of containers and the content within those containers worldwide. In that scenario, there are algorithms

which are running and those algorithms are really workflows which are saying how much content can go into a container and we maximize that content and the shipment of that. And it does a fantastic job in doing that and automated in what we do. Significant value in that case is about 30 million return in just one quarter when we did it. What's different is leveraging that infrastructure, not replacing that. So now what I call Gen.A.I.

can complement that infrastructure which is already in place with supply chain management systems to insights which effectively are providing well there's a geopolitical event there's one going on right now yes okay if you look at that geopolitical event was kind of telegraphed about two weeks ago all right and

How could we have reacted and understood that so the Gen.A.I. could take the geopolitical event, right, then interact with existing systems and say, what are the implications of a pretty major war now going on in the Middle East, and then replan, resimulate.

give you options and recommendations to mitigate the risk before it actually happened. And so this is where I see the traditional AI and GenAI, Secure GenAI, working together to solve real problems.

Jeff (09:50)
Wow, that's powerful. That's a really great example of that tool. So I've heard you talk about in your tool, agentic AI. Is that what you're referring to here? Agentic AI, or is that something that's different?

Boz (10:03)
So what happens here is there's two real key words here. There's GEN-AI, which is effectively if you decompose it into two parts, language models and agentic. So let's start with the first piece. The agentic piece is just like a co-worker.

but it's a digital piece of software, okay? And it's automating actions and tasks on your behalf. So in the scenario we just had, it's going back into those routing plans, finding the data sets and saying, okay, how could we replan ⁓ another scenario, okay? And potentially doing that action for you.

So actually doing the action. So productivity can be significantly improved if we're doing that. And that's kind of the agentic piece. The model piece is providing the intelligence, the context, and the intent around how decisions could be made. And it's captured the organizational intelligence, as they called it, so that the governance and the rules by which we operate. So for example,

when we stuff in a container, what size of container should it be? That's organizational intelligence, or the pricing logic used. It could be organizational intelligence in the small language model, which is being picked up and then applied by the agent. Does that make sense?

Jeff (11:31)
That does make sense. Yeah, that's a very good example of how it works. One of the biggest concerns I hear from supply chain and trade teams is this idea of trust. ⁓ They are in their roles, they're responsible for accuracy, compliance, and risk. And a lot of times I hear skepticism about AI outputs. So when you were building Sustain360, how did you think about that trust gap?

Boz (11:59)
That's a great question. Let's step back a little bit. I think this trust gap is huge. I think it's one of the biggest things we've got to, as an industry, look at. I think, in my view again, customers, consumers, policymakers, and governments are not fully understanding the implications and the consequences of this. We step back and look at it philosophically ⁓ and say there's something called thought and there's something called thinking.

Humans generate thoughts, and these GenAI models, language models, are very good at automating thinking. What can potentially happen is that thinking from these models are influencing our thoughts. That's a philosophical view on the merging of thinking and thought, in my view. Now, if you've got a model...

And that thinking model is now a large language model, which you do not know how it was trained. What data sets were used? ⁓ Who built that? And what biases are in there? And more and more recently, has it been contaminated ⁓ by an adversarial actor? And if that is the case, these are black boxes.

Jeff (13:18)
Right.

Boz (13:26)
And so how do you build trust in the outcomes on a black box you have no visibility of or transparency into? trust becomes really, really key. So in industrial regulated environments, engineers are judged if they know the empirical information and data which went into their outcomes. So building on general public LLMs can be very dangerous.

That's why we believe sovereign AI, our trust is at the heart of everything we do. And the outcomes which come out of it. we can actually, within the customer's environment, where the client controls the model, not the other way around. So in the first paradigm, the data moves to the model. In this paradigm, the sovereign AI paradigm, the model moves to the data.

in your environment. You control it. OK. And so now you can trace back. ⁓

completely the recommendations and the outputs which came out of it. And an auditor, an external party, can check those off. That's kind of the key is building trust in where we're at. Just another kind of example of it, I'm here in this podcast because we met and you said, oh, this guy's got some knowledge. He's got some experience. He'd be a trusted advisor to give the listeners his insights around sovereign AI.

and large language models. So we built trust. And because of that trust, I'm a credible voice, which can be checked and validated. My credentials are there. My history is there. It's on LinkedIn. And so we built trust so that can be on this podcast. ⁓

Jeff (15:19)
And so

that's really, that's the essence of the sovereign AI. It's this ability to trust the data because it's data that you have, ⁓ what I'd say, walled off as part of your own data or as opposed to the large language which is getting data from everywhere.

Boz (15:23)
over.

Exactly. You can verify the data. hearing this podcast, I'm very favorable in what I'm

Jeff (15:50)
Boz, you and I are really watching the same thing happen. know, AI and supply chain have both become this new hot space. And suddenly everyone is building something. Now, and I have people calling me all the time with different ideas and how trade could work. But I'd like to understand from your vantage point, what's driving the explosion of tools and what's getting lost in the north?

Boz (16:15)
That's a great question. And so there's kind of two factors which are coming together. Firstly, supply chains are volatile. The volatility is driven by a social media event. One tweet to whatever you want to call it is now allowing complete change in supply chains to occur.

Jeff (16:34)
Or one tariff.

Boz (16:36)
One tariff change, So a political event or a geopolitical event just drives everything out. And so the discussion around supply chains has moved from an operational topic to a boardroom topic. Big change. Because it's having a material impact both at national level as well as company profitability.

Second thing is the maturity of these GNI tools and their capabilities is just getting better by the day. The power and the capabilities these tools have allows the leaders of tomorrow to realize that if they don't do something, they may not be relevant tomorrow. And now what's been lost in that noise is that the foundations by which we sit here today have been laid.

If it wasn't for the internet, if it wasn't for social media, if it wasn't for digitization of ⁓ ERP systems and the like, which provided the ground and the data, provided basically the fertile ground by which gen AI and language models can be created, how we exploit that is what's happening today.

Jeff (17:52)
Right, that's good. I think that's a good way of looking at it. ⁓ So, Boz, one of the things I noticed when I looked at your tool, because you gave me a demo and so on, and I was most impressed, but one of the things I noticed is that you are focusing today on something that's become strategically important, and that is critical minerals. And so let's kind of park on that, because today, critical minerals ⁓ have become something that's become national security.

security interests, governments are paying close attention. And so just really curious, are you seeing governments engage with this kind of system level intelligence and what role do you think they'll play going forward?

Boz (18:34)
So critical minerals ⁓ are in all modern-day electronics, from your laptop to your cell phones to your defense systems ⁓ to your washing machine. They're everywhere. ⁓ And historically what's happened is that the concentration of these critical minerals, there's about 63 of them, the US has defined 63, and most ⁓ countries ⁓ like the UK, Europe and

Australia aligned to those 63 and off the 63 11th are classified as rare earth elements and they're very kind of what's the word important to mine these is very very difficult okay and also to refine them right and what's happened is the concentration of these critical minerals is in a selected few countries in particular China yes

China now controls roughly 80 % of the rare earth elements, both in mining and refining. When you go to other minerals...

They may not have the mining capability because that's spread out to other countries in Africa and Latin America and so on. But the refining of those critical minerals is in China. Refined critical minerals are then ending up in specialist components, which are in electronics, like specialist magnets. So that concentration of...

creates risk because it now can be used for geopolitical purposes. So, give an example. The country of Congo, DRC, it produces 80 % of the world's cobalt, which is required for the batteries. mid last year, they decided, so it's just been a supplier, just carrying on supplying.

They decided to manipulate global commodity prices for cobalt. They put trade restrictions on, export controls, quotas in, three times. Drove prices on cobalt up by over 120%. Just one country impacting the entire EV industry for the world. That's the impact critical minerals and over-concentration can have. So what are governments doing?

The US government has created Project Volt. Project Volt was a $10 billion investment into building national reserve capabilities.

There is in plans right now the Secure Critical Minerals Act being proposed in Congress. And that is also to secure national reserves for the US. Other countries like the UK and Europe and Australia are doing exactly the same thing. So critical minerals has become a very, very hot topic because of the concentration of risk in certain countries.

Jeff (21:43)
Right, no that's a really great explanation. What I'd like to ask you though is you have this tool, Sustain360, and I'm going to use your tool. How is your tool going to help me with this whole critical minerals or rare earths?

Boz (21:59)
Okay, there's two sides to the use. Okay. There's the demand side and the supply side. Okay. On the demand side, you're an electronics company. Yeah. Okay, and you're making whatever widget there is or componentry there is. Right. Okay. Well, how do you know your exposure to these critical minerals? Because they're usually in your tier three and tier four supply chain. Okay. You know your tier one, but you really have minimal visibility in tier three and tier four.

So we can take an electronic product, a physical product which you're selling, it through our software, it through our sovereign language model for critical minerals, and identify which critical minerals are being used and where they're coming from and understand the source of origin and the risk. Now once you have that baseline, you can now look at that and say, okay, what's the impact of tariff changes, quota changes?

and mitigate the risk. Now when you mitigate the risk, some of those mines, this now comes on the supply side, some of those mines may look like they're in countries which are friendly to the US. But the ownership of those mines is not.

They're owned by adversarial countries. So our tool will also tell you who owns the mine or the refiner, so that you can say, OK, I've got clear visibility to reduce my risk for my adversarial content and plan alternative routes or alternative methods. On the demand side, if you're a miner or a company producing minerals today,

this huge explosion of each government's setting aside huge repositories of critical minerals as reserves and you'll be mining in various locations and you're a friendly miner effectively, Maybe in Europe or here in the US. Well how do you take all those reserves you've got, an existing mining capacity and now pivot to a potential new world where you're now being asked to turn

certain mining capabilities into these critical minerals so that the particular government can reduce its dependency on China. So the demand guys are saying it's an investment model for them. Firstly, baseline what capacity, what production, what reserves, what licenses we have worldwide. Then look at these potential demand both from data centers, EVs and so on and government surplus.

requirements, stock binding requirements. And now we'll do an investment model, which says how do I turn an existing mine or an existing reserve into a strategic asset to meet that demand. So it's two sides of the equation, if it gives you a feel.

Jeff (25:05)
So it sounds like if I was using your tool, I would be in a much better position to make good decisions when it comes to this really volatile and tricky situation.

Boz (25:11)
Correct.

Key decisions, transparency and clear visibility are those decisions.

Jeff (25:21)
⁓ And without

your tool I would be left trying to search all over the place and trying to figure out and I wouldn't even know if I trusted ⁓

Boz (25:28)
You'd

have spreadsheets, you'd expensive consultants, spending a significant amount of cost to do this exercise.

Jeff (25:38)
Okay, that's very helpful. It gives me a good sense of ⁓ what I could expect if I purchased your tool for critical mineral. ⁓ Let me shift a little bit because many of our listeners are related to trade practices. It's kind of where I came out from. A lot of people know that. for our trade listeners, ⁓ I just wanted to confirm, Sustain360 is not a customs automation tool, correct? But trade is a major input, right?

into this fly chain into your tool. How do you see a trade team using the intelligent platform that you're providing?

Boz (26:15)
Absolutely, we are not a customs AI trade trade. That is not what we do. What we are is providing the earliest signals which are occurring in whether they're in the social media environment or on websites, on government websites or in the news media about changes to export controls, policy changes, rules changes in sources of origin, country exposures.

potential wars like we talked about okay and sustain 360 takes these changes in policy signals and allows you to model those changes and impacts and simulate investment decisions so that you can make the right choices for your business the right course of action for you.

Jeff (27:05)
So it may put that into real terms. So it may be that we could export or cross border in Country X. But based on some of the modeling and the intelligence that you would gain from using your tool, we might realize it may be more cost effective to go through Country Y. So it gives us the ability to make good decisions about how to minimize our risk, how to minimize our cost, how to streamline.

Boz (27:23)
Correct. Absolutely.

And what financially makes sense to do that. Absolutely.

Jeff (27:33)
the cross-border.

Okay, that's helpful, very helpful. I'm glad you explained that. So finally, ⁓ Baz, if you're a CEO or a supply chain leader and you're listening to this podcast, ⁓ what's the one capability that they should be building now to prepare for the next decade of supply chain volatility? Or maybe it's not even a decade, maybe it's next year.

Boz (28:04)
what time horizon could be subjective. But at the end of the day, think the one capability, given the volatility going on in the world, and I think the volatility will continue. We are going through a seismic shift. We had 30 years of globalization, supply chains, which are efficiency driven, cost driven. And now we've got a period of reshoring.

And that reshoring and national interest in terms of protecting supply chains and critical minerals that we talked about is going to be a generational change. So the one thing supply chain leaders should consider is how do we build resilient supply chain networks with the intelligence so they can react to volatile changes.

Jeff (28:52)
Well said. Very well said. I think that if I was the CEO, I would take you and say, okay, so what do do now? Well, Boz, before I wrap this up, I just want to offer you anything else that you didn't get a chance to say or you want to say to the listeners here that are listening to this podcast.

Boz (29:14)
I think the kind of back to the philosophical point. ⁓ We as humanity need to understand the implications of what this gen AI is really doing. ⁓ to the younger generations. ⁓ Morally, this can influence our thinking. We've seen it day in, day out. So the rules and the standards and the governance around language models.

is the onus on government societies to really take a harder look at the implications of what could happen to behaviors because of these models. And there's really a philosophical view and a personal view on that.

Jeff (29:56)
Thank you for sharing that. very important. Thank you. Well, let me go ahead and just wrap this episode up. ⁓ Now, as we do, I keep coming back to something you said today, Baz. ⁓ Supply chains aren't just data problems or logistics problems. They're systems problems. And in a world where climate shocks, policy swings, and geopolitical pressures are accelerating, leaders need intelligence.

that can reason across the entire system, not just report on it. Agentic AI, sovereign AI, and the kind of scenario-driven modeling we talked about today aren't about replacing expertise. They're about giving teams the clarity, transparency, and confidence they need to make decisions when the stakes are high and the information isn't complete. So whether you sit,

in trade, supply chain, compliance, or even the C-suite. The question isn't whether AI will shape the next decade. It's whether the tools you rely on will help you navigate that complexity or leave you reacting to it.

Vaz, I want to really thank you for taking time and explaining these very ⁓ important concepts and your particular ⁓ tool that you're developing. Thank you for sharing that with us and for joining me on this podcast. And to everyone listening, thank you for being part of this conversation. These are the questions that will define how we build resilience, how we lead, and how we prepare for what's coming next. This is Jeff Rittner, and you've been listening to Rittner Reflections. We'll talk again.

real soon.