Technology leaders panel – Developing digital roadmaps for the next 12 months and beyond

SPEAKER A

Absolute pleasure to introduce our first session this morning, our technology leaders panel. Allow me to introduce Chris Sass, who will be moderating the session. Ladies and gentlemen, remember, if you haven't already, please make sure you're posting on LinkedIn about how much you're enjoying the event. Be much appreciated. And my boss won't sack me, so enjoy. Chris, over to you.

SPEAKER B

Well, welcome and thank you for the introduction. I'm Chris Sass. I was downstairs, I apologize. I get sucked into technology, I can't help myself. And I was looking at demos of some of the vendors down there and making sure I understood what they were offering, which kind of dovetails right into the panel. Today we're going to talk about where we're going the next twelve months, what we see taking place. And it's nice because of who we have on our panel. We basically have users and people that have experience building the infrastructure in place. So rather than me introducing our guests, what I'd rather have them do is take just a couple minutes, introduce yourselves, give a little bit of your background, and then we'll get into the conversation. So why don't we start with you, Magesh.

SPEAKER C

Hi. As Magesh Nair, last few years I've been on ionside, which is on the selling of ETRM in commodity side, and prior to that I've been the buyer in selection and implementing a cioco for Castleton. Prior to that, Conocophillips as well. About a decade. So decade at ConocoPhillips, close to a decade at CCI and a couple of years at Ion. So hopefully my perspective on both sides of the equation there helps.

SPEAKER D

Great, thanks, Chris. Casey Becraft I'm the head of it for 6.1 commodities. I've been in the industry on the tech side for approaching two decades now. I spent the first half of my career working for a number of ETRM CTRM vendors and then made the jump to the client side, spending time at Louis Dreyfus, north american power, Sorense Castleton as well. And of course, of course currently at six one commodities. So I think during that time it kind of gave me a real perspective on this space, both from a technology standpoint, saw a lot of evolution during that time, and a lot of change in methodology as well.

SPEAKER B

I think it's timely talking to you because I think you're building a lot of infrastructure from the ground up right now. So you're putting a lot of thinking, a lot of research into what's current in what's the best practices of today. What I'd like, to start with is talk about data management. So much of the problem that we're focused on today in it and other systems is the data. What kind of, where are you seeing data management going today, and what are you seeing in the trend there?

SPEAKER D

Right. The constant demand that we're trying to meet is making data available to our users at the same time putting, make sure we have security around that, both internally and externally, making sure that the right people internally have access to data that they should, and those sets of data, and then also externally working with our cyber partners to make sure that we're securing that proprietary firm information. So I see that's kind of one of the biggest hurdles we're facing right now is around the security of that data.

SPEAKER B

So let's unpack that a little bit because it's pretty high level still. What does that mean? And what kind of tools or strategies are you taking to make that work?

SPEAKER D

Right. So it's permissioning and putting process around data sets that certain groups should have access to, building out those data lakes and curating that data to make that available to specific groups and users. Obviously, we don't want everybody having access to p and l information. That's high level information that we like to keep at the executive level or for those who should be privy to that information, simply stated. And then from a cyber standpoint, putting restrictions in both from who our external consultants are who are coming into our ecosystem, how they have access to our systems and data, as well as working with our cyber partners to constantly build in robust defenses around our databases.

SPEAKER B

Now, you come from, both of you, I think, have a history of CCI in the past. So lessons learned to do differently. Since you're starting with a green field, what are you not doing that you might have inherited if you were in a legacy infrastructure? What's unique to your new infrastructure?

SPEAKER D

Right. I think, you know, the approach that I'm taking as a smaller firm is as many supportable solutions as possible that I'm plugging in place here. We're not an it shop. We're a merchant energy trader. And, you know, for, so there are certain parts of our technology infrastructure that are proprietary, I would say definitely our own research and data that's used for commercial decisions. When we talk about operational and transactional data, I'm trying to make that as supportable and robust as possible so all those different components that fall into those ecosystem are fully supported. Of course, there is some parts of that that we do have internal development around, but really trying to make that work as much like a swiss clock or watch as possible.

SPEAKER B

So magesh, you've built huge infrastructures. Why don't you share a little bit of your perspective?

SPEAKER C

Yeah, I think you used the word greenfield. That's a very interesting way to think about. And it's a green field, it's a whiteboard. You get to start fresh, make the right decisions. What are the current decisions? Involve the consultants, like Casey said, all that you could do and start off with a fresh mindset and do it. But when you inherited a. When I joined CCI, it was called Louis Dreyfus Hybrid energy 10 zero year old french company infrastructure as 400 green screen. So keep that in mind. I'm not talking 1950s, gentlemen. It's 2011. And in 2011, if it's green screen, you walk in as a global CIO, you are asked to now change the tires of a car that's moving at least 50 mph. Right? So that's an interesting way to think about. I don't want to break anything. You still want to run the company, you still want to publish your p and l valuation risk. Everything has to work. But except you're moving them from the stone age of as 400 green screen to all the way to CXL and all the other modern technologies. So you had to go very methodically planning out what that switchover looks like. It took about 18 months for that roadmap to even get there from. I would think of like I would lay out the infrastructure since you used the word. It's even deeper than EGRM. We had to go first to data centers, we had to migrate them from email itself, Microsoft Outlook for the first time in the company, whatever Lotus one, two, three on mainframes were running, I think before. And then after that, get your data centers, windows, all that stack built in. Then you bring in your ETRm pricing systems, SAP for your ERP, right? All that had to be brought in and then you're migrating. At that time we were not eight trading floors at the peak, we were eight trading floors globally. When I joined, I think it was about four. There are a couple of folks, my colleagues from CCI here, correct me if I'm wrong, three or four trading floors at least line of businesses wise, you're talking about all these businesses who let me one at a time, migrated. So, which means you had to keep the old world and the new world going right without breaking it. So it was an interesting way to think about changing tires as an analogy that I gave that I could think of since I lived through it the first hundred days of my strategy work was all about thou shalt not do any harm and make sure the shop runs. And yes, I know what I need to do, needed to do get them out of the stone age into the modern age. Easier said than done, especially when it's a moving cardinal. If it's a green field, it's a different ok.

SPEAKER B

But I'd imagine 90% of the folks in the room aren't greenfield. I think that's a unique place. And so we got a good baseline where both you guys are at. If we look what the panel's about, what are we going to do and start planning forward. So let's start going forward a little bit. Let's assume you've got the infrastructure. Casey, you described your infrastructure that you're building out. How are you planning for your data and your data security to evolve next? Twelve months, 24 months, am I going to be needing AI access for some new tool? Am I going to have some new vendors coming into the ecosystem? What kind of thought process are you making there to evolve forward?

SPEAKER D

Right. So I think as I'm building out, again, you know, early, early stages, as I'm building out that ecosystem, the approach I am taking is best of breed solutions to manage our transactional lifecycle. So we're talking specifically about that, those data streams. So as I'm building that out, knowing that I'm not pigeonholed to that technology or obviously it's hard to do an EGRM replacement, that's where those challenging part, but still as I'm building that out, be able to have plug and play solutions. So for my price feed, knowing that in future if there's a better solution out there that I can take that out and plug that into the APIs for my ETRM system or for our bolt on application, making sure that that's robust enough to meet our needs to make up for any gaps in any of those functional areas there reporting solutions. So really kind of creating this giant ecosystem of best in breed applications that are really fundamental and special for whatever that nuance is.

SPEAKER B

So how is the general vendor ecosystem supporting that best of breed philosophy? Are they playing nice in the sandbox together? And what kind of learning curve have you had there?

SPEAKER D

Right. So obviously some vendors have partnerships with other vendors, but not all do. And actually I think in yesterday's CTRM session some good points came up around the use of APIs more and making those more widely available. I believe it was molecule yesterday you mentioned that that was one of their biggest products was making their API available. To their customers so that they can put in that plug and play methodology regardless of who's trying to integrate with their system. So I think the rolling out more integration tools like that allows for that to be more sustainable.

SPEAKER B

I think you mentioned data lake before. So data has been this evolving thing. I'd love to be a data consultant because you're always employed, there's always a job to change the data. We're going to have lakes, we're going to have silos, we're going to do whatever. What's the trend we're seeing? You're saying data lakes is the trend and what kind of data are you trying to capture these days?

SPEAKER D

Right. So I think there's two sets of data, right? There's transactional data, and then there is that research, fundamental commercial data. And I. So I think capturing both those data's data in an environment and databases that we can make accessible to both those groups is what we're eventually trying to accomplish. So for operational folks, getting that transactional data to them to make, you know, better decisions and operational processes, and then from a commercial side, making all that research data available to folks easier, both how we capture and store that data, and then both how we, and then, as well as how we visualize and present that to the end users.

SPEAKER B

How about images?

SPEAKER C

So I think APIs is not a new invention. It's been there forever. It's not this lack of APIs, it's how much money do you have? And how open minded are you in terms of walking into an implementation or a strategy, knowing that three years from now, regardless of what your selection is, the merchant is going to change. They're going to venture into, depending on the capital and interest rates and whatnot, access to it, they're going to go into new commodities, and your existing ETRM selection is not going to be enough. You're going to bring in another new ETRM or whatever. All these things will happen. In our business. If anybody thinks there are millions spent trying to put the genie back in the bottle, basically trying to get all the systems out, I want to be in one ETRM. Good luck. Keep working on it. Just like your data common, you'll be employed forever. If you're a data consultant. Yeah, you'd be employed as a. I will bring it to one ETRM consultant. Nobody brings it to one ETRM. By the time you finish it, there'll be something more. Unless your firm has decided, we will not do anything new anymore. Right. And that's probably a stagnating, dying platform. Right. Any growing vibrant merchant is going to look for opportunity and volatility is what brings these new opportunities and we know it's always going to be their volatility. So from technology standpoint, I would say start with integration in mind rather than application in mind. If you're excellent in integration, you really don't care whether vendor plays nicely, molecule plays nicely with ion, or ion plays nicely with core. Who cares? At least get it. If you're software analogy, if I use it for hardware, if you're buying a laptop, make sure at least it has a USB port. If it has a USB port, then tomorrow's mouse and tomorrow's keypad and whatever else keyboard will show up in your world and you're going to plug it in. So if you start with integration always helps. For me, it helped significantly because in mechanical village days, we started off with the Enron days and all that. If you think about it that way, early, two thousands, we had a messaging bus first. Put a messaging bus. Get good at integration. Then you can start shuttling data from any silo that you have to any silo. That's how you handle legacy companies. Greenfield is a different story right now. He's allowed to dream because he's a new guy, new company, all that. You are allowed to dream that. I'm going to keep it simple. I'm going to have one system, one database. The dreams just you wake up one day and you realize somebody will walk into his office and say, we are going into whatever, renewables tomorrow or something, or we just acquired an m and a company and it brought its own etrm. Now what are you going to do? So having that mindset of I'm an integrator first rather than an application selector. Application selector is what nineties were. SAP was supposed to be one system. How is that working out for you? Same thing with ETRM. One open link is supposed to be everything ETRM. We can do a poll in this room, how many are absolutely thrilled with your etrm, and everything is running in only one etrm. Show of hands, case closed.

SPEAKER B

All right, so switch gears to ctRm. I'm going to call it ctrm instead of just etrm. Let's be a little bit broader in the technology there. What are you seeing trend wise taking place? So we're looking forward. You're trying to make decisions. I think there's agreement. I would agree that the customers that I talk to and the institutions I talk to have multiple etrms. I don't know anyone that has a single one. It's a crowded ecosystem, there's a lot of new players, there's a lot of existing players. What are you seeing taking place? What's coming up on your radar for CTRM in the next 1224 months?

SPEAKER D

Yeah, I think the 1224 months in the CTRM space, let's be honest, they don't move fast. So it's hard to say the upgrade.

SPEAKER B

From three years ago.

SPEAKER D

Exactly. So it's hard to say there's a revolutionary change that's going to happen here in the next year or two. But the trend I do see is a move away from the monolithic large CTRM systems of the past that require heavy support, both internally and by the vendor to more agile SaaS solutions. Again, one of the things I think keeps the monolithic ctrms is they've evolved and built over time and some of them very custom to, to specific businesses. But definitely the trend I see is more agile SaaS support solutions that require less support, less onus on the client to maintain and more on us on the vendor to maintain and host.

SPEAKER B

Thoughts there.

SPEAKER C

Future of ETRM? Well, they don't move fast. Having worked for one, I can one it has many products in Aion, as you guys know, the risk of disruption is so high because of all the integration in ETRM. To be fair to vendors, once you're in in a large company, you're not going to be able to upgrade everything tomorrow and say, I have a version ten which is completely AI, let's upgrade it. You're not ready for it, even if they can build that product. So let's be honest about two ways, right? Both vendor can write software, writing software is at least issue there integrating and implementing. And that is why whatever you paid for your software, whether it is SaaS or licensed software, it's an illusion. It's less cost and all that. You're just renting software. But by the time you bring that new animal into the zoo, you got to make it play nice with all the animals in the zoo. So that's integration work, reporting work. Any software you bring in, you had to do one of those. Both of those, right? Integration and reporting and that you spend a lot of money. I don't know if there is any research that can tell us the number. With my experience of two decades plus in ETRM, I would say at least five x you spend on implementation by the time you put an ETRM right with all the integrated for a mid sized company at least. So I think the upgrades, there's a reason why every other industry you can talk to, they'll tell you energy industry is very slow in technology. It's not up to speed. And that, and this, it's not because idiots work in energy industry, it's because it's complex, right? If you take equity trader for that portion, everything is standardized by exchange. And even in our industry, energy derivatives is standardized, right? Chicago Mercantile exchange will say standardized contracts. This is a symbol, all that, that is why you can put in. And if he has to select an ETRM for derivatives only, he would be done in two weeks probably, right? Net sitting here, coyotex Sungard, if you remember our story, systems like that available, you can just get a derivative system, put it in, you're done with ETrdeh. But the minute you become physical, it's complex. The minute you touch physical, commodities, storage, transportation, scheduling, all that complex global, more complex illiquid markets, more complex. We are layering complexity. So once you successfully wrestle that bear to the ground, really implement an ETRM, the least you want is one more fancy it guy feature to come in and kill your shop. I don't care what AI, right? So that is why our industry rightfully moves slowly.

SPEAKER D

If I can echo that real quick, I think there's a reason nobody's ever created a silver bullet with all the bright minds that we have in this business, as you mentioned, because it's extremely challenging and verges on, I don't want to use the word impossible, but it's moon landing esque.

SPEAKER B

All right, so you have a monumental task in front of you, but you have to deliver. Your shareholders really don't care that it's challenging. You've got some responsibilities. Magesh mentioned the word that I don't think you can get to a technology panel these days without talking about, which is AI. So complex things. Maybe AI is better suited than human for some of these things. What is your current take on AI in near or longer term? What are you thinking?

SPEAKER D

Right, so right now that the struggle I have with AI is not the technology itself, but the application of that technology within our business. I think the easy, like low hanging fruit applications are data reconciliations and optimizing those. And I think we've heard from the clear docs folks the past couple days, I think that's the find those correlations in data that requires less manual intervention. I think that's definitely where AI can immediately short term.

SPEAKER B

And is that real time? Is that something you could plug and play today, you think?

SPEAKER D

Hopefully. Hopefully? I mean that'd be the idea. But I think even some of the struggles that they have is mapping of data and find those correlations in data. And I think as those AI models reference more data sets and learn over time, I think those correlations will become more and more unknown. And hopefully, I'm not saying completely eliminate, but greatly reduce the amount of time that's required for manual data reconciliations.

SPEAKER B

Okay, so that's a swim lane where you see the near future for AI magesh, do you agree?

SPEAKER C

Yeah, I think Genai, I'm not talking about the whole AI, let's talk about Genai, because that's what brought AI into everybody's head quickly. The way anytime a new technology came into my view in the past when I was a CIO evaluating technologies, whether should I jump in or at least kick off a proof of concept budget for that, this is an analogy useful. For me, at least it was useful. Is this technology a vitamin or is it a painkiller? Because vitamins are great. We all want to take it, supplement it. But how many vitamins, what's the dosage? And then if you forget, are you dead today? Are you crying? If you didn't take vitamin D today? No, Genaa is exactly a vitamin. Okay? Right now, tomorrow it can change. Today you probably have chat, GPD or one of those in your phones, and you probably used it enough times, and you probably did your own work without Gene also today. Right? So that's the vitamin angle. Yes, it's great. Yes, it's better. Yes, it's good for me like a vitamin. But if I don't take it, it's okay.

SPEAKER B

The difference is the security risk in your network because the genie's out of the bottle.

SPEAKER C

Yeah, that's a different topic. Security is an altogether different topic. If you're talking about gene, I can identify it.

SPEAKER E

Right.

SPEAKER B

I can choose not to take my vitamin C this morning, because I just don't need to take it. But if I open my AI platform inside my infrastructure, and my data isn't as secured yet as I had hoped it would be, isn't there a challenge?

SPEAKER C

Now? If your shop is already a mess, AI isn't going to make it messier, it's already a mess. You're just going to deal with those issues. If you look at analytics, the four things, what are the descriptive, prescriptive and predictive, and all the analytics? There are four stages. If you're going to implement AI, you better be at least in the three before you go to inference and prediction analytics using AI. If you are a math quant based research platform in a mergere, even with guys, people doing already, Matlab and Python work and time series data and all that, that data is already sanitized and because you're deploying millions, it better be good. Now, operational transaction data, we already talked about ETRM word. I'm not going there. But pre trade analytics data, if you're deploying capital and that's the bread and butter of your platform, you better be good. Otherwise you're going to go out of business deploying capital with the messy data and AI isn't going to help.

SPEAKER B

All right, so that brings me to are we doing all this in house these days? Is a trend to have build your teams and hire the people you need? Is it to hire partners to help you do this? Are you getting services? What do people want for pre trade for your infrastructure? So I'm going to look at Casey.

SPEAKER C

Pre trade is all internal in my experience, because that's proprietary.

SPEAKER B

Proprietary. But Casey's building infrastructure.

SPEAKER D

Yeah. I think for our applications of AI, again, it's identifying those use cases and then once identifying, then I think it's going back to my team before it's given our, my calculus is a little bit different being a smaller organization again. So I'd be looking externally for that consulting opportunity to someone or vendors already have AI technology built into and again to plug into my ecosystem.

SPEAKER B

How about more generally in your infrastructure? You said you're not an IT shop, so your core competence isn't being an IT shop, but you need these systems to function and run the business. How much of that is in house these days and how much of that are you looking to have vendors be responsible for?

SPEAKER D

Right. I'm looking to again look for vendor as many vendor support solutions as I can plug into my ecosystem. But with that said, I'm never going to be able to move fully away from that. Things like my bolt on application where we're constantly making up for gaps in functionality, that we have that flexibility, that again, meeting our stakeholders demands. End of the day, they don't care how we get there, it's just the fact that we got there. So having that flexibility to build, do internal development and make up for functional gaps in my vendor supported ecosystem won't go away. But with that said, having a lean team with the right people in place to deliver on that is, that's my strategy.

SPEAKER B

So how is the staffing going, building that team, finding the skill sets, what are we looking for today that may be different than in the past?

SPEAKER D

Right. I think for us at six one commodities, I think we're looking for people, even on the IT side, that are right cultural fit for our organization, that are flexible, they have a natural Hungary thirst for knowledge, and that they can collaborate across teams with little or no supervision. Again, we're a lean shop, so everybody's got to do their part. So those are the people we're looking for. And I think that is one of our, I think our advantages in hiring folks is that we give people this opportunity to work, worked relatively autonomously and be a contributor versus some of the larger organizations.

SPEAKER B

And Magesh, when you're putting the team together, what are you looking at these days?

SPEAKER C

I always gravitated towards entrepreneurial cultures, having worked ten plus years in Carnacle, Phillips fortune five at that time, 100 year old oil company. Not exactly entrepreneurial, right? So you spend long time there, then you gain the appreciation for what it means to really not just have ideas being allowed to implement them, right? So when you hire smart people into a bureaucratic culture, you're killing their spirit, right? They're going to leave eventually. So merchant energy private equity is a fantastic place where you come in and you just let loose. Smart people, give them clear vision, expectations, empower them and drive fear out. Because if you ask people to be entrepreneurial, you have to accept some defeat, some failure, easier to ask, take ownership and be entrepreneurial. And then if you are not willing to take any mistakes, not the fatal ones, at least the small ones, then they are not going to take initiative. Driving that culture has helped me immensely. Very proud to say people who I've hired even now, continue on in CCI's of the world and wherever. And that shows that, you know, they were allowed to perform to their peak and also rotation, right? People want to grow, right? It's the Rockefeller statement of how much in his deathbed, how much money you need. Little more. Same thing. That's a human thing to say. I want to grow. I want to learn something new. Very few people would say, I'm not going to learn anything. No, we don't need those people either, because they're not going to survive after a point in this space, at least with the complexity and challenges and all that. So when you bring in the right people, empower them and then give them clear vision, and then let them loose, support them when they need you, and then rotate them every two, three years, completely take them and give them a new opportunity. Extend if you do that over 1015 years, if they work for your platform, they are your future leaders, c suite leaders, not just some mid tier.

SPEAKER D

If I could add to that, real quick, you should mention the entrepreneurial spirit of our industry. And I think that's what I mean, me personally, that's what's kept me in this business for so long. And I think that's part of what I try to sell, bring people into our industry, is that entrepreneurial atmosphere, environment where things are constantly changing. It's not a widget factory. It's not. You're doing the same thing. It's literally wherever the commercial business takes us, we're there to react and deliver on that. And I think that's what keeps things exciting.

SPEAKER B

So you actually just turned over the question I was thinking in my head as you were speaking. You said, we're there to react, and this is a digital strategy session. So strategy implies proactively. So how does the digital strategy that you're, let's say, in the organizations you've been in or currently in, how is that aligning to the business? How is the company coming about that? Or is it more of a reactive thing? Is this part of the strategic advantage?

SPEAKER D

Right. Actually, it's funny you mention this. I just had a conversation with some of my colleagues about this in the past week about we were in a vendor selection process and it kind of came up, where's the strategy of the business going forward? We can only make decisions based on the information we have in our hands right now and understanding the strategy of the business as it's being communicated to us from our senior leaders. So we need to make the best decisions around that. Who knows? Maybe in a year from now, things go in a completely different direction from what was communicated to us. But unfortunately, those are some of the reactive parts of this business, and you need to be flexible and to be able to accommodate those changes.

SPEAKER B

So, but architecturally, so knowing that your stakeholders are going to come to you with last minute requests, what are some of the considerations you're looking at? I mean, we covered a few of them already, right? We talked about the best of breed kind of approach and being able to drop other things in. Are there other considerations that you're taking in your architecture that allows you the flexibility that two years from now, when I come with a new mission critical app that you didn't know existed today, how do you implement that? How do you drop that into the infrastructure without breaking everything?

SPEAKER D

Right. I think one of the biggest things is kind of going back to the data equation, is normal, is trying to create as much of a normalized data structure as you can. Obviously, not all just talk about transactional data. Not all data is created equal, but trying to normalize that data as much as possible so that we can plug and play technologies to support a business need as much as possible. And again, going back to that ecosystem of best in breed, as you mentioned, applications within there that we can plug in and have that integrate with our ETRM, or if there's a gap in the business need, functional business need for something, we can find something and send that data to the ETRM for processing. But having that flexible ecosystem that we can again plug and play things, as well as a normalized data structure that at the end of the day all that data can go to and be commonly reported.

SPEAKER B

Are there still elements in your ecosystem that you can't get the traders to stop using? Excel spreadsheets or your settlement or things like that that aren't digitized yet.

SPEAKER D

So I don't think we're ever gonna move away from Excel.

SPEAKER C

That's the right answer.

SPEAKER B

Well, there's a newer company, not in.

SPEAKER C

My lifetime, but it was not in my lifetime, not having Casey wish list.

SPEAKER B

So I guess I bring it to the audience we've gone through. Are there questions from the audience that you'd like us to address? Things that you're looking at or thinking about in the next twelve months or 24 months that you want to ask and get an opinion on? Anybody have any thoughts? Go ahead.

SPEAKER C

Sebastian Esposito from beacon. So, just talking about digitizing your digital roadmap, can you talk a bit about your iPass strategies, how you're looking at cloud, how you're looking at hyperscalers and incorporating them into that strategy going forward.

SPEAKER D

So when we talk about a cloud strategy, I think it's. I mean, we're trying to move as much as any legacy on prem applications that we have. We're looking to move as much of that as possible to a cloud ecosystem.

SPEAKER B

Are you asking more about orchestration within the cloud infrastructure?

SPEAKER C

100%. And with respect to. We talked about security events of cyber with the cloud migration strategies, maybe how that orchestration is happening a little bit more about that.

SPEAKER B

You want to answer that?

SPEAKER C

I'm not sure I fully understand the question, but I'll take a shot. I think I mentioned the integration is the first thing that I always approach, because going in with the clear vision of cloud on premise, existing ETRM, new TRM, it's going to change. Building for change and negotiating with your stakeholders for change, that's the first thing a technology leader has to do. If you fail at that, if you do the nineties thing of you give me the requirements and I will build a system for you. Good luck. You're going to fail. It's a matter of time. Something new will happen. And yes, there is new requirements. Do you want to win the argument that this is new requirements or do you want to win? If you want to win, start with saying I'm building for integration. Whether it is orchestration platform that, for example, let's take cloud. Forget ETRM for a moment, let's take cloud. If cloud was the solution, is everybody hundred percent in cloud? No, you're going to have a hybrid cloud. There's some things. You're going to have it in private, especially for large companies in our space. Let's say CCI. For example, in my experience, when I had to get a valuation engine to run fastest possible to get my mark to market p and l out by 630 on the same day after market closes, with the trader marks and whatnot, I need the fastest infrastructure. I had to invest at that time. Oracle exadata. That's not cloud. I had to buy millions of words of hardware and bring it inside. Not just security reasons, performance reasons. So there is some level of cloud is great if you're only seen as an it and a cost center. Great. Then just rent everything and say everything is done by vendors. Have 800 number and you may sleep peacefully. But if you're running any large firm, you will have complex infrastructure. Complex includes onsite, offsite, college cloud, private cloud, hybrid cloud. Hyperscaler is coming next, right? How many are ready for a data center with hyperscalers and connecting to that and running AI on that? That's always going to be there, that ramp up. If one of your quants want to do an Nvidia Blackwell 200 chip level AI tomorrow, you have to ask the question. It's pre trade analytics, it's not transactional data. Are you going to compete with buying Nvidia's latest chip with Facebook's and whoever else is ordering the latest chips? As a commodity company, how many can you buy? Do you even have people to do all this? So I think starting with integration and then negotiating change. Change as in future unseen change. How are you going to navigate when change comes to us? Us, which is not just business giving you requirements? Business giving business and it requirements. If you do that well, change will come and you'll negotiate it.

SPEAKER B

I guess I'd ask a follow up to you. What specifically were you thinking of when you were asking about the infrastructure? What problem statement were you asking specifically?

SPEAKER C

Sure. Any preferences for AWS versus Azure, for example, you recently heard about Google GCP deleting an entire database of pensioner funds.

SPEAKER B

Are there preferences?

SPEAKER C

Are there standards emerging that you're applying in your cloud strategy? You can use a dependency graph to speed up cloud and provision s three for spot compute where you don't have to invest in massive hardware investments. So there's other strategies from a design and architecture point of view that you can use in your cloud strategy. So I'd like to know a little.

SPEAKER B

Bit more about that. Do you have a religion on cloud?

SPEAKER D

So I would say right now we're a Microsoft shop, so we're, I think, pretty tied to Azure and then developing out in Azure. I think that given, you know, plays well in that Microsoft ecosystem of things. So that's a, that's our path forward right now. With that said, I mean, we're constantly evaluating what our options are and maybe it makes sense for some specific application or data sets to be in alternate cloud solutions.

SPEAKER C

If I'm building for AI, I'm definitely thinking GCP. All the other operational data can be agnostic. It could be Azure or AWS, one of them or both of them, it doesn't matter. They're not going to be significantly different there. But for AI, if I'm putting in a pre trade quant type infrastructure, I'll go with GCP.

SPEAKER B

Any other questions? Yeah, look at the mic so we can hear you please.

SPEAKER C

Thank you. So in your digital roadmaps, let's say you want to implement the new ETRM, how to choose the partner. Any advice which criteria you would use to find the trusted partner? Hopefully he will be there for some quite time and support research and development. Just in general, what is your advice would be because there are many offerings around a very interesting company with very interesting ideas. But are they sustainable? What your approach would be.

SPEAKER D

Really good question. When you talk about the eTRM ctRM landscape now versus ten years ago is really interesting. I actually think we were almost in a better place ten years ago when we talked about support abilities given consolidation by multiple vendors under a single umbrella. I think we lost a lot of the supportability of those applications along that path. And as a result of that, to your question, it does make you start to think about what's the longevity of that firm. If I'm going to tie up my horse with these, you know, with this, with that vendor, how long are they going to be bound? Are they going to get bought by a private equity firm and they're going to lay off after support staff and then everybody knows everything about that. That product. I mean, that's part of the risk you take. So I think it's, if that's the part of the supportability, concern and roadmap of that vendor, I think it's asking the right questions.

SPEAKER B

I mean, trying to drill in, no regret move. I mean, if you have APIs and integration and your data is free, the reason you get locked in is workflows and data.

SPEAKER D

Right.

SPEAKER B

So once I start building my workflow in, I'm kind of stuck. Right. And so are there any no regret moves when you're selecting a new vendor?

SPEAKER D

Yeah, I think, again, it's trying to, as you build out that ecosystem, again, not even touch talking about the CTRM system, but all the ancillary applications around that, again, trying to think about, hey, if I need to replace this, how easily can I do that? Or am I locked in? I think that's part of the calculus, at least I'm looking at.

SPEAKER C

So just to clarify, the question was on ETR vendors or the partners who would help you implement both.

SPEAKER F

Both.

SPEAKER C

Okay. Well, I worked for one called Ion, you know that. So the trouble with that space right now, ETRM, if you ask, 1015 years ago, Gartner Quadrant, you know, if you're familiar with it, if a CIO is selecting, there are four or five names on the right hand side, top quadrant. All of them are ion products now, right. Whoever was in that quadrant, and now it's the Volkswagen. Ion is the Volkswagen of ETRM. The analogy is you start with Jetta and then you grow as you grow richer, you go all the way to Porsches and whatnot. So that's the analogy. So everybody, when Ion started acquiring ETRM, start, it's going to become one ETRm, like SAP? No, that's not the idea. They are all still separate products sold separately. If Casey wants to buy tomorrow, the cheapest SaaS solution, he can start with aspect and grow all the way to open link when his platform expands. Right. So that optionality is available with ion, leave the supportability and loss of workforce or private equity acquiring. And all this is going to happen in our industry. Because if we were growing crazy product software, why did open link sell itself? Right. It stagnates after a point. They were not selling that much. They were not growing. Private equity comes in when you're stagnating and you're not able to justify your existence, and then your shareholders wants an exit. They just gave it to somebody else and got out. So that's going to happen regardless of non ion vendors, if you choose today, it's a matter of time, right? It's not going to sell like a b, two c app. It's a b, two b app. At some point it's going to reach the saturation. So vendor selection, you have to be very clear, that might happen today. If you don't like ion, let's say, for example. Nobody says that. I'm just saying hypothetically. So if that is the case, and because of that one reason you went and picked somebody else tomorrow, can you guarantee that they won't get acquired tomorrow? You can't echo us independent yesterday, today it's not. It's powered by core, right? So these things are going to happen. Leadership is about not just collecting requirements and serving. That's just service mindset. It guys fall into this trap of service mindset. Leadership is anticipating change and building for one. Right. And then when you get there, you're not surprised. At least surprised means you plan for it. Right? So ask that question, how am I going to manage that change that I don't even know what it is. So integration, orchestration, architecture, all these are not buzzwords. You got to really have people who understand those and then have a partner like a Deloitte or sapient or there are so many vendors who could help you do the proper selection based on the firm size and other requirements. They've done this for decades now, so it's nothing rocket science. And this space is not moving that fast either. So any one of them who have done good implementation in your region, your commodities that you play in, your size of the firm, your aspirations of your firm and all that, come up with a good selection criteria, go for it and then keep the integration. And as he said, make sure your data, you're basically renting warehouses. This is analogy I used always to free myself from this selection mindset. ETRM is just a warehouse. Your data, you're renting that warehouse, putting it in. And if that warehouse is not going to work, at some point, you should be able to get your data and move on. Right? If you're able to do that, well, integration well and extraction well, go for it. Unless you are spending millions and millions of software, then yeah. Be aware that you spent that much money. You're not going to get out for a little open ticket with some vendor who's not closed tomorrow. You can't select another ETRM. That's the unfortunate thing in this.

SPEAKER B

I think we have time for one more and we missed one question, so I want to make sure we get there.
SPEAKER G

There's been a lot of discussion around AI and you know, we can think through all the benefits we can get from AI, but just wanted your thoughts on the risk that you see with the introduction of AI in this space.

SPEAKER C

So just gender AI in particular, I think furious one is it hallucinates, right? So you gotta be careful. It's almost like it's an intelligent, unbelievable curiosity, unbelievable intelligence, let's say. But then for the sake of giving output, it will give you some output. So you have to check everything. So which means that risk is there. Hallucination of AI, Jai. The second is a public LLM is trained on every public information and then you bring it in and then are you going to throw your documents at it and let it go outside the door? So you have to make sure how is the vendor giving you a space where it's protected by firewall and technology leaders need to understand that. To say I'm going to let it loose on my shared drives in whatever place. They're not exactly the crown jewels of your company, which is p and l and whatnot, but some other places, let's say pick a recruitment HR, right? They can happily use gene today. There are ways to protect and fire wallet and put all the resumes you are getting in a shared folder, let loose and train your public LLMs plus into your documents. And you are searching for candidate. I don't care if you get 200 resumes or 200,000 resumes, you can have the AI train on those documents and then you can start asking questions like any candidate with this experience and that experience, this region and at least five years and massive uptake and productivity on those kind of things, right? And NLP, like if you're getting PDF's and research reports for traders, how many of them are, for example, if it is a gas daily PDF that you subscribe to, yeah, it looks great. You're subscribing to research every day, you can't read 20 pages to find out what is the point today? Are you bullish? Are you bearish? Are you neutral? Tell me your summary position on gas daily on whatever, right? Those kind of things. Think of it like a speed reading with a JNAi tool. So if you take that approach, I think today you can implement it. Of course security needs to be ensured.

SPEAKER B

So one risk that I see is generally people use it to do coding. And so if you have a trader or someone doing scripts and things like that, they may not act as they intend to act. So you really need to understand it. But I think it's coming into your environment whether you want it to or not, through your vendors and through their software, you're going to start seeing it. It would be just like any other part of the software stack where you have to worry about the security of it. You're just going to have workflows that deal with it. But I think it's coming. I think we've already talked about the use cases where you see it. I already see it in settlement and I already see it in other areas where people are using it. But more on the it shop, I see people coding with it and it can do the coding fairly well. But you have to know, understand the coding of what you're doing because it may not do what you think you asked it to do, and there's risk there of that. Getting introduced in your infrastructure is what I would see. Casey, do you have any final thoughts here as we bring this together?

SPEAKER D

Yeah, I think I echo what Magesha said. I think what the concerns around AI is, the output of AI is only as good as the data inputted to an AI model. So I think that's being able to trust those outputs and the data that goes into those outputs is definitely invalidating that is concerned going forward. And again, to the point of your firms giving your first proprietary data to an AI model and making sure that data stays securely internal and doesn't make its way externally is also the other biggest concern.

SPEAKER B

You'd use a private, you wouldn't put it in a public AI, right?

SPEAKER D

Absolutely.

SPEAKER B

You'd keep it internally and that that exists, that already those choices are there.

SPEAKER C

Yeah. I think the private AI we have to be clear about. I don't think anybody in this room, correct me if I'm wrong, in a commodity sector, want to create a public LLM. Those things are several millions and millions of dollars left to the magnificent seven companies to do it or whatever in the world, right. There are only five or six public LLMs, so you definitely need public LLM. You cannot just completely build an LLM from scratch, no matter even if you're Citadel, right? Citadel bought global license of chat, GPD, by the way, so. Or at least I read. So the point I'm making is you still want public llmdeh. The point is, whatever my documents that is going to get added to that, that needs to be private or you can only secure your documents. Don't have the illusion that you're going to create an LLM in your form. That's.

SPEAKER B

No, I understood that. Well, I think that's all we have time for. I want to thank you all so much for joining the panel. Hopefully it was informative and thanks for your time and enjoy the rest of the day.

SPEAKER D

Thank you.

SPEAKER A

To start us off there, ladies and gentlemen, a big thank you to Magesh, to Casey, and of course to Chris. If I could possibly ask our next panellists to make their way up to the stage.

SPEAKER H

Morning, Mark.

SPEAKER A

Okay, ladies and gentlemen, we will continue with our next session. Delighted to introduce our panel. Allow me to hand over to our moderator, Alexander Renault from radar. Radar, to get us kicked off. And we're going to be looking at data, the most valuable commodity. Alexander, thank you.

SPEAKER H

Thank you, Ross, for that introduction. So we had four panel members, actually, but we're missing one. I guess they dropped off. So we're sticking with this group. Welcome, everyone. Yes, so I'm Alexander Agnew, first time moderator, so bear with me, please. Radar. Radar. That's our business. We provide consolidation tools for the industry to better your market risk management. So data is definitely at the heart of what we do managing that. And we will be talking more with this esteemed panel around that. What's nice about the panel, that we have a combination of data organizers, as I would call it, data consumers. So I think what I'm going to ask them to introduce themselves and give their perspective also on data as the most valuable commodity. As the introduction says, if you want my perspective on it, data is great. Some would say, basically, time is the most valuable commodity. I'm a chocolate buff, actually, so cocoa is my favorite. But anyway, I'm going to turn to Janee for introduction and give your perspective on data as a valuable commodity.

SPEAKER G

Thank you. Thank you, Alexander. I'm Jeanne. Jeannie Amatia. I'm regional analytics manager for North America division, Louis Dreyfus Company. In my perspective, data is very valuable commodity because it helps to unlock the insights, especially for a trading company like us with like insights. It drives the data driven decision making, sustainability.

SPEAKER H

That sounds fascinating.

SPEAKER C

Please welcome the panel. Darcy Curran, Jad Orr, Magesh Nair, and Chris Elise.

SPEAKER B

Thank you very much.

SPEAKER F

Many of you don't know nature's fine, so nature's find is for quick context. Is a startup based in Chicago. We make different food products out of a fungi that one of our co founders discovered, which is very different. We have products in both the alternative dairy and alternative meat categories. And my role is really to oversee the implementation of data science based solutions for manufacturing and supply chain. Excited to provide a perspective that's closer to the manufacturing plant today about how we're using data. And, oh, data is the most valuable commodity. To me, at least, it's the most valuable and tangible commodity because it helps us, allows us to be smarter about how we use our tangible commodities, make it, make them more effective.

SPEAKER H

Mark?

SPEAKER E

Thanks, Alex. So I'm Mark Fleming Williams. I'm the head of data sourcing for a Paris based quantitative hedge fund called Capital Fund Management. And data. Well, I'm the head of data sourcing. So data is massively valuable to me, clearly, because it kind of defines my role.

SPEAKER B

I.

SPEAKER E

More broadly, I see data as the most valuable commodity for 2024. Just the example, I would say, is just look at what's happened to the car industry in the fact that it used to be all moving parts and petrol, and now it's largely electronics and algorithms, and that used to be largely driven by petrol, and now it's quite a lot driven by data. And I see that happening across all the different industries. And for CFM, that's incredibly valuable because data gives us the ability to measure all these industries. So that's why.

SPEAKER H

Very good. Well, thank you for those introductions. I'm going to turn back to Janine. I imagine in the enterprise that you're operating in, you have a lot of different type of data that you're operating with. So give us some more insights into that data and how you're building a future proof infrastructure.

SPEAKER G

Sure. So, Louis Dreyfus is a global company. We have a presence most parts of the world. I'd like to say Australia to Argentina. I am responsible for the data assets or data requirements for North America division. So from contracts, any sorts of data requests, reporting analytics requirement for North America division. So some of the requests I get could be from trading to different division, different platform, different function. It could be all sorts of requirement. It can be in house, our proprietary data or something we do not have in house yet, and they still have a question. So it could be a public data. My role, the scope is like very wide. It's a fun job. I don't have a dull moment. It keeps me really busy. So, coming back to your question about the data infrastructure, before we jump into data infrastructure, I was in this room earlier where there was a panel about digital roadmap. I think in order to build data infrastructure, you need to have a digital roadmap. First, you need to understand what your business is heading towards, what's your business change, what's your business innovation, where we're heading towards as a business, and then support that business with our emerging tools, the best of the technology tool sets. Right. So when you have a clear picture of the digital roadmap, then you can, then you can build your data infrastructure, some of the do's or don'ts for the data infrastructure. I'm a firm believer of fair principles. So your data has to be findable, accessible, interoperable, reusable. Right. While trying to do all that, I suggest to use cloud technology, your data architecture has to be scalable and flexible in terms of when you build a data lake data warehouse, something we've been hearing a lot from earlier. Panel, data privacy. When you build this data warehouse, be careful about the data privacy. And then once you have this solid data infrastructure tool set, whatever is the right tool for your organization, govern it with a proper data governance structure. Right. The accountability, reliability, the proper compliance for your data structure. And maybe just to add on top of the tool, your technological tool is only as good as your adoption. You need to have a data culture in your organization, be it a data ambassador. We call it data champion. We need to make sure that this sound, the latest emerging technological infrastructure we have built, is adopted from executives to day to day operation. So, yeah, that's my few suggestions for data infrastructure.

SPEAKER H

Some excellent insights. Thank you. And I think we're going to come back to some of those points around solid foundations, but also the governance that you pointed out. I'm going to turn to mark, and in your role, I can imagine in the fund management that the appetite for data is untouchable. So how do you go about sourcing your data, getting that organized? And where's the end point for that?

SPEAKER E

Sure. So I feel a little bit different at this conference, because while commodities is important to me, it's just a part of what's important to me. We are a coin fund of $14.5 billion under management, and we trade all sorts of things, not just commodities, but equities and FX and various other things as well. So as head of data sourcing, I am interested in data on all kinds of things, and I am very much the mouth in our organism, and I'm not that involved. I'm not the stomach. So I'm not actually organizing the data myself. I'm just finding it and feeding it to the people who are, who are better at it. So I'm better at talking about the outside data world than what we actually do with it, partly because we also don't want to tell you what we do with it, I'm afraid. But so in terms of how we go about how I go about finding it, what are we interested in? So we're a quant fund, which means that we want a large universe, that the data will give us eyeballs over a large universe of some kind of process that's happening in the world. So that could give us a good understanding of a supply chain or the oil industry, or it could be what consumers are doing in Chipotle. Not just Chipotle. It would have to be all sorts of stores. And so we're interested in finding a dataset that could be a. That could be credit card data, or it could be a company which is providing perhaps some kind of machinery to the mining industry or something like that. And we can use that data set and compare the history of it to what happened in the markets during that day, and then find an investible signal, and then buy it on an ongoing basis and hopefully make money with it. So what that means for us is we want wide coverage. We want something which covers a large field. We want a long history, because the longer the history, the more we can feel comfortable in the signal that we found. Ideally, we'd like a history that covers various crises, because we're training the computer to recognize, oh, this is another 2008 coming, so sell everything or whatever, and so long histories. And it's incredibly important for us as well, for it to be point in time. And for those who don't know what point in time necessarily means, then it's that the data should look exactly like it did, as if we had been buying it from you on that day. And so what we hate is if the data is improved later, you know, somebody made a mistake on January 1, and then on January 5, they improve it. And if we saw today the January 5 number and not the January 1 number, then it wouldn't be point in time. So that's a kind of simple example of it. So that's essentially my job, is to scour the earth for interesting long histories, wide coverage data sets, which will give me some kind of visibility over processes, and it can be all sorts of. It can come from any direction.

SPEAKER H

Okay, very good. You hear a point here about the sheer size of data. We're going to come back to that in a minute. But I want to turn to Omar, because you're a data scientist by profession, so I assume you don't want to spend too much time really collating the data collecting, et cetera. You want to really get hands on with it. So tell us more about the different tools that you use as a data scientist today.

SPEAKER F

I think you're absolutely right. I hate spending that much time cleaning data and getting it organized. I'd rather be spending time on generating insights and adding value for whatever organizations, organization I'm in right now. In manufacturing, we're kind of at a phase where we're doing a lot of reporting analytics, but we want to move to this phase of true process mining where you're collecting data, automatically generating insights, and then adjusting your process parameters for manufacturing or supply chain or whatever in an automated fashion. That's kind of the vision, that's what we want to work towards. We're not there, and it's partially because we're at this phase where data sources kind of exist in silos and it's really difficult to pull them all together. I'll bring a few examples. So, in manufacturing, we have, we're always trying to look at ways to improve yield, improve quality outcomes, and make better supply chain decisions. That's kind of the goal. We have a few frameworks for us to do, few tools that enable us to do that, and they've definitely been modernized with the help of advanced analytics tools. Let's take, for example, root cause analysis. In root cause analysis, you're trying to identify kind of the root cause behind the deviation in product quality. For that, you need to be able to pull data from many, many different sources, whether that's your ErP data, whether that's your quality, laboratory data, sensors. So time series data, you need to bring that all into one unified playing ground and then analyze that altogether. So there's a taxing process, bringing the data together, but it's also taxing to process that data. And as you've heard from the prior panel, AI is really exciting. This is one application where we've seen some promising results in it, parsing through the data set and identifying, identifying your deviations. So we've seen that in line in a few applications, but it's only possible if you're able to get all your data sources in one place. So kind of the learning from this experience is make sure your data is not in silos, but also contextualize your data. So good schema, data management and good metadata management are also very helpful. Another tool I could think of is in process diagnostics in the manufacturing line. So that's an area that's been really helpful to use computer vision. So we use a lot of computer vision in our production lines and they could process a lot of imaging information quite instantly and look through at scale a lot of your production deviations in this area, kind of the learning we've had is they've been rolled out as proofs of concept so you've set up your supporting architecture for that proof of concept without thinking more holistically about how this point solution fits into your broader architecture. And that means you're creating a new data silo, which goes to my earlier point of data silos. But then I guess the learning here is try to think about how these new proofs of concepts pilots fit into the broader architecture and work from there.

SPEAKER H

Very good. I think we are hearing a common theme here, that before you're able to actually consume data and do the analytics, that there's a lot of organization that needs to be put in place, foundations, as Ginny also said. So I want to zoom out a bit and reflect on actually, why are we talking so much about data. I did some homework for this panel and looked back in time. So if you think about it, over the last ten years, the volume of data that is produced daily has 60 times fold increased. Right? So I have some statistics here. 328.77 million terabytes we produce a day, and that's 60 times what it was ten years ago. It's huge. Now, if we dig in there, 50% of that is video, a lot of it's by the consumers, but within the business environment as well. So I do want to go back to the entire panel. Data is so broad, so how to make actually a valuable commodity out of it. Where do you put the emphasis to really help the business in their decision making? So take us along in that respect. And where do you see the direction of this going? Because I think it's going to continue to compound tremendously. So how do we navigate through this and really get the value out of it?

SPEAKER E

I mean, from my perspective, the fact, so I mentioned the importance of long histories, and actually, as a result, the fact that there are a whole load of companies emerged this year, last year, two years ago, they're not that interesting to me, actually, because they've only got two years of history. And anything before two years ago, they would have perhaps backfilled, and so it wouldn't be point in time. So my dream, actually, my dream source is a company which has been around for a very long time and has been generating data for 50 years in the course of their business, and not realizing that it might have value for someone like me. And actually it's the old data quite a lot that I want new data as well, but I quite like old data. So there are vast quantities out there. But actually the types of data which are useful with a long enough history, with a wide enough coverage, which a point in time remains limited in some ways, so it remains an art. What I do.

SPEAKER F

I think also how you store it and how it fits into the architecture, going to your point, Ginny, is really important. We've used data lake house architecture, which is somewhat controversial in the data world. But the reason we opted for something like that is you're getting a mix of the data warehouse where you're storing tons of information, but then you're getting use cases and scalability from a data lake architecture. So there's a bit of both. It's kind of a new concept. So a lot of the best practices haven't been fully developed, which is bad because we're talking about best practices in data today. But I think having that mindset where you want to be able to store large sums of data, but then being close to the analysis aspect is helpful in my perspective.

SPEAKER H

So still you need good human judgment, but you're also mark as well, richness, depth in the data.

SPEAKER G

Jenny, your perspective, data landscape is evolving, right? Like you said, we're capturing more and more data. You see it within the organization. Also, like, previously, we used to be interested in audit history for like a compliance point of view, one random example. Now we are doing analytics on audit history also, right? Analytics based on like, so how often it is changing and the reason behind why it is changing. So we can avoid those changes needed for the future cases. We are definitely capturing more data points than previous. These more data points are led or as a result of like, more insights needed for like a more streamlined process or to drive an innovation. So to come to your point of like, how to make these data more valuable, again, it comes to business case, really, it comes back to, sorry to harp on the point again, your digital roadmap, your digital strategy, where are we trying to get to? What are we trying to get to and where we're trying to go? It really comes down to your business case, and then we can come back to how do you structure your data? What are the additional data points we need to capture, or what are the additional data we need to acquire? So I'd like to focus on the digital strategy for that point.

SPEAKER H

Very good, very good. Thank you. Can I just mark? Go ahead, Chuck.

SPEAKER E

Another thought in which occurred to me, which is actually an exciting thing about the most recent years from in my space, is that frequency is increasing, that historically it was where you might get weekly or monthly data, and now it's daily and intraday is happening more and more because there's more that functionality is coming into the businesses and coming into these providers. And so that's, for us, that gives us that much more ability to do things with it. So that's an exciting thing.

SPEAKER G

And in addition to frequency and also timeliness, so example, previous if a public data release within 15 minutes of a public data, if it was acceptable, now they want to see it within 15 seconds because there is a technological capability. There is a possibility now with the new tool set. So we are also seeing the change in not just data frequency, the timeliness of also, how soon can we get it?

SPEAKER H

So to that point, actually part of the homework I also did was look at what they consider the three major trends in data and data management and speed and volume is definitely number one. But number two, which I really don't know how to react to, is that the diversity in data is actually narrowing, was the argument. So I have a two fold question, kind of, and it comes back to a question also asked yesterday in one of the CTRM panels. We promote single sorter too. Okay. But if you allow a lot of flexibility on the front end, people are going to be creating their own truth. So, and do a lot of data mining, etcetera. So how do you govern within your organization that the interpretation off of that data is actually correct and not just our own quirky perspective?

SPEAKER G

I think we need to understand that statement a bit better of like a single source of truth. Right. Especially for an agro commodity industry like us, where we trade different commodities and throughout the world now, different markets have different structure, different futures, different taxes. Right. So to your point of like having single sort of truth or having structured data, I think it comes more to the consolidation. We need to have a strong governance in place to define metadata of like a core metadata of the company in a way where we still allow and provide the detailed granularity at the operational or ground level, wherever that ground, wherever that plant, that region, wherever that particular commodity is, we will not be able to define one single source of truth or one structured data for all commodity in place. But then we will have to find a way to define a consolidation or metadata to scale or to consolidate that report in order to be able to see it to how do we define the consolidation? So that comes back to our data governance or defining the metadata for the compliance of the industry.

SPEAKER H

I hear recurring themes, solid foundations. Would you come back to Omar, you care to react to that?

SPEAKER F

Yeah, no, I totally agree. I think metadata management is where my mind first went to. And schema data management in addition to that, where you're, where you know the specific format of your columns, you know what each sensor tag entails. Thinking manufacturing. Again, sorry, but that's extremely important when we build this source of truth. But then this broader aspect within the organization.

SPEAKER E

I mean, in terms of single source of truth, from our perspective, we've got a, we know if we're on the right track because money goes up, and if the data is bad for us, then the money goes down. So that's the way that we can kind of know if it's good or bad. On the diversity point, I'm intrigued by that statement, and I'd like to understand it better in terms of what less diverse forms of data there is. I mean, it might be, I mean, looking at my space, my world, an awful lot of data providers emerged during COVID because perhaps because people didn't have anything better to do or because there was, because actually I think the hedge fund industry, like everyone else, had to understand this new world. And so we're suddenly reaching around for interesting data which might explain this new world. But then I think a lot of those data providers kind of disappeared as soon as they arrived. So that could suggest some kind of diversity which has disappeared. But it's interesting if we are already seeing some kind of streamlining because it feels early in the data evolution.

SPEAKER H

My thoughts are that it potentially creates some tunnel vision. And if I compare it to our social media feeds, the more we feed that with what we're watching, it does kind of narrow the field, what we get to consume, even though there's still a much broader selection out there. So I think that maybe part could be what that trend is explaining.

SPEAKER F

I might have an interpretation of that. And just like from our experience, we've been collecting, for example, tons and tons of images which have been super valuable, but we run those images through models and we get certain metadata about the image from those models. That's kind of one way where the diversity in data formats is going towards. And we use those like tags or columns for our analyses. Right. So we're taking a visual image and we're making it something more numeric. That could be another way that I interpret your statement there, Alexander.

SPEAKER H

Very good. So we're going to turn a bit to a question of investments, and I think that the answer will be a little different from Mark as I get a sense of data is part of their core business. Right. And finding that information asymmetry that edge. But still, how are you encountering unlocking investments within your organization, being aligned with senior management workflow, et cetera, around this new or developing discipline. As a data manager, data and analytics, etcetera, data scientist, are you finding resistance, or are funds being unlocked? Are people really on board with this? So go ahead, Mark.

SPEAKER E

But I'm going to attack this the way I want to go for it, which is from an investing in data perspective for us, then it's quite neat, actually, in that the data pays for itself. We'll do a three month trial, and during that three month trial, we'll perform a backtest, and then the backtest will sort of tell us how much money we can make with the data on an ongoing basis, which is great, because then when we're having the price negotiation with a provider, then we know pretty much what it's worth to us. And so it's kind of de risked for us. It's. We're putting a lot of effort and work into that three month test. You know, data scientists, quantitative researchers with PhDs are spending time on it, and a lot of the tests don't work out. But by the time we've done a test, we know that, but this data should, if the work's been done correctly, and then it should pay for itself, and then. So my price negotiation task is kind of simplified by that. So investing in data for us when the system works becomes a bit of a no brainer.

SPEAKER G

Sure. So investment in data. I think there's a huge enthusiasm in investment in data, for its analytics, for its capabilities. We do measure it with the KPI's. Also, right, to Mark's point, like maybe it's a three month trial, maybe it's a quick PoC, but other than that, also to measure it with the KPI's, such as what is the time saving and not just time saving, efficiency saving, and what is the dollar amount saving. So there's always some kind of a KPI metric dependent on your data infrastructure or depending on your analytics project. We always have some kind of a KPI metric to assess that data investment in a measure.

SPEAKER F

So at nature's find, it's kind of a greenfield opportunity because it's a newer company, a lot of younger folks in there, and everyone's interested in having the latest and greatest. So we have some buy in from the team members. A bit different than a lot of my prior experiences, but I will say kind of to Janine's point is thinking about KPI's for us, since we're a manufacturing quality place, does it improve our yield, does it improve our quality outcomes by having more accessible data, better data, and trying to tie a capability or a data source to that is really, really helpful.

SPEAKER H

Very good. Yes. No, what was that question I wanted to ask? Yes, we are going to turn to it, Janice. So we debated it last night, but every panel, of course, towards the end starts staring back towards AI. We are going to go there as well. But maybe we can take an angle, just highlight how AI is helping more productivity in your work and if you can be a bit more specific on it, application or some kind of hurdles that you're facing. And actually using this in the business takes time. Right. And Mark also highlighted that you need long history rich data. So what are you encountering around the theme of AI?

SPEAKER G

So on theme of AI, I think there's a lot of buzzword, there is a lot of potential. Personally, I think there's a lot of potential. I think we're already seeing much efficiency gain compared to how we were doing things traditionally in the past without AI on a personal productivity front, like a typical gen AI case, I think it provides a lot of value in three c's to create, to communicate and to compute. There is definitely a lot of efficiency on that front. And like the panel discussed earlier, in order to bring AI Gen AI to your organization now you're going to use one of the existing LLMs in the world. We're not going to create the LLM for our company now to your thing now. Next step is really are we ready? First of all, in terms of AI, are we ready? And I hear very unanimous answer from management to operations that they are now. Next step is are we ready in terms of infrastructure, is our data standardized, is our data structured, is our data warehouse, data lakes ready? So are we ready in order to integrate these LLMs of the world to our data, which is going to be more in house, we are not going to expose our proprietary data to the LLMs and make it publicly available. So challenge over there is now we want to, but how do we do it? Do we have the right skill set, right talent, and do we have the right culture in the organization also to move towards it? It is definitely a curve. It's definitely not a switch. It is definitely a curve. It'll take some time, but I think we are heading towards there.

SPEAKER F

I definitely agree with those statements. I think in the manufacturing space, when we think of AI, we have traditional AI, and then I would say the generative AI, which is what people are excited about nowadays. We certainly had a lot of applications in the more general AI use case, whenever you have a repeatable task you want to automate, or if you need to parse through tons of information. It's been very helpful in those areas. I would also include some of the computer vision examples I shared earlier in that area. In terms of the LLM aspect, we kind of experimented with a few retrieval augmented models. So those are types of llls where you're actually referencing specific parts and documents, and we've used that for standard sops and manuals that operators would use in the production line. It's somewhat exciting, but we're not really there yet in it being applied. The one thing I would just add is, is applications of AI in a lot of organizations have all been tied to cloud usage. This is one application where you have to think about a hybrid application, because you might have to do some of these computations in line, in the production line. So having that be part of your strategy is also important.

SPEAKER H

Thank you.

SPEAKER E

So the interesting thing from our side actually has been on the licensing side with AI. So when chatgpt launched, then obviously everyone got very excited. But from anyone who owned large quantities of data which was text based, I would say the description would be more like panic in terms of understanding the risks from their point of making their data available to outsiders and the risk of somebody training their model on their text data, and then as a result losing an awful lot of value. And so it became much more complicated to have these licensing conversations. Prices potentially went up, but I would actually say it was just a massive jump on the brakes from a lot of these owners of large quantities of text. But they also, depending on their kind of, their market approach, they also don't want to be seen to be kind of being left behind. So it's been interesting to watch as they've gradually been coming to terms with what does OpenAI mean? What kind of protections do they need from a compliance perspective in order to allow this text? What assurances do they need from us? So it's been an interesting, that's been the interesting challenge from my seat with AI is dealing with people who own lots of text data.

SPEAKER H

Very good. Thank you for those insights. So we're going to take this down back to Earth and talk some more about the nuts and bolts and the angle I want to take. This is the three of you are not necessarily a tech company, but you're dealing with a lot of technology. So how do you balance building things yourself? Partnering with vendors? A lot of the techniques are more accessible, but still you're building out a tech stack that maybe you need to maintain or not. So how do you balance that approach?

SPEAKER G

Sure, I can go first. So you're right, we're not a technology company, right. We are an agricultural commodity energy trading company. There is always a discussion of buy versus build. We all have a department. We want the high skill talent resource pool. Right? Is there a skill set to develop? Yes, probably. But do we want to develop in house or do we want to buy off shelf and customize for our need? I think it is use case dependent, right. Some of the P and ls we probably want to do in house, but maybe some of them mark to market. There is best of the tools in the marketplace which we can adopt to it and which we are. Right. So it really depends on the use case. And it is always a debate of buy versus bill. From a personal point of view, having worked in different organization and different projects, I'm an avid advocate for buy and customize on top of it for your personal need or for your organization's custom specific need. Rather than build from the scratch, why reinvent the wheel? Right? Take it off the shelf and customize to your need?

SPEAKER F

For us, I would say it varies a bit, again, because we're a bit more of a greenfield opportunity. We have a team of five data scientists and we're only a company of 200 people. So, relatively speaking, large data team, that's internal. I would have say our biggest learning, at least from this, is just having a really good partnership with our cloud partners has been extremely helpful. They've been tremendously helpful in allowing us to kind of think about how we set up our architecture and also the different tools that they have available for us to use.

SPEAKER E

We're a bit unusual in a very competitive market because we're sort of the only game in town in France for quant funds. So we have kind of, and I'm not speaking as a frenchman, but CFM has kind of evolved in a slightly separate. There's an awful lot of quant funds in New York, there's an awful lot in London, there's not so much in Paris. So we have got, that gives us some insularity, and it might have encouraged kind of a preference for building over buying. It's kind of in our ethos. And there's a feeling that if we developed it ourselves, if we own it ourselves, then we have something that other people don't have, which in a competitive world is very important. So, yeah, we would lean that way. I think when something becomes commoditized, if a technology becomes commoditized, and if you are just causing yourself unnecessary headaches building it yourself, when it's available on the market, and we do that too, but on the spectrum, we probably lean more towards build than buy.

SPEAKER H

Very good. So, mixed response. As expected, I've exhausted my questions and I know the audience has been listening attentively and imagine there are some questions out there as well. So, Russ, if you'd like to walk around and pick some people at random, if no one's sticking up their hand, and then my panel will gather their final thoughts here.

SPEAKER C

Firstly, thank you so much for sharing your views today. So I think a lot of you, like, shared about the challenges in historic data management, data infrastructure. So what are your views about managing future forward looking indicators? Because that is one of the most important things, specifically in time series analysis. And as history does not predict future, and how do you see, like, those kind of things coming towards, in terms of data management and specifically for you, mark, like, data sourcing, because there can be various sources for forward looking indicators. And how do you ensure that this data is like the most valuable commodity for like, all levels of management? Because I see that real time analysis is like growing up. And for that, involving people at the operational level, getting that data is the most important thing for forward looking indicators.

SPEAKER E

So if history doesn't predict future performance, then we're in trouble, because it's kind of how we try to work. So, yeah, we are essentially combining an awful lot of histories in order to try to understand a market compared to what has gone before. So forward looking indicators, I mean, forecasts are interesting, and if we've got a way of understanding of testing someone's forecast, then that's, then that's interesting and seeing how accurate it is. But no, we're not afraid of trying to use history to predict the future, definitely.

SPEAKER C

But that's what traders are afraid of.

SPEAKER H

Omar, I bet you have some perspective on that as well. Have you created that glass ball?

SPEAKER F

I mean, I could just speak from the perspective of more generalized predictive analytics, not really in commodities. You definitely want to be, you want to have that forward looking indicator, you want to be able to look at your forecast, make sure your forecast is in line with what you're expecting. That's all I could add to this question here.

SPEAKER H

Janique, to react.

SPEAKER G

Or, I mean, I'd echo to Mark's point of, like, we have historical, we have like multi, multi years of historical data, and these historical data should be able to provide you some kind of a future indicators. This will not predict the black swan events, not those exceptional cases, but other than that, we have had like a, several market updates and down cases, it should be able to predict or provide us some kind of insight once the black swan events happen, what the recovery curve would look like potentially. Right. So we should be able to do the forward looking analysis on that based on a sound, structured historical data.

SPEAKER H

Very good, very good. Must be more questions out there. There we go.

SPEAKER C

If such thing exists, how do you deal with fake or manufactured data, or simply incorrect data? How do you flag it? What kind of tools, how can you avoid to act on the data which might be wrong?
SPEAKER H

Actually, I really like that question because the word fake news came in my mind also yesterday talking about this. So indeed, how do you mine it? Correct and valuable. Go ahead.

SPEAKER E

So, a lot of my job is to try and reveal it by asking a million questions of a data provider and trying to reveal inconsistencies in what they're talking about. Generally, they won't lie about what their data provides or how they've gathered it, etcetera, but they won't also want to tell you the things which they're not proud of. So you need to keep asking the questions until they kind of, they're kind of forced to reveal it. But then beyond that, so that's the first line of defense, is my team, who are asking a million questions of them. And then the data is handed over to our data science and quantitative research teams, and they have a certain amount of internal data that they can reference, check against, and see if things stack up. And that's an awful lot of what happens in the three month trial is actually making sure that we can trust this data. I mean, to give you an example of point in time, then, if a merger happened in 2017 and it's a 15 year history on a dataset, then perhaps the temptation in an unrigous data set would be to just backfill on the post merger brand in a dataset, whereas we know we've got a referential back from 2013, which shows that there it was pre merger, and this is what it should look like. And so we can expose when there's poor data from our perspective. So there are various checks. There is also, once data is in production, then there are various checks to make sure anything crazy gets kind of flagged as well, so that it doesn't lose a hell of a lot of money. But so, yeah, it's something we take very seriously, and we've got various kind of levels of defense against it.

SPEAKER G

Others can react just to add to it. Right. The fake could refer to the aid fakes, but also maybe it could lead to our data accuracy, data reliability. Right. So when we talk about the data insights or analytics dashboard, it could be your insights for what happened in the market, what's happening, or it could be also an actionable insight on your data accuracy. If there's any outlier, that's also an actionable insight for you. May not be as interesting as the market insight, what happened in the market. But equally important, action item. If we have a data inaccuracy, if we have a data inconsistency, that's also an action item. And equally important, I would say I'd.

SPEAKER F

Quickly add, for areas where we control data entry, you want to do some data validation on the point of entry. So, example of operators in production line, right? Like if they input a number and this number is ten times the average of what it is historically, then it would flag that to the operator, for example.

SPEAKER H

Very good. So not all data is created equal for sure. Do we have more questions out there?

SPEAKER C

I think one of the questions you already answered, like, I wanted to know what would be a systematic way to.

SPEAKER G

Check that data is always accurately, like it's always accurate.

SPEAKER C

But I think you talked about the outliers.

SPEAKER G

For some time. What is my next question is that, let's say if there are delays in.

SPEAKER C

Data and our business decision is based upon that data, what would be the best practices to, like, get an approximate guesswork and give business what they want.

SPEAKER G

But also tell them that, okay, this.

SPEAKER C

Might not be accurate enough.

SPEAKER E

Is that about missing data?

SPEAKER G

Missing data, yeah.

SPEAKER C

Or delayed data?

SPEAKER G

Yeah.

SPEAKER E

I think that's a stomach question rather than mouth questions as you go.

SPEAKER G

I'll try. So we rely so much on data for our daily operation, daily decision making. In the perfect world, all the pipelines should work. We should get the data on a timely manner, as we have defined. Again, in the perfect world, sometimes the pipelines break or sometimes there is delay in you, always based on, again, to the earlier point, we have so much historical data. So based on your historical data asset, you should be able to have an estimate or some kind of a forecast, right? Even if your pipeline, even if you're expected data does not arrive on time, you should be able to have some kind of an estimate and forecast to what it should be. But in order to assure, like, we have a data on a timely manner, some of the pipeline should be. I don't want to sound like some of the other pipelines are the stepchild, but some of the pipelines are more critical than other, I would say, like having P and L at the 530, having P and L at the end of the day, position at the end of the day, some of the reports cannot wait. Day plus one and has to be d zero. Right. So some of the pipelines are more critical. So it comes down to your, again, digital strategy, digital roadmap of where your priorities are, what your priorities are. Right. Like it could impact the planned day to day operation, which means a significant loss of an operation. Right. So you need to prioritize and put, like, criticality on the pipelines accordingly, would say.

SPEAKER H

Anyone else care to react?
SPEAKER F

I guess. Okay, I'm another stomach on this panel, in Mark's analogy. So if like, let's say your data point that's missing is something that's historic, not something forward looking, there are different approaches to filling that data based on its own value and the value of other indicators in addition to it. So there's some synthetic data generation approaches that I'd suggest looking into.

SPEAKER H

What I'm hearing, basically, and it's a recurring theme, is that managing data is not just like putting soybeans in a truck and then moving it. It really is a data factory that needs a lot of hands on work. You need to be on top of it. You need to indeed identify anomalies coming in, fill the gaps, and making sure that the output makes sense. So, yeah, think about that, building your proper data factory pipeline here. Any other questions? I know we've gone a bit over time, but thanks to the other panel, not us, we're doing good time management here. But anyone else, no more questions. Okay, then I have a final question for this panel. So we went on a data journey. Yes. A very valuable, maybe most valuable commodity, but I'd like to ask them what besides data is your most valuable commodity? And I'll tell you mine. No, I'm not going to give it away yet. Go ahead.

SPEAKER E

Can I say love?

SPEAKER H

There you go.

SPEAKER F

You, Omar, family, slowly. Pickleball is taking over.

SPEAKER G

Yeah. For me personally, it would be family, obviously. And what I was going to say, if I were to answer first, from an organization point of view, my team, the resource. Right. I would focus on resource rather than the tool.

SPEAKER H

Very good. I like that. But yes, time with loved ones, I think, is the most valuable commodity. So let's park it here and go call the home front, check in and enjoy the rest of the conference. Thank you to my panel.

SPEAKER G

Thank you all.

SPEAKER A

Thank you very much indeed. Great panel. Great session very much. Encourage you to continue the conversation over coffee. So thank you all very much indeed. I.