Welcome to The Chemical Show, the podcast where chemical means business. I'm your host, Victoria Meyer, bringing you stories and insights from leaders, driving innovation and growth across the chemical industry. Each week, we explore key trends, real world challenges, and the strategies that make an impact. Let's get started. Today we've got a great conversation focusing on the use of technology and AI in business.
I've got Alan Spanos, who is the director of data solutions with ICIS and Chad Alan, who is the director of technology strategy at ICIS. We're going to be talking about technology, the role of AI in business and analytics. Risks and solutions and more. Welcome to the chemical show guys.
Thanks. We appreciate it. Thanks for having us.
Absolutely Alan, let's start with you Can you just give a brief intro to who you are and how you got here in your role at ICIS?
Sure thing. So I didn't start my career in chemicals. Um, I started in a management consultancy company After doing aerospace engineering at university And then I spent about 14 years In aviation in the uk in a variety starting with Moving on to data and data engineering roles after that and leading a internal practice, looking at that, within the UK's largest airline. And then at the end of the pandemic, I moved on to work for ICIS, um, performing the same kind of role, but for our customers.
So making sure they have what they need to do their job.
Awesome. Chad, how about you?
So I'm, I'm from a technology background, so I've been working, uh, in the parent company for ICIS for about 18 years. Um, I've been with ICIS for about eight years. Uh, so I've come from a computer science background and, just really into technology, uh, but, got into ICIS because of the interesting data and all of the very interesting problems you can solve here.
So your role today is director of data technology strategy. What is. What does that mean in the grand scheme of things?
Great question. Um, so it, it means, I do a lot looking at sort of the longterm view. Um, on, the business and how technology, helps support the transformation of the business where it needs to go looking at sort of emerging trends, and, you know, things that are going to make the difference both for us and for our customers.
Awesome. Well, that'll be perfect then for today's conversation. Before we jump into a little bit more about technology. Can you guys just explain a bit about who ICIS is and who they serve?
Yeah, I can jump in on that one. So, um, ICIS is, effectively a information and analytics provider for the chemical and energy industries.
We have a variety of different services and products that we offer from price assessments for different raw materials and commodities, but also analytical services for things like supply and demand, analytics, and price forecasts and margin analytics in the industries that we serve, um, also part of Red X, which is a global data analytics information provider, uh, working across many industries from health care, science, legal and data services, which is what we work in, uh, covering areas like
chemicals, but also aviation and other industries.
Awesome. Thank you for that. In fact, uh, I had not appreciated just how big the parent company was and all the pieces it touches, which is really exciting. And then, listeners of the chemical show will know that I have talked with one of your colleagues, John Richardson, from ICIS many times. So we'll, in fact, link to some of John's episodes on, uh, on our show notes.
And as we promote this episode, uh, because John, of course, always brings a wealth of Insights into what's going on in the world of chemicals and polymers and more. So let's just get into this in terms of what do you see as the role of technology in chemicals today?
From a business perspective, the industry is very competitive. So I think, what you see when we talk to our customers and the industry in general is that everybody's looking for competitive advantage. And one way of delivering that is trying to be as efficient and creative as possible with new technology that you're able to deploy.
Um, so we see that customers talking about fantastic innovation in their operational processes, making sure they can really get the most out of what they're producing and who they're producing it for. And also more recently, and in the space that we operate in, making the most of understanding the market correctly. So, um, how much should you produce? Who should you sell it to? How you should use it?
Those are all types of questions that we can help with, and we're seeing customers digitize those processes, and try to apply AI to them as much as possible to make better decisions and to really do things faster in such a changing environment.
Yeah, that's interesting. You know, you talk about technology. In fact, what comes to mind is I think about is the supply guys especially would run things like, an LP model, linear programming. And yet, I don't know if we talk about that anymore. Is that a predecessor to AI? Is it part of AI? Where does that, something like that fit in?
Yeah, so I think, um, in my whole career, probably everything I've done up until now, today we would call AI, and it's such an umbrella term for, kind of, lots of, in my mind, it's mathematics. So, you start from incredibly simple formulas, linear programs, or just basic, Mathematics you'd learn at primary school in effect in the UK, all the way up to the modern version of what you'd call generative AI these days, which would be much more complex and, I suppose difficult mathematics, right?
And I think the difference these days is we have so much more data available to pump into those models. And also the compute power is a lot cheaper. So often people will give you the kind of adage that when I was running an Atari. For my games console back in the day, my phone has got 10 times more processing power than that, or a thousand times, right? And I just put it in my pocket.
So you just have a lot more compute power and a lot more data and better mathematics to solve those problems these days. Yeah, and it's,
it's all, it all builds, right? It's not like, it's not like AI emerged overnight and it was, it was everything was a whole brand new, you know, so all those techniques, we would have looked at any kind of data analytics, historically, all that mathematics stuff. That is the. Core of what fed into AI. A lot of the core things around AI and how it works are actually very old ideas, but the technology in terms of the compute power and all of that just needs to catch up.
And once it became powerful enough in the cloud computing revolution happened, all of a sudden we could start just throwing tons and tons of processing power at solving problems. And that's over the years become more and more powerful to the level you're now seeing with like generative AI and really getting into what people have always thought of as being artificial intelligence.
Yeah, I mean, and in fact, I think when I first heard the term artificial intelligence, which was probably 20 or more years ago, I was like, what, holy heck, are you talking about? Um, and I think people still are. What the holy heck are you talking about in our machines taking over the world? So I think we'll get, we'll get into that cause I think that falls into some of the risks maybe later, but when you think about, what's going on with technology, what are your clients.
And how are they engaging? Because I think of ICIS as a data company. Um, at least that's my lens on it is that there's always this wealth of data that, uh, clients go to you. I go to you, other people go to you for data and information. So what's different today in terms of what. Your clients are asking.
Yeah, I think if I start with, under that umbrella of AI, kind of what you'd term as more traditional AI use cases. I actually run the strategy and work with customers every day looking at data solutions as my job title. And what does that mean? That kind of means. Customers that they, they're not comfortable or it doesn't meet their needs just to read content from our website.
They want it in bulk and they want to feed it into systems and they want to run their own AI models and things like that from it. What we see these days is that's becoming more common that customers in the industry want to do that. And probably the main driving reasons are we're needing to make decisions more quickly because the environment around us is changing. Us is changing and to do that manually these days is just not sufficient. You just can't run that in a spreadsheet anymore.
People want to be more efficient. As I say with the competitors in the industry. Everybody's looking for marginal gains in terms of shaving time off of processing and running their operational processes and with the data available, they can now make smart decisions as well. So they want to run smarter algorithms and against our data and their own. And that's really what we're seeing that, customers are pushing forward and those solutions are becoming more popular.
I think Chad can give you a good answer as well about kind of the more. Generate I kind of use cases and what we're doing there as well.
Yeah, I
think that's that's a different space. So I think the perception of us as a data provider. That's true. We certainly are a data provider and we do create that. But we do see ourselves very much in an analytic space, right? We do some pretty cool stuff there. The gen ai I think is a slightly different beast. I think the explosion of chat GBT and what that actually means because, it's really going into the whole thing of looking at individual productivity, right?
And in some ways it's quite difficult 'cause it's sort of like it can do so much, but, you know, where do you get started? So we're seeing like different levels with our customers in terms of like. We have some who are quite savvy with it and get it very quickly in that. And you can see them immediately saying, I use chat GPT to use this. I'm using Ask ICIS to do all of these tasks. And what they're getting out of it is a loss of productivity, right?
It means they spend less time doing the jobs than they're doing before. And in other areas, people are still trying to get their heads around it. The difference between using a chatbot based solution versus using a traditional search where they're still just trying to type in a keyword and just trying to get results. But, it is that evolving to that space of kind of asking questions and trying to have a conversation with, an artificial person, if you want. But, those who can really get it.
They can really get advantage and there's a lot there about the skill set of actually asking good questions is, is, is something that everyone really has to start learning in the world of the AI revolution, as I like to say,
absolutely. In fact, that whole aspect of prompt engineering, right? Figuring out what questions to ask, and then to keep asking. Um, I know myself just from my use of chat GPT over the past 2 years, and how it's evolved.
I can get to the types of answers I'm looking for rather quickly because I know exactly how to start asking the questions, how to amend those questions, how to keep driving it to a solution, but it's, you know, the first question you ask is usually, just like in real life, frankly, the first question you ask is probably not the right question. You have to keep asking questions.
Yeah, absolutely. That's very much the case of, advanced users with, with looking at using AI technology. They'll ask a question and if it doesn't give you the answer to the one, you just go and reframe your question a bit, you know, knowing is that it's smart and it has this amazing amount of knowledge, but it's not always that smart and maybe it just needs a little more context and help from you to, to get the answer you want.
So let's talk a little bit about Ask ICIS. So I know that you've recently rolled out this, Ask iCIS, which I guess is an AI based tool, to really, I guess, interrogate the data and get answers. Right? So can can talk about that and how you guys are using it.
Yeah, Ask ICIS is our generative AI assistant, as we call it, that is sort of a. specializes in these sort of the chemicals in the energy sector. And it's there to, you know, like a chat GPT, you can ask it questions and it will answer those questions for you. Unlike chat GPT, it's not just a general purpose tool. You know, it's not there for answering every question. It's focused specifically in the markets and it's built on the data and content that we're taking and create with an ICIS.
So when you ask a question about, like what's happening in the styrene market or what's what's going on there. It understands more context about what styrene is and what you've used it for and it's pulling content from our price assessments or forecasts or from from our news to help answer those questions. And it gives you a much more powerful answer as a result.
So what is it? So what are the use cases? How are you seeing your clients using it and really getting value?
So I think the big thing there is people are looking at a really key time saver, right? It's that they can get the job done a lot quicker than, um, just going and doing the work themselves. You know? So we have some people who are saying that they, they go and use it and they can save hours of work.
They normally take them hours to go look at our content, look at other content elsewhere, and ask that, and now they can just have a single session with Ask ICIS, ask some of those questions and get what they need, you know, and it accesses all the content we have around market dynamics and companies, topics, events, supply and demand fundamentals. And so
I'm assuming is it, I'm assuming it's available only to people that are already subscribing to your services and data. Is that right? Yeah, that's correct.
I think that the important thing that we do with Ssk
ICIS,
that, Some of the like, Bard, I think does it now from Google and some of the other Gen AI tools is originally they didn't tell you what information they were using to give you the answer. So, in effect, they would just say, hey, here's the truth. And a lot of the time it wasn't right. It was a lie, right? It was the AI getting it wrong.
And what we've done from the beginning and some of the learnings we've taken from our sort of sister companies across Bellex was ground the answers in the articles that we already have. So. We'll give you an answer and say, Hey, you've asked about styrene in this market. Here's what we think is going on. And here's the relevant articles that we've used to create this response. And then you can click through if you subscribe to those and see.
And if you don't, you can talk to us about if you want to get a subscription to those articles. So trust for us is a really big thing in using AI, and I know we'll go on to talk about that. But I think I would never trust the answers if I didn't know where they came from. And that's really
important. Yeah, that was when we were developing Ask ICIS. That was really the heart of where we started, which is, you know, our customers really value the content we do. They know they can trust us. They know we do our research, we do know what we're doing, and the tool must do the same in that sense. So, it's very much built into the actual design, to make sure that we know that if you see it, it's something that somebody in ICI has said.
Because that's, trust is the absolute most important thing, and when you just use a chat GPT or looking at that, they talk very confidently, right, when you use these services, and that always gives you the suggestion that is, it's always right, but it's not. And it's very hard to tell when it's right or not. If you can't go and say, can I just look where that content came from to see if that's correct.
Well, and I really appreciate that as well, because especially the referencing back to the specific articles, because as we know, things change over time. Right? And so when you get an answer and say, Oh, the answer is X. Well, okay. The answer X was correct back in January. But these other three big things happened and now the answer is Y. So understanding and being able to kind of trace some of that I'm sure is super important to your clients and really to everyone. Right?
Because we, when we ask a question, and somebody is coming in to Ask ICIS and be like, Oh, I'm asking this question because I have to go talk to my president, the board, the boss, a customer, whatever, you want to go back to that individual with confidence. And I think that's a, that's
a common, requirement, I think, for any analytics or AI back from before people talked about generative AI, you know, we got some great. Data science teams working on price forecasts as well as what we do with price assessments and one of the key things that we work on with customers is explainability of those analytics because As you say if you're using a third party to create them If your director or cfo or cmo asks you a question about well, how did that work? Get to that.
That's a strange number. I don't understand it.
You
need to be able to explain that to them. If you can't, you lose trust immediately. And I think no AI tool is really going to be used by customers or any users if it can't explain itself. It's very important.
Yeah, that's great. And that's actually a great segue. Let's talk a little bit about maybe some of the concerns and the risks that we think about as we think about technology and AI and how those are being mitigated.
Yeah, sure. I mean, the big, the big thing that people worry about is hallucinations, right? It's the phrase everyone's kind of using. It's a bit of a disservice in some, some cases to the machine. It says it's hallucinating and that it's just hallucinating. They're actually making very, very, very educated guesses, right? That's how they actually work. But the one thing that people worry about a lot is if it goes and tells you so. Right? Can you actually handle that?
Can it actually manage that hallucination? Yeah. And that's something as far as how we, we took and built it in a pattern, it's fairly common in the market to make sure that, when we train, um, Ask ICIS, we tell it that it prioritizes our view. In our content right over any other view.
So it's it's always pulling from our content first, which we make accessible to it and effectively does a search and pulls back the information and then it tries to answer the question that customer has based on that information. It looks at and it's it's not drawing on as much the the LLM. To answer that question, it's using how its ability to read and its ability to reason to help answer that question. But the data it's pulling from is ICIS content.
Got it. So it's, so it's a little bit less about generative, right? So I think this aspect of generative being creative and so creating answers that don't exist, but rather generating and then testing back to what is already known to be true.
So it's, it's generative in the sense and it's creative in the sense of it makes content that's more specific to what you're trying to find out. Right. But it could be made out of six pieces of our content. Right. So it could have pulled it from three news articles, a price report and forecast because you are asking about something that hit all of those areas. And so it, it then it creates a unique piece of content for that particular answer, but it's all based on our content.
Yeah, that's helpful. What are the other risks that you see or that people concerns that people raise?
I guess one of the other ones that comes across a lot is the, is AI going to take my job? That's the, the big one that, that most people say.
Yeah, do we still need the ICIS consultants if we have this?
Certainly it would be my answer. And I think, I think you need to look at AI and the tools there. They make it easier to do your job or to do somebody's job with that, but it's not just a like for like replacement, like really, our content creators in our business, they, they answer very hard questions, right? They're looking at very complicated situations and apply subject matter expertise. Most of that's not in the question that's being asked, right?
So when we actually look at the content, we're looking at these things, There's a lot more there to add that. So what element, you know, um, it's fine to say that as a piece of information, a tornado wiped out interstate 95 when Hurricane Milton, was hitting Florida. That's great. It's more important to know that that's a major, uh, Transport hub to the port of Tampa, right? Which actually could potentially is going to affect some customers there. So that's hard to figure out.
I think as well with any AI, not just the sort of more modern generative AI. It's really a lot of the cases only as good as what goes into it in the first place from an information perspective. So we're in a privileged place. I think in ICIS that we have a lot of good information that we can feed into these things and actually, Make them more intelligent to help you make a decision.
I think as well, like with, not just Ask ICIS, but the, the tools that we create, but also internally, how we use a I, um, a lot of it is around in the industry. For sure, and every industry has this challenge is how do you scale up your workforce? And how do you make them more literate to use these kind of tools?
Right?
And both from a benefit perspective, as Chad was talking about from making you more efficient, faster doing your job and so on. But also the risks and pitfalls very, very early on, like, you know, because we're At heart data analysis company in a group in general, we had guidance around employees using generative AI tools. Don't put confidential information in chat GPT. It's just going to harvest it and start spitting out to other people and
I'm
sure a lot of companies their employees are doing that and they don't realize. So, um, yeah. Making sure that you get good training out to your employees, that you teach them what these tools can do, why they're good, why they're bad, or what are the risks around them is very important. And we have to always keep up to date with the latest and greatest new tools that are available to do that.
Next year there'll be another Chat GPT like thing that will come out that we've got to make sure that we use it in the right way Like I like to think of AI analytics tools like a builders toolkit. Someone gives you like a brand new drill You don't just Try and do something with it straight away. You read the manual, right? Like, so, try to make it, you should do that. Or you drill a few holes in the walls and see what happens. I mean, hey, maybe, maybe I'm that guy who reads the manual.
But I think it's important that, they're fascinating. They're brilliant. They're, they're important tools and they are. Chad and I will often geek out on, it's a very exciting time to be in business to use these tools in your personal life as well, but you should make sure that you do a little bit of due diligence to understand which tool to use at what, uh, in what way, and also, I mean, not just risks, like there's a big risk that you run up with lots of costs.
Some of these tools can be quite expensive. So if you can solve, um, my mantra has always been in the AI space, right? If you can write a linear program to solve a problem to a level of accuracy, that's acceptable, why would you deploy a neural network? It's like 20 times more expensive, right? So having that cost benefit analysis, when you're deploying these tools is important from a business point of view as well.
Yeah. Well, especially I think when you think about the. Maybe the broader social context of it is the power Usage that goes along with AI as you say the these sophisticated neural networks I don't need to ask it how to make a peanut butter and jelly sandwich Because I can get the answer anywhere else But asking it things that I can't find the answers to becomes critical so that I'm using the resources wisely Yeah. A hundred percent. Do you, folks, you talk a little bit about training.
Do you guys provide training to your customers when they start using Ask ICIS? Because obviously you want them to use it effectively and get the benefit.
Yeah, absolutely. So, so we have a customer service team in effect or customer success team, which not just Ask ICIS, but they're available to our customers for any product that we ship. As part of onboarding and general queries, if anybody has any questions, uh, we can do that as interestingly, a use case for those types of algorithms that you can ask them how to use them as well. So it's
actually really good. I hadn't thought about
that. Yeah, Chad, just don't Ask ICIS specifically anything that he helps.
Yeah, I think it's, it's something I know that as we develop it, it's always about making it easier and easier. I mean, I think a lot with these chatbots is trying to make them as simple as possible. Like it is just a box, you type in a question and you try to have a conversation with it. But I think, there's always elements. How can we make it easier? How do we help people to onboard? And that's something we're always looking at.
And we can evolve things, the tool and how we approach things, of course, too.
Awesome. Love it. So what's next guys? What do you see as So Ask ICIS is the big thing for 2024 and obviously continuing to roll it out into 2025, which is when this episode is going to be getting published. What else? What's next? What should we be looking for from, from your team, from ICIS?
I think more innovation, I think, is what we, we want to kind of promise to our customer base and ourselves that we're going to keep trying to push the boundaries of what's possible and what's sensible in our space as well. And I think whilst my head of product would kill me if I said anything specific, from my perspective, that's more investment in some of these tools, like Ask ICIS, growing its capability, but also in my space.
new analytics for customers to use to help them make decisions across our industries, but also keeping up to speed with developments in the software and analytics market if they want to take data directly from us as well. We've done some investments on that recently as well. Um, and we need to make sure those tools are up to date when our customers come asking us to help them. So yeah, keep watching us.
We'll try and post as much as we can on socials and things like that personally and as a group. And, Yeah, I can't be more specific. I'm afraid that if you want to be braver than me, but I think it's,
you know, as always, it's with with all generative AI tools that people use today, right? It's not always going to be about specific features or things you're looking at, but it's just, it's a just how do you make the experience better? It's a lot of a lot of fine tuning. Of how things work or behave and respond and looking at particular use cases on how does it answer particular kinds of questions or things like that. Right.
So it's it's it's hard to say exactly what's there because we're looking at, feedback from customers and seeing what's important. And, that's what we're looking at. Yeah, that's what drives.
I think what I would say is, well, if you want an advanced look at any of the stuff we're developing at the moment, we have like a custom advisory panel. You can apply to and we'll be testing like tool prototypes and things like that with with those customers. So, um, how does somebody
get on that panel? Is there a website or an email or can I just how do we direct
people to you? Yeah, we can, we can supply that if I don't know if they're still recruiting people, but I can give you that information. Well, and,
and for sure, we're going to include a link to Ask ICIS. And I know if people are already ICIS customers, they can probably talk to their salesperson, the relationship focal point to learn more. Awesome. And I think, uh, you know, if nothing else, as we head into 2025, learning how to use AI effectively, including tools like Ask ICIS becomes a real differentiator for companies and for individuals.
Yeah, absolutely. There's, there's a lot. I think customers, people working in this industry to think about about how, how can we use this technology to help improve our productivity? How can we help our individual people get more productive? Time to do the things that add value to your business and less time doing, you know, some mundane tasks. That's where these tools are very powerful to get people to focus on the stuff they tend to like doing more as well.
So, we honestly, we feel like these types of tools and us as a partner with our customers, we want to augment what they do. We want to make them better. Ask ICIS as a name, even think of us as basically, you're the superhero in our story and we want to be a sidekick, right? We want to help you, we want to help you get done what you need to do, and we want to make you look great to your boss. That's effectively what we would like. And that is a
great way to end. So, Alan and Chad, thank you so much for your time today. This has been great.
It's a pleasure. It's a pleasure. very
much. Absolutely. And thank you everyone for listening. Keep listening, keep following, keep sharing, and we will talk with you again soon. Thanks for joining us today on The Chemical Show. If you enjoyed this episode, be sure to subscribe, leave a review, and most importantly, share it with your friends and colleagues. For more insights, visit TheChemicalShow. com and connect with us on LinkedIn.
You can find me at Victoria King Meyer on LinkedIn, and you can also find us at The Chemical Show Podcast. Join us next time for more conversations and strategies shaping the future of the industry. We'll see you soon.