Season 1 Episode 13: Chris Wallace, InterGen, on the Possibilities and Limitations of Data and AI

This week we are joined on the Hypercube podcast by Chris Wallace, Head of Data and Analytics at InterGen, to get his take on how data and AI are impacting both the organisation, and the wider industry.

Chris takes us behind the scenes at InterGen and talks through how the data team is working closely to support the trading team and the business more broadly. He cuts through the hype of new technologies, focusing on where the application of established technologies like predictive modelling and automation can have a tangible business impact. He also highlights the limitations of AI tools like ChatGPT, emphasising the importance of trust and confidence in the models being used, and gives us a frank and realistic assessment of the current landscape.


In this episode, we covered:

  • Taking a cautious approach to cloud migration.
  • The challenge of finding the overlap between on-premises and cloud skills
  • Which data tools are currently adding the most value.
  • The need for data teams to educate the wider business
  • Delivering data tools that work for the end user

The weekly Hypercube podcast sits down with leaders in the energy and utilities sectors to explore how data analytics can help businesses make smarter decisions and accelerate business growth.


[0:29] Chris gives a quick introduction to InterGen and his role with the company.

[2:42] Chris explains the need for caution when tackling cloud migration in the energy industry.

[4:00] Chris discusses some of the skills challenges in the energy sector and attracting the right talent.

[5:18] Adam asks if there is anything now holding back the energy/utility sectors.

[7:52] Chris shares his perspective on the current use of AI and data science in the industry.

[9:31] Chris shares his view on educating non-technical stakeholders on the limitations of AI.

[13:16] Chris highlights the need to be cautious about what can be achieved with LLMs.

[15:06] Chris talks about some of the ways AI and data are being used to support energy trading.

[20:53] Chris talks about implementing tools that work for the end user.


Hypercube Podcast Transcript

Title: Chris Wallace, InterGen, on the Possibilities and Limitations of Data and AI

Host: Adam Sroka
Guest: Chris Wallace

Adam Sroka: Chris Wallace: 

Intro: Welcome to the Hypercube podcast, where we explore how companies in the energy and utility sector leverage data analytics to make smarter decisions and accelerate business growth. I’m Adam Sroka, founder of Hypercube, a strategic consultancy that supports asset owner-operators, traders, route-to-market providers, and energy services companies to unlock the power of data.

If you’re interested in hearing real-world examples of how data and AI are advancing the energy sector, this is the show for you. 

Adam Sroka:  Welcome back to the Hypercube podcast. I’m delighted today to be joined by Chris Wallace, head of data analytics at Intergen. Chris, do you want to explain to us who you are, and what you do? What do Intergen do? 

Chris Wallace: Yeah, absolutely. Um, Chris, I am a head of data analytics at Intergen. We, I guess, Intergen as a company are. Primarily a power generator. We’ve got power stations in England and our headquarters in Edinburgh. And in our headquarters, we’ve got a large trading team and other business functions and the data team that I head up, we support a lot of the activities of the trading team and other parts of the business as well.

And that’s supporting kind of using AI and building models and data engineering and BI and.  All the things that go along with reporting. 

Adam Sroka:  Yeah, you must have quite a broad remit across, like, the strategic, the technical, the commercial scope. How do you manage that?  

Chris Wallace: Yeah, so I mean, the data team within Intergen is relatively new, so we are still building out that portfolio of things we’re involved in.

Largely, we can sit within the part of the business that contains a trading team. And that’s like the, by its nature, the kind of the things that they do, they’ve used lots of data and code analytics to some extent for quite a while. So there’s a kind of natural place for that to grow from. And other parts of the business, OD, they’ve kind of maintaining and managing the plant.

They have used systems for specific purposes for a long time. They tend to be focused around knowledge-led things though. So if you know a particular Okay. Problem can occur for you looking for a particular type of scenario. You can track that over your data. But one of the things we’ve been looking at, um, working with them on more recently is trying to do some of that using things like AI, but it’s definitely a kind of work in progress as building out that capability around the business. 

My experience of these places is lots of on-premise, like legacy systems, that kind of attitude to some technology. Do you see that? It’s like the move to the cloud prevailing.

Adam Sroka:  How’s the tech landscape? 

Chris Wallace: Yeah. So, I mean, the move to the cloud, it definitely happens. And it’s happening with N Integer and,  other companies in the industry.

But I think like you point towards, there’s a reluctancy to change things too quickly, given, I mean, the assets Intergen operate and similar companies that think of these as critical national infrastructure to, to be able to power podcasts. So you want to make the changes in a way that’s as cautiously and carefully managed as you can.

And obviously like as on-prem things become more expensive and maintained and hardware becomes more of a niche thing and played. costs become cheaper, then yeah, there’s definitely a move in that direction. But yeah, it’s done cautiously and it’s done kind of as and when. And really the, the downside for companies like Intergen to do that slowly, there’s not really a downside.

It’s a lot, if it’s functional and it’s doing what you need, then that’s great. And then the penalties on the other side, if you do something That doesn’t happen the way it should, or you expose yourself to any cyber risks. And there’s directives that encompass a lot of that for the kind of civil purposes in the UK.

And the penalties associated with getting any of the cyber aspects wrong can be devastating for a lot of companies. So doing things cautiously, carefully, slowly. It makes sense for us to stand in for everyone, I think, in that area. 

Adam Sroka: This may be a thorny hot topic, but what’s your take on skills and talent and like bringing in the right people and expertise and the trends towards like new technologies, cloud-first, open source, all that, like really open blue sky stuff that you might see in Starbland versus your needs to maintain kind of untrendy technologies. Is that an issue? Would you see it becoming an issue? Are you going faster to counteract it?

Chris Wallace:  Yeah, it’s definitely a challenge to find that that skill set that overlaps with the on prem management of database services that may be a few years old now, but also the skill set to work in the clouds.

So I mean, yeah, that was absolutely challenging, probably more, more acute than it is if you are a startup and everything is in the cloud and you can hire a data engineer who only needs to know how to do things in Azure. Yeah, definitely for us, we need people with good skill sets to straddle both sides of that.

And I think we’ve done it successfully, and you can,  it’s knowledge that you can build. I mean, a lot of the things that are in the cloud are not exactly the same as they are on-prem, but a large proportion of the content is similar, and it’s more understanding the infrastructure than it is a SQL server.

It could be here, it could be there, it could be in another country. 

Adam Sroka: What would you say then, like, is anything holding the energy utility sector back, not just specifically Intergen? Do you see, like, what’s holding the pace of change back? Is anything impacting, like, yeah, the ability to drive positive change, more revenue, lower risk, all that? 

Chris Wallace: Yeah, I mean, obviously I can speak to what other companies are thinking in the same space, but I don’t think there’s necessarily anything holding back in the sense that regulation doing things cautiously, because it’s such a, an important part of the national infrastructure.  It’s probably the right way to do it, and yeah, I don’t think there’s specific technical challenges necessarily that are insurmountable just now.

It’s just a case of doing them, doing them properly and making sure that they’re as robust as you need them to be. And obviously financially, economically, if you’re a business, you want to make moves where it makes sense for you. If you’re sure there’s a vested edge to having some of these things move faster, probably they would.

All right. That balance with doing things cautiously, carefully, you know, it’s a kind of, I think just different businesses, different industries, kind of by the nature are going to do things at a slightly different pace. I’m sure we come back 20 years from now and follow up podcasts,  almost everything will be in the cloud, apart from probably the physical infrastructure of connected things you need to dispatch.

Adam Sroka: Well, that’s So, it just makes me think, it’s quite a refreshing answer that, because, I don’t know if it may be me, but it’s the worst out of people, but it’s rare to get a techie say, no actually, I think we’re at the right pace and it’s moving right. I guess that suggests that there’s a pretty good culture and like a feeling listened to by the non-technical parts of the business, which is how that’s true at IntoGen, and if so, what’s the secret and how do you replicate that elsewhere?

Chris Wallace: Yeah, there definitely is a culture of wanting to look and see what’s next, how to do things and the unique performance side of things, smart engineers looking at data, trying to make decisions better, do things practically efficiently. And the trading side of things, there’s lots of smart people thinking about building models, and it’s not a case of you get to a point where you’ve got all the tools that you need right now, you’re making good trading decisions, you’re making a profit, you’re happy,  constant like in the environment of constantly, okay, right, what’s the next thing we can look at?

What can we build? What can we improve? What are other people doing? And I think you can do both of those things with maybe a slower overall move into the cloud, because you can do elements in the cloud that have. You can institute infrastructure that connects back private cloud and secure ways of getting small pieces of things done if you need resources that are there that you don’t have, without doing everything all at once.

Adam Sroka:  What do you think are some realistic applications of AI and data science kind of approaches in the energy utility sector? And where’s the gap between like, realistic view and the hype? 

Chris Wallace: Definitely you get questions from people regards to JetGBD or GoPilot. So are there any of these kinds of early learning things that are on BBC News, that get in people’s social media feeds? 

And I’m sure these things are delivering value in lots of cases, and I’m sure there’ll be situations where we use them as well, but in, kind of, the applications we’ve been looking at, and probably in the energy utilities industry, there’s so much scope to use things that are a few years old, ten years old, more than that.

To drive more tangible things, I think, so I don’t know how long chat GPT has been a publicly accessible thing, but building core part of a business and making critical decisions based on something that’s a year or two old feels, cause it makes me uncomfortable. I’m sure tools like this will definitely have their place in the future, um, particularly when it’s looking at unstructured data and wanting systems where you can have a model that gives you an answer and kind of perfectly formed sentence.

But. Yeah, for us, it’s looking at either automation or scaling things that humans have to do manually or we can simplify or change building models that can help you predict or forecast things, classify, maybe not the kind of things that are going to be keynote Databricks or Microsoft Summit, but the amount of value you get out of smaller things, especially when you’re starting to build out this kind of capability in a business, sir.

They’re definitely the place to start rather than immediately just get an open the eye if you can and throw questions at it. 

Adam Sroka: Do you have any tips or tricks for aspiring or struggling sort of techies?  that are having to manage expectations then to the business. Like, do you have a stock answer for what are we doing about chat GPT? Or how do you simultaneously raise the kind of education level of those non-technical stakeholders, but also kind of bring them back to reality a little bit and keep them enthused? Anything there?

Chris Wallace:  Yes, uh, it’s important. I guess it depends kind of like where your, your team is or how mature the capability is within a company, but you want to make it clear that you’ve kind of got visibility of these things.

You’re up to date on what the latest tools and technologies are that they’ve had or that they can relate to. But it’s about building kind of confidence and trust in AI as a whole concept. I mean, when you said the E words, that there’s such a fuzzy and encompassing set of things for situations like that.

So it’s trying to find specific business use cases and specific asks and problems and breaking them down to just the most fundamental,  Thing that you’re trying to solve and then saying, well, is it a technique or an approach we can use to solve or support with that problem? It’s not an LLM because that’s for a specific set of use cases that are actually much more remote from the specific question that you’ve got.

And also, in the cloud, you, if AWS or whoever you use, these things are generally available with not a vast amount of setup and infrastructure. If you’ve got situations where you need to use these, you can make it clear to commission your leadership. Yeah, if we’ve got a use case for this, we can definitely pilot something quite quickly, but in terms of bang for a buck and solving the most tangible problems first, yeah, we, we start lowering it further back in the evolution of, I would say, AI one more time. 

Adam Sroka: I want to ask a follow-up question around, uh, do you have like a big tactical bet that you think, like, what are you steering your resources towards in your team? Like, what do you think is going to unlock the biggest chunk of value? And how do you get people on board and get by and get people excited for what can be quite an abstract or like weird, nerdy, techie kind of topic sometimes?

Chris Wallace: I mean, people will get excited if you interact with JGPT, Yeah, this is cool, this is something from the future, this is something from a movie. But it’s not going to solve or necessarily improve their day-to-day, or their objectives, or their bottom line of trying to make money out of something. It’s tying it back to a benefit that they can see tangibly.

Usually that will get people excited. Maybe not that little bit of hype or, I don’t know, quick buzz that you get from interacting with an LLM But you can show somebody that you can deliver a significant benefit or significantly improve something that they’re doing that they want. That’ll get them excited.

It’s just maybe a small bubbling excitement.  

Adam Sroka: Yeah, okay. I get that. Yeah, I just want to speak in their language at their level and solve the problems that are like at hand as well you can get carried away with like potential futures but actually, the day-to-day is hard enough sometimes and like these are clever people with hard jobs and there’s things to be done and sort of low hanging fruit to be won I suppose.

Chris Wallace: Yeah there’s nothing wrong with necessarily having an idea of longer-term projects or ideas that people have that you could potentially use these newer things with and get them excited and build interest there. But unless you deliver something in a reasonable amount of time, people become jaded pretty quickly.

And why am I spending money looking at this tool? Or you were talking about this tool two years ago, but we’ve not built anything that uses it. Yeah, it’s a balance that you have to work, but I know lots of people that will be massively excited if I show them a time series forecast that’s a couple of percent better than the one they had last week. And that’s, that’s excitement for them. 

Adam Sroka: Are there any applications of, or anything you’re doing at Intergen to help with the navigation of like policy changes, market dynamics, like some of these complex systems that are really poignant to what you’re doing and how you’re driving value?  

Chris Wallace: Yeah, nothing that we are kind of using like operationally.There can be kind of imprintable ways that you can use some of these tools to summarize very large documents potentially, but that, you still, that’s still a can reviewer to do that and then have a good bit of confidence that it’s doing it properly. If you can have potentially things on the horizon where there are large volumes of documents or things that are updated.

Scanning through them to pick out key things, but I mean a lot of that you could probably extend back and say well that’s kind of natural language processing along the lines people were thinking 15 years ago, whether you trust in the LLM to go through and draw truthiness conclusions that you would act on, I think that’s, that’s a little bit, I think it all comes back to like what kind of impact or risk is associated with the decision you’re going to make.

from the output of this model. And I mean, if you’re Facebook LLM and it just your site very slightly so that you make 10 cents per user less on advertising in a month, you don’t really care because in aggregate, you’ll make lots of money elsewhere. But if you’re a company, and Interjet are a huge company and you’re making billions of decisions, but thousands of decisions, then you want to make sure each decision is as good as it possibly be. 

Especially when any one of those decisions can cause catastrophic headaches for. The business, the shareholders,  the markets, I guess, as well. 

Adam Sroka: What about some of the kind of harder science y stuff? Like, I always think if energy trading, there’s going to be risk and lots of big brain stats and stuff. Do you see much into play there? 

Chris Wallace: Yes. I mean, that’s a big focus of the detail projects that we do, is with the trading team. And obviously lots of people in there already got strong math stats, and programming backgrounds. Also, actually, kind of variety of different other backgrounds are suited to the the personality types of trading, but yeah, what sort of scenarios and problems.

I mean, the amount of data that’s available, it’s electricity industry in the UK. So regulated, the amount of data that’s produced, even before you get to third-party or external sources is huge. So there’s almost like a. We’ve certainly got a pipeline of ideas and POCs are constantly moving to kind of investigate and work on.

So yeah, no end of demand there.  

Adam Sroka: And I guess the, for some of the techie folks that might be listening, any like favourite tools, models, libraries, any sort of technologies on your immediate radar that you’re, you’re thinking could drive a lot of results for the team? 

Chris Wallace: Yeah, I mean, a lot of the tools that we use, think like lots of these problems, you always want to start off with the simplest type of model first.

I know there’s been a lot of chat recently, the foundation models, someone by Bruno Reynolds singing or dancing or time series model that you feed your time series data into and I’ll give you a fantastic prediction of what’s going to happen looking at these things. But we’re kind of aware of this that so much and.

The energy industry is driven by external factors, so you don’t want to be looking at a time series one thing, you want to be looking at a time series that’s got 10, 20, 30 features that are all related in some way. Thinking about things like that, but a D to D building models using the kind of the usual time series kind of techniques.

I mean, lots of times you will explore different approaches and you end up back at boosted trees and you think, well, why did I try anything else? And for us, some of the challenges have been like understanding what kind of time horizons you look back at. Rather than necessarily the modelling, it’s pretty easy now to take a bunch of different models and benchmark them all relatively quickly, but it takes a bit of kind of human analysis and chats with the training analyst teams to understand exactly what data you should be using.

Adam Sroka: And you said there, it’s relatively easy, but that takes a lot of like investment in a, in the right backtesting frameworks. That sort of stuff to do that kind of analysis, right?

Chris Wallace:  Yeah, sorry, I should have said it was relatively easy the 100th time. The first time it takes a while because you’ve got to build your pipelines and where your data comes from and how your models are run and orchestrated and where your results are stored and how you calculate all your metrics and compare them.

So yeah, in that sense, there’s a bit of work upfront. But once you’ve got that kind of template and you can decide which features, which timescale, which type of model, What type of parameters you’re interested in chaining, you can automate that quite a lot. And then you can go to have lunch and come back and see what’s looking interesting.

Adam Sroka: Do you liaise with traders to iterate on what you’re doing quickly? Because like trading is hard, right? And it’s like, I wouldn’t say I deeply understand it. I know what they do, and I’ve spoken to loads of them over the years, and I know how they work, but I wouldn’t say I could do the job. On the flip side, the same people always say like, that scientists is really clever, blah, blah, blah, techie.

So you’ve got two really technical fields trying to collaborate on a hard topic, like sort of fuzzy bets and answers that are easily measurable. But what’s that process of like, Trying stuff out and bringing them, like, working with them to develop new things? How does that work? 

Chris Wallace: I guess you touched it up there, it kind of blows my mind still.

The knowledge that traders will have of the market and of everything that’s happening and there’s a reason they sit in front of ten screens because there’s ten or twenty or thirty different things going on at the same time. So they come at things with a kind of fundamentals approach, so thinking of like mechanics of the market.

If this happens over there, what the consequences are going to be here and here, and what will that mean for our price or for any other number of things. And usually the kind of things that we get involved with working with the trading team are where There’s something that they’re doing and they feel they could be making a decision that either incorporates other information or there’s a point at which they don’t know how they should be making a decision.

Like, this is kind of what we do, it’s part of the process, is there a way we could be making a smarter decision here? Or somewhere where there’s a decision to make and it’s not quite clear. The kind of the human or the market fundamentals rationale for making one decision or the other. And often they’ll come to us and say, Oh, we’ve got this kind of scenario that we think we’ve got a kind of gut feel that there’s a an opportunity if we can make a better decision here, it could be good.

And these are the types of features or types of data that we think potentially are related to what’s driving the underlying thing here that we’re trying to understand. So that’ll be usually the starting point and you shape up something like that. You get a scope of what kind of data is we involved.

And then There’d be a bit of either a data engineer bringing some of that in, and then the data scientists digging away, trying to either kind of validate some of the hypotheses that the traders have, or alongside that do a bit of analysis to see statistically does anything stand out,  and then it’s talking to the trading team and the analyst team again to understand these are kind of what the things that we’ve seen in the data, this kind of, feel right to you?

Does this make sense? And then from there, narrow it down and then think about a process that we can ultimately tie back to something that they can use day to day. You could go away and build a model that looks quite, quite interesting and theoretically pretty cool, but if it’s not something you can operationalise and have in front of somebody to use, then it’s not probably worth developing.

Unless you’re going to write an academic paper on it, maybe.  

Adam Sroka: Yeah. And developing the interface of like, how is this new insight going to be consumed and used is almost harder than actually developing the insight. Like, because depending on the cadence of which they’re making decisions or like how they’re operating, that the specific individual, right.

Sometimes they got preferences and quirks about how they work and you’ve got to kind of adapt to that.

Chris Wallace:  So it’s, yeah, I mean, definitely. Yeah. There’s like a data science tendency to like, Oh, I could present this. Amazing model to you in 12 graphs. There’s only 12 graphs and they all look amazing. And then you talk to the end user and they say, Oh, could you give me a table onto the 10 values and that’ll be fine for me?

Adam Sroka: Just want the screen to go red or green. And that’s it. Now the other quirk about traders talking about stuff they know, always know the weather everywhere at all times, which is lovely, but it’s always handy to just have human sort of weather charts walking about with you. Well, look, Chris, it’s been a pleasure as always.

Thank you very much for joining us today. 

Chris Wallace: Oh, thank you. 

Adam Sroka: And look forward to catching up again soon. 

Chris Wallace: Excellent. Cheers for having me.  

Outro: And that’s it for this episode of the Hypercube podcast. Thanks for tuning in today.

If you have any questions about the topics we covered, you can reach out to us on LinkedIn or check out our website at You can also join Beyond Energy, our Slack community of data leaders from the sector. There’s a link to sign up in the episode description. We’re just getting this show off the ground, so if you like today’s episode, please leave us a rating, review, or subscribe wherever you get your podcasts.

It all really helps. See you next time.