Oil and Gas Startups Podcast

0:00 What's going on, guys? Welcome back to another episode of the oil and gas starts podcast.

0:06 We get icon signs in the house, guys, not a startup, but you have a new product. That's pretty innovative. I've actually seen it. I've actually broken my golden rule of actually seeing things

0:16 before actually having on the podcast. So I do know a little bit more than I would normally like, but it's been awesome. I didn't know you and get the rest of the team over the last couple of

0:25 months. Yeah. I think it's fair to say. I mean, it's a new product It's not a new company. We've been around for like 21 years, but the mentality with this one is suddenly more of the start of

0:34 mentality. It's a new direction for icon science. It's one which I've been working on. I've been in this space for about 10 years now. But what we're doing in terms of the accessibility, the

0:47 generative AI, the machine learning, everything that we're trying to do, basically, for those of you who joined into this, we are looking to put the consumer friendly face on data management,

0:58 try and take data management out of the archives. and get into the operations, that's a big thing. I want to talk about the cool, sexy, latest stuff that you guys are doing, but first, what is

1:08 ICONS, give you a little bit more of the origin story, you guys have been around for 20 something years, what are you guys traditionally known for, let's dive deep into that. Yeah, good way to

1:16 start. So we are the geoscience guys, that's been our big thing. Rock physics is what we really do. Don't ask me too many questions about that. I come from the data management side, I come from

1:28 petaphysics side, when it comes into rock physics. I'm like, okay, yeah, we do that. But essentially we work with operators around the world, different areas, onshore, offshore, and

1:39 basically do subsurface characterization, try and get insights of the subsurface that can make better decisions. We have software, we have services. We started out of Durham, I think it was in

1:49 the UK, in like 2000-ish. Our headquarters are in Serbton, which is

1:58 in West London We've got offices in Houston, Kuala Lumpur. Calgary, and then we've got various support guys all around, and we've worked with all the Supermajas, a load of the NOCs, but also a

2:06 lot of the smaller guys as well, and consultants as well. And in that rock physics space, we're pretty well thought of working with the market leaders when it comes to that side. So what kind of

2:15 data we type in one of the geoscience side? So it's a little bit of everything, but it's a lot of well-bought, a lot of logs, a lot of seismic. But the thing that really sings to my heart is more

2:25 of the, well, I turn the awkward data, the stuff that doesn't sit so nicely, so a lot of the rock and fluid data. So whether that is core samples, whether that is mudlog, whether that's

2:35 cuttings, whether

2:38 it is a lot of the PBT type analysis. And that's really where the curate story comes in. It's trying to find a home for all the stereotypical data that people think of, like log seismic, but then

2:49 really getting into the rock and fluid and the awkward stuff that a lot of the times when people go, how do you manage that? they kind of shook their shoulders and go, We don't. Yeah, it's in an

2:58 email, or it's in a folder, or it's in a Excel spreadsheet, or our vendor looks after it for us and a whole lot of that kind of stuff, so. So was it bread and butter historically? Was it more on

3:06 the software side and more on the server side? It started off, I believe, and I'm

3:11 probably speaking way beyond my history here, but I think it started off as an Excel spreadsheet. That was the first thing that they did, and then they turned it into a product. But I'll serve

3:21 this side of there, 'cause a lot of companies, particularly in this space, over here, they don't have their own self, like huge, great geoscience communities. They certainly don't have a team

3:29 of rock physicists. We do whale bar analysis as well, geomechanical studies as well. Super important, but a lot of companies don't have that expertise to pull 'em. So that's where our consultancy

3:39 comes in. It's like, you need someone to turn around a project to be able to feed into that. Well, that's where our concerns come in. We can really help you understand the subsurface, understand

3:48 how you can place your wells better, how you can drill those wells better, how you can complete them safely. Safety is a huge part of what we do, pressure studies. So I sound like your services

3:58 side is not just like in a typical software company, it's usually for like to implement your software and get them running, but it seems like it's almost like more consultancy for you guys for them

4:06 to be successful in this space. Particularly on the GS science side, yeah. Our QI guys,

4:12 quantitative interpretation, they're experts in rock physics and pedaphysics and geophysics, they come together to really create those those end to end and we can do as little as much so we can

4:22 either take like the pedaphysical, pedaphysical inputs and do something with them or we can basically take the whole thing and through our network of our own consultants and our own partners, we can

4:32 basically kind of turnkey the whole thing. So yeah, we have got a standalone services department. When you talk about curate, which is about implementing those data management solutions, then we

4:42 then we're very much about the implementation of our own solutions. We're not really looking to do data management as an isolation. We're very much about joining get products in there that will take

4:52 a lot of that heavy lifting away. So I feel like there's not, and I can be totally wrong, I feel like there's not like a significant amount of players in kind of the GG space as there is in

5:04 production or drilling and completions. I feel like it's, you know, maybe a little underserved, there's fewer players. Tell me more about that space. Is it like, are there certain trends that

5:14 you're kind of seeing or are there any like misconceptions around kind of the GG space? Yeah, no, it's a good question So I think it varies if you're in the conventional or the unconventional space.

5:26 The conventional, there's a huge amount of GG. Well, particularly if it's like deep sea, I mean, it goes all the way through. If you take an offshore installation, they generally, they're

5:36 longer turnaround projects. They are more expensive at each well. I mean, each well could have like 18 months worth of planning going into it before you actually go and spud it. HPHT,

5:46 obviously massively complex, massively dangerous if you don't take it all into account So I think GG in there has a much heavier workload. it's part of the characterization that goes into it. If

5:57 you're dealing with areas which are heterogeneous, like there's a lot of conflicting different types of geology that could come in there, a lot of uncertainty, then gene-g is a huge place. When

6:07 you get into the onshore world, I think gene-g's kind of almost been cast to the side. And I've been spending a lot of time thinking about why this is recently. And I've really been looking into

6:17 kind of the Permian as the kind of the case study here. And so if you look at the Permian, I mean, it is just, it's like this giant monster of just productivity. I mean, what? It's the fourth

6:29 biggest production of the world. If it was a country, it would be right up there. It'd be bigger than pretty much anything at OPEC outside Saudi Arabia. But everything's been in tier one. And if

6:39 you look at the MNA that's happening at the moment, then those prices have been pushed up because we're kind of ringing a sponge in those tier one areas. It's like, all right, let's infill and

6:47 infill and infill. And if you look at any of the studies out there, then it's very much there, okay. your spacing should be X. Well, now we're like X divided by three is where we're in, because

6:57 we're just like, well, can we rinse more and more out of that? But you'll see that the production is tailing off those new child wells that have been drilled, which is pretty much all of them in

7:04 tier one at the moment, a lot less than the wells are replacing. And so companies are now in this space where they're like, okay, well, what am I going to do next? Either I can try and flex the

7:13 cash and try and buy my way into those tier one assets. And that's where you've seen a lot of the MA. Or do you start looking around the other areas and go, tier two is tier three really that bad.

7:23 And if you look at it from a rock quality aspect, not really, tier two, tier three, kind of the same thing, but the risk's bigger. And so like, when you're planning these things, it's about

7:32 portfolio, you want to have as minimal risk as possible. And you get into two and three, that standard deviation of how successful your well is going to be is going to amplify out. So how do you

7:42 get a constraint on that? Well, you understand the subsurface. And I think that's where gene science really comes in now is like, okay, well, we're looking at an area which is less will

7:50 understood. How can we improve our understanding? And I think that's where geoscientists is almost gonna have like a second coming when it comes to the onshore unconventional space because we're

8:01 moving from areas where we know what's gonna happen. We're very understanding of those areas into a little bit more of not exploration but kind of on those edge cases of, we probably need a little

8:12 bit more information about that. And I think that's where geoscientists kind of being the wizards of the subsurface. That's where those guys are really gonna come back into it So I think that is

8:20 kind of where I see is going. Are the definitions between, you know, everybody knows what that implies, tier one, tier two, tier three, acreage, is there actually like definitive definitions

8:32 of those or is it very subjective? I believe, I mean, it's hard to pin down, but my understanding is it's not so much on the quality of the rock, it's about on the return of investment per well.

8:44 Like what you expect to receive And it's like, I think it's like 30 above is what makes it tier one. Now whether that's an industry level

8:53 I've probably changed this from asset to asset. I'm probably not the expert who's speaking about this, but from my understanding is that it's more and about, it's the amount of paid that you'll get

9:03 back based on the investment that you put in. It's not actually the measurement of the productivity of the quality. You could have a super productive well, whatever, it costs you like a billion

9:12 dollars then it's not tier one, clearly. And that kind of also stops you just drilling for longer and longer and longer and fracking for fracking and for longer, 'cause obviously the cost goes up

9:21 So it's, as with everything in the world, it's tied back to economics. So you said, you think you're gonna see a resurgence of kind of GD professionals back in the space. What do you think is

9:32 gonna cause that? I think as we get into the stage now where we are drilling more tier two, we've seen a lot of companies kind of investigate in those areas a little bit more. And sometimes it's

9:43 like, oh, we're gonna stay in a Permian? It's like, I think of it like the real estate market. It's like, all right, I live in central Houston. I rent, it's like, all right, Am I going to

9:50 buy a house in central Houston? And I'm like, down there expensive. So I could either try and move out into the suburbs, where I'll get my house with my money, which is kind of like moving to a

10:01 different area. So like moving up to say an asset in Wyoming. Or I could try and do a fixer upper around about where I am, which to me is kind of more like a tier two, tier three. I'm not moving,

10:13 but there's a little bit more work needs to go into this. And so I think that's kind of the analogy there. And I think that the geoscientists are the ones who are going to be able to give you that

10:21 understanding of what needs to be done here Like, how am I going to complete this area more effectively? How am I going to make sure that I have the optimal design plan for this? And how am I going

10:33 to understand the uncertainty? That's the huge thing. That's what investors hate. Investors hate uncertainty. And that's why they love tier one, because it's like, I'm going to drill the well.

10:42 I'm going to get oil out, move on to the next one. Great. Tier two, it's like, okay, well, all of a sudden my ratio of the good wells, the really good wells, the really crappy wells and the

10:50 stuff in between. is completely warped from where it is in tier one, and that obviously makes the investment less assured. So if we can reduce that uncertainty, then we're in a better space. And

11:01 that's basically what geosciences do. They take subsurface data, they reduce uncertainty, and they feed the business. Yeah. I think the, we've really, really struggled to attract talent into

11:13 the industry, kind of across the board, and you're seeing this with enrollments or a drop of enrollments in patrol engineering and geology and things like that. And so I think naturally what

11:22 happens is, it's just simple supply and demand, and we just don't have enough talent in the space. And I think people that are in the space are gonna get paid significantly more than what they're

11:32 getting paid today. And then it's gonna become a competitive again, and then that's probably gonna draw more people back into patrol engineering and geology and the - The cycle. The cycle, kind of

11:42 just like naturally. On GD data,

11:47 you know, I'm very familiar with

11:50 production data and that's how I got my start of the industry. And so I've been very privy to a lot of information about how even the super majors work and the independence and even today, just kind

12:01 of the status of what it looks like. What is the current like status of most G and G data in this space today? Yeah. So, yeah, that's a big question. So it like if they're not using you guys,

12:16 what is it like, what does it look like? Is it just Excel spreadsheets? Is it just like, yeah, Google Drive? I mean, so when let's take log data, it's the most like abundant, I would guess,

12:28 in terms of sheer number of files that you're going to get. Comes in this format that's called Lass, which is a log ASCII standard. It was developed by authentic Canadians like 35 years ago or

12:39 something like that. There's last one, last two, last three. No one uses last three for whatever reason. So everybody's been using last two for the last 35 years. And basically it's an ASCII

12:49 file on steroids. And what do you do with it? Well, you keep it on a well folder somewhere. You keep it in your application. You keep it with your vendor. A lot of it's public data these days.

13:01 I mean, let's face it, these wells and you, a lot of the areas in tier two, we haven't drilled new wells. They've been kind of left in the back burner, but we have got decades and decades of

13:10 vertical wells that were drilled in the heyday of the Permian. So conventional early day Permian And so it's that information. Where's that sit? Well, it sits with the Texas Railroad Committee.

13:22 It sits with vendors like IHS. It sits on hard drives and it just kind of sits everywhere but nowhere at the same time. So when people say, where's my data? It's like, okay, well, I mean, how

13:35 big an existential crisis do you want me to give you right now? 'Cause a lot of the day you need, we'd actually have. And so yeah, there isn't a lot of management processes. And then you start

13:45 looking at all the different data types that geoscientists use. And it's not like we just have one type. We have logs, okay. So that's on like a half foot resolution. We have core data. Well,

13:54 that could be done on a nanometer resolution. And then we've got images which come with that. And then we've got like

14:01 test measurements and fluid flow pressures going through that. And then it goes all the way up to the other scale from nanometers all the way up to seismic, which is like kilometer regional scale.

14:11 And then we need a system that's somehow going to manage all of this, bring it all together and make it accessible. And so that's when you kind of get these solutions where it's like, oh, I'm

14:18 using a finance system to manage my subsurface data. And you can't just go, but why? It's like this is such a specialist subset of data that if you start using a generic platform for this, yeah,

14:30 OK, it's going to ask, answer some of your questions. But when your geoscientists really want to know what information is that, what can I use? What is useful? What's my most relevant data?

14:39 They just don't have those answers And so that's where a lot of time is spent, where people go. You look at the processes to get data from where it is now to utilizing it. It's a huge chain. It's

14:50 like, I've got to find out where that data is. I've got to look into that data. So I've got to pass it somehow. I've got to figure out whether I already have this data. Do I have exact match of

14:60 that data? Or do I have a different version of this data? It's like, okay, am I looking at the most recent version? I was there another version, a better version of the truth. I should be

15:07 utilizing it out there. How does it relate to other information? Does it make sense? Has it been processed correctly? Has it been tagged correctly? Is it stored somewhere, can access it? And

15:17 then at that final point, someone can actually do something, start utilizing it. And you hear horror stories of companies saying, oh, it took us like three months to get all our data ready for

15:25 this particular project. And then I'm like, think of how quickly we drill wells in an onshore environment. It's like factory drilling. It's like drill, drill, drill, drill, and then it's like,

15:34 and then look at the stereotypical geoscience workflow, getting your data together. And it's like, these do not add up. It's like, if you're trying to make insights, you're trying to be like a.

15:44 power source for your own company or I'm going to create competitive advantage by understanding the subsurface, we're going to be getting those insights out there quicker. And so you need to have a

15:52 centralized place where if you need to get access to the data, everyone knows where it is. We've done all the vetting side of it. So someone can get that information and put it into whatever they

16:01 need to. Is it an application? Is it a machine learning script? Is it some kind of new tech that doesn't even exist yet? Well, whatever it is, you've got to have a foundation of that. So that

16:11 information is freely accessible. So it becomes a stream of data, essentially. And that's pretty much, like, so you also need a part of curate, right? It's tackling that issue, yeah? Yep,

16:20 100. Our whole role is to try and centralize subsurface data. So it's freely accessible, but answer a lot of those previous processes. So which is my best quality of data? It's a question that

16:31 geosign to spend all the time asking themselves. It's like, all right, I've got these logs. Is that the best version we have? Is it another version? Where the heck did this data even come from?

16:40 Because if it's been through so many different processes. Is this the raw day I'm looking at? Or is this someone's kind of pseudo mock-up of the data? And how much of it is, can I take and how

16:50 much do I trust? Confidence is a huge part of it. So our tool basically is designed to try and collate all that information, clean it up, standardize it, make it accessible, provide in situ

17:00 tools for people to be able to quickly screen that data. So log visualization, seismic visualization, maps, without having to export it into another application and then be able to confidently say,

17:11 that's the data I need or the other way is, what workloads I want to achieve or what wells can I use that for, being able to answer those questions quickly. And then we've got all the integration

17:21 tech to be able to feed it directly into applications, be able to feed it into machine learning scripts, whatever it is and basically get that data flow going so that it's an always on ready source

17:31 of information to be able to action.

17:34 How's the response? Good. Yeah. I mean, there's two ways you can look at it good in terms of what could be achieved. And then that horrible moment where people look at where the data is right now

17:47 and go, Oh man, this sucks. This is going to be difficult. And I think that's where data management, traditional data management has got a bad name. It is because it is no secret in our industry

17:59 that the way we handle subsurface, geoscience data sucks. We're not good at it. We never have been good at it. Trying to clean that up, traditionally, has been a big massive service operation.

18:10 It's like you look at some of the vendors out there and it's like, well, whatever you pay for software, you're going to pay three times as much in terms of service deliverable, just to clean it up

18:17 and get it implemented. Counting to your initial point, that's what a lot of the consoles do. The whole approach at ICON Science is to try and automate that as much as possible. We want to

18:26 productize that so that you guys get deployed a system that could handle your data, and our whole tagline is subsurface data managed. Because that's what we want to do. We want to

18:36 to have your subsurface day to manage and quite frankly we're not set up to do. those big, long, drawn out data loading, but that's not our business. That's not what we're interested in, and we

18:45 don't think it has to be that way. You look at the steps that generative AI, machine learning, it's taken over the last few years, and you're saying that that can't be applied to improve how we

18:54 handle data in our space. Absolutely, it can be. And so that's where we're really at. We're trying to replace a lot of those tired processes which are very resource heavy, and replace them with

19:04 technology. Do you feel like, do you feel like with this, you know, there's a lot of use cases that I've had this conversation a thousand times recently where the biggest thing that you're selling

19:16 against when walking into some of these companies is an engineer or somebody internally who's building their own things, right? It seems to me like this is a little bit more of a complex problem,

19:28 and it doesn't seem like there's a lot of other solutions out there. Do you find that where there's some sort of GD professional who's kind of like cobbled something together internally, Is it less

19:39 that and more so just like this is a insurmountable data management problem that we're just never going to get the other side of and so we'll just keep doing it the way that we do it. Both. The

19:48 second one probably more though, we always say our biggest competitors would be perfectly honest is doing nothing because it's like, this is just way too much for us to even deal with. The number

19:59 of companies who have the specialists inside and that's what our data solutions team does is they're not there to really do a lot of the data loading, but they help to guide it and try and make it as

20:08 painless a process to scope what would make a good data management solution. In terms of building their own solutions, often what you'll find is that there will be legacy systems which will have

20:19 been existing around. So someone at some point will have built an access database or someone will have put together. People love to call like a series of Excel files in a well folder, a database,

20:31 and I'm like, there is nothing database about this This is just a slightly more organised mess, but. It puts you a little bit further ahead, but it's not a database.

20:42 The one area we would say where we do get people maybe designing their own things is data scientists, because data scientists need a whole lot of data. But what they'll do is they'll take certain

20:52 aspects, certain areas of data, clean it up for their particular needs, and then be like, all right, that's done me. And then they'll go off to their next thing and do whatever it is that the

21:02 business wants them to do. Meanwhile, your geologists, your subsurface engineers who are looking for this data. I'm talking a lot about geoscientists, but a lot of this data is valuable to other

21:10 related subsurface disciplines. It's not just about the geologists, but it's like, okay, well, that didn't solve any problems for me at all. You've got the data you need, but it's an

21:20 operational idea that hasn't helped at all. And one of our biggest success cases, actually with the previous version of this software was with a company who invested heavily in data science

21:34 Their Decigns team was. super keen to make their own core data management solution because they're like, oh, we can do this. We're super smart. We can do this. And someone at the right level

21:43 went, yeah, but if we get in a solution that can provide you guys value and also the rest of the business, it's now a little bit more cost effective and a little bit more intelligent use of

21:53 resources. And so they're able to top them down from building their own thing and went, okay, well, we'll do this and we'll just hook up to using an API. And that's kind of the

22:04 internal politics that we sometimes come across. That's interesting. So on on curate, you mentioned generative AI ML. What do you guys do to that space? So yeah, I mean, the first thing we've

22:19 done is what everybody immediately does when they see gen AI and they go, Hey, chat bot. And I'm like, sweet chat bot

22:26 We like to think ours is a little bit more intelligent than some. We have put some thought into it. But basically what it does is if you look at the subsurface. And this is cross-all, and this

22:34 used to be for 8 of all data that you come across as unstructured. So I only 20 of its structured, and what is every single one of our workflows in the subsurface that's in an application dealt with?

22:45 Well, the structured data. So 80 of it just gets lost. And so what we've done is we've it into AI implemented basically,

22:56 so that it's a generative chat, but you can ask it any questions, and it'll go away, and it'll look at your own document So it's kind of a RAG system, it's not completely generic, it is very much

23:05 tuned on individual customer's data, to be able to go away, answer those questions, and then it'll take a little thumbnail of the actual source document as well, and link you directly to that.

23:17 'Cause the one problem we've got a generative AI is we enter stage where no one trusts anything, it's actually generating. So you've got to have that link back to that reference data, you've got,

23:25 these are the answers I've got, this is the reference, have at it. So that's stage one and that's kind of low hanging fruit and we've had this conversation. It's kind of like it's the first thing

23:34 you have to do. It's about accessibility. It's a great tool, but is that the limits of where you can take it? Probably not. So the next thing for us as well, can we do it to get more on the

23:44 inside side? Can we use it as a tool to help subsurface experts, management, whoever it is, start seeing insights and data on great scale? Because if we're looking at the Permin, I mean, we've

23:55 got customers who have got 53 million wells in their curate platform. And all of them have got logs, but a lot of them do. We can't go through that manually. Look at every single log and look for

24:06 something that's interesting. So can we use generative AI basically to be able to start suggesting areas of interest light? Hey, this is an area that you may not seen before, or this data looks

24:16 weird, or here's something that pops out, which is, you know, kind of interesting. And that's where our next stage is going. It's more that insight generation. So you start looking at it and it

24:25 goes, all right, cool Now, this is something you might be interested in. and it helps you hotspot into those areas. The other thing for us, and this kind of goes on the data side, is that China

24:36 turned unstructured data into structured as much as possible. There's so much data out there, which you look at a report,

24:45 they're just floating around there because no one's ever gone and tagged that data. Like, where did it come from in the first place? Like, well-identified operator. What the heck even is this

24:53 information that's in here in the first place? It's just code report. I don't know what the heck it is Without someone going in and opening it and accepting all that information, it's gonna get

25:02 lost to the ether again. So that's where Genitive AI come in, do some of the categorization and extract key identifiers. So that when you look at it and curate, you go, all right, these are my

25:11 logs, this is my well data. By the way, do you know you've got all these reports and this is the information that's in them and now you wanna go and query it? Okay, well, now you can do that

25:18 with the chat bot, but it gives you that high level knowledge first. So a lot of stuff happening on the

25:26 Genitive AI front. Everyone, every company that I'm talking to is us included. Yep. There's so much cool stuff that's happening in the space. I think there's a lot of challenges from a data

25:40 management perspective that I think that it's very easy to want to jump to the most cutting edge sexy thing that's out there and be like, yeah, let's prompt everything. But it's like, we need to

25:50 go back to first order principles. It's like, well, if your data's absolutely dog shit Yeah, your chatbot's only going to give you dog shit. Yeah, 100. That's the thing, which is the big

26:02 challenge with us is there's a lot of data out there which is locked in and go, what the hell is this? I mean, I did it, I was a pet of physicists. Before I did all this, I was a pet of

26:11 physicists. And one of my first ever jobs was to go into the UK national data repository and try to find data for a field-wide reinterpretation. And I was like, I had to go through everything

26:23 single fold when I was looking at these things and I'm like. the hell is half of this stuff? And it's like, I opened one and it was like, it was a picture of a sunset. And I'm like, that is

26:31 nothing to do with an end-of-well report. Why the hell is this in here in the first place? And it was just like, so you end up with all this crap that's basically in there. And I'm like, well,

26:39 if that's what your chatbot's drained on, well, that's what it's been tweaked on. Yeah, good luck getting some kind of sense out there. 'Cause after a while, you're just gonna end up with

26:47 pictures of pretty sunsets and all sorts of other stuff. And as soon as that happens, you're just gonna go, well, this is crap. I'm gonna not bother using this I'm gonna go back and do my old

26:55 system. And that's how these systems get broken all the time. That is one of the biggest risks that you kind of run when deploying any sort of AI feature set into a product is that if you're not

27:08 getting accurate answers right off the bat, it makes you lose confidence on its ability to be able to actually perform. I've experienced that even recently I mean, it seems like even with GPT-4,

27:20 Um

27:25 It ebbs and flows and just sometimes where I'm like absolutely blown away and I'm like, this is remarkable. And other times I'm like, this is so dumb answers that you're giving me.

27:38 I was doing a little bit of research on somebody and I asked for some factual information and it was like it spit off stuff that was blatant lies. And

27:50 it's very easy And I think this is going to lead to like, there's like broader implications of this, like societal implications of like the degradation of like critical thinking where people start

28:02 to just assume

28:05 that any form

28:09 of AI is accurate and the answers are going to be superior and it's going to lead to, and you see it, I see it on social media now, every day, where it's like, people are just blasting out like

28:18 these extremely long, um, and like well crafted, but very ineffective, like social posts and something like that. And it's like, just use your fucking brain people, instead of just taking what

28:32 GPT is giving you at face value and then just like running with it. Well, I mean, yeah, I've got to the level where I've got a very good AI BS filter now. It's like, I can look at something and

28:45 go, yeah, that was AI. I mean, if you look at my LinkedIn profile at the moment, I mean, it is, yeah, okay, so I run that through it. And this was when I was still quite naive and I was like,

28:53 oh, yeah, emphasize my ability to about all the things I've done about structure. I basically fed it my CV and I was just like, okay, make it and a nice little LinkedIn profile. And after a

29:03 while, I read it and it was like, every second world was like strategy or strategize. And I'm like, that's not what I meant. I didn't literally just mean putting BS words that mean strategize. I

29:12 was like, so yeah, it's very easy for that thing to just, it goes mental. It fixes the cold start problem of like, you know, hey, I need to draft, something, give me a framework, give me

29:24 some ideas to think about it, use it as a baseline to be able to revise and ultimately come up with something, but it's

29:36 so weird to think about where this is going to go and even three to five years of people who embrace it, learn how to use it, but also learn how to use that as a tool to be successful and not just

29:48 take it for what it is. Those are the ones you're going to thrive Yeah, 100. I mean, we see this with every technology, to be honest, because people that buy into this, like, it's a silver

29:57 bullet, it's going to change the world, and it's like, it doesn't exist. There is no silver bullet that's going to change the world. These are tools that still have to be utilized by someone with

30:05 a brain who understands the problems they're trying to solve and then apply them. It's just another way of doing that. It's a nice way of summarizing information. It's a nice way of providing some

30:14 insights, but you have to engage your thinking muscle at some point and get to the stage where it's like, all right, yeah, this makes sense, So this doesn't make sense. Yeah, what are you,

30:23 what are you most excited about with what you guys are working on? I mean, I would be lying if the genitive AI stuff wasn't just because it is cool. I mean, for everything that we've just said,

30:34 if you tweak it in the right way, the way that we've got it kind of in chain of reasoning right now, that's the super core stuff. I can't talk too much about it in the moment because it's still

30:46 kind of early days, but this idea being able to break down quite complex workflows and basically be able to trigger off parts of it so it can kind of like you ask a question and it either fills it in

30:57 with information or it knows how to pull an API or do SQL queries kind of trying to because one of the issues with data management software has been traditionally there is that you kind of already

31:08 have to know what's there if you're going to find it in the first place it's like okay what's the data model that's been utilized here and traditionally you'd get like geotexts who would be like

31:15 trained in geo speak but also trained in data management speeds and someone say find me all all the data for Area X. And they'd be like, all right, I know how to do this. Type, type, type, away

31:23 we go. Well, those people don't exist anymore. So now we need a solution where someone can ask that same question. And it's able to get the unstructured and the structured information together.

31:32 And so there's a lot, it's pretty complex once you get into that because generative AI is terrible when it comes to structured data frankly, 'cause it's just a series of numbers and values. It

31:43 can't do anything with it. So you need it to run through more conventional technology. And so it's kind of utilizing generative AI to pull together a particular chance of your own technology. So

31:55 like, well, I need to query the SQL database here and I need to get information from this API here and then combine it together and put it into something which is sensible for you. But it's kind of

32:06 like a balancing act. And then the other things, the automated workflows, just because automated workflows as they have existed within our industry a been little always bit have,

32:18 They've existed like period. I mean, whether it be through spreadsheet macros, whether it be through something like a Zapier, when like, you know, or an if-this-then-that, when something

32:27 happens, it triggers things. But I think especially with generative AI, like that can take them, potentially, to an entirely new level. And you're seeing products that do that, so now you can

32:36 build out workflows with 12 steps. Yep. You know, the now can be executed repeatedly, all day, every day And take into account the outside information, 'cause the big thing is it's like, All

32:49 right, find me all the data that's been tagged like this. What if that day wasn't tagged like that? But it should fit that. And it's like, well, your heart logic, which is if and all, isn't

33:00 gonna fit all of that. So how can we get that kind of information? And I think that's really where gender decay comes in. It kind of expands out the bubble. And if you, again, you have to be

33:10 careful with it, but if you tweak it into a certain way, and you start getting it to self like, it starts almost understanding its life. Okay, well, this information could be considered to be

33:20 something like that. I will return this, but I'll also flag it up and say, this wasn't the exact match for your filter, but at least it's something that's kind of useful. That'll help you get

33:28 there. So yeah, that's a

33:31 big part of where we're trying to get in is the after a while of traditional logic. You can't clean up every piece of data out there. Eventually, you're gonna need something that's flexible enough

33:40 so that it can be like, okay, close enough, make sense And that's a balance. Do you have any unpopular opinions about the GG space?

33:52 Oh, many. I mean, unpopular to who I guess is the question, but I think a lot of our workflows are designed for an area that doesn't exist anymore, to be frank.

34:05 The fact that we need so much data, we need it to be in a certain format, the fact that we need it to be presentable to this particular technology, this particular time is just so again. against

34:16 how the rest of the world works. And it's like every single one of our platforms, it's like, well, I have a proprietary way that I start this technology. And we hear things like OSDU, which I'm

34:26 not gonna get into 'cause that's a whole other thing, but it's like, is that breaking down the silos in the way that we think it's gonna do? Is it breaking down the silos in the way that the

34:34 business needs? We gotta be quicker. It's gotta be about streams of data. I mean, if I go online and I wanna find literally any piece of information about anything, I'm gonna Google. And it'll

34:47 give me the answers pretty damn quickly.

34:50 If I wanna do that in geoscience and a very specific question, it takes forever. And it's like, we're just not at the same cadence of where drilling is. Again, it's fine if you've got 18 months

34:59 to plan a single well. But if you've got like 18 days to plan a whole drilling operation, yeah, that's not gonna work. Yeah. So that's my big book bear with it. We're just too damn slow.

35:14 1, 000 and I think that the GPT is, and gender and ABI as a whole was ushering in. If nothing else, I think it's raising the bar for what our expectations should be out of more like enterprise

35:27 systems. So it's taking what we're seeing in our consumer lives and now applying it to our business lives. And now we're saying, well, when I'm at home, I can ask at GPT, anything can get most

35:38 likely at a half SD's in answer. Or let's give me start in the right direction Why can't I do that with my work data? I mean, I'm getting to the level now where I am super critical of any time I

35:49 ask for data and it doesn't give me the thing that I want, even in a consumer world, like Alexa. Alexa is the most annoying thing in the world to me at the moment because it's just so dumb. It's

35:59 like, I'm like, why does this not have generative AI in it? It's like, okay, so the fact I'm British, all right. First thing, you don't understand half the things I say. That's fine, I'll

36:09 take that. When you do. There's not a British Alexa you could talk to. She speaks British sometimes, but I don't think she understands British particularly. I mean, my entire in-laws are a

36:21 Scottish. Yeah. They have such a rough time with Alexa sometimes, just trying to get her to do the most rudimentary things. The one who's actually best with it is my five-year-old cousin, like

36:31 him and Alexa are like best buddies, but it just infuriates me when you ask her something. She's like, I don't have an answer for that right now. And I'm like, Come on, it's 2024Yes, give me

36:41 something that's at least close to it. And I don't know, but in our space, it's a whole nother level. I look at it in GEG and it's like, why do I have to click so many dumb buttons to get

36:50 anything? Like just one line, this is what I mean, get it to me. It's cool, it's funny you mention that. It's cool to see how my kids interact with technology. You know, and I'm sure it's the

37:02 same way that my parents viewed me interacting with technology when I was a kid, but like they, we have Alexa's all throughout our house and the will be. ask the questions, we'll have it play

37:11 music. Like, and like they don't even think twice about it. Like this is just like life for them. And even my three year old like knows how to navigate the phone so well. Yep. The other day I

37:23 was at Pickleball and my phone was about to die, but I had my laptop with me and I was fully charged. And I was like, they wanted to watch something and so I put a YouTube kids on my laptop and I'm

37:32 like, hey, I go, I'm gonna go play game. And they're like, I look back as I'm walking away and they're trying to touch the screen on my MacBook and they're like, what is this thing doing?

37:43 They're trying to swipe on it and I'm like, oh, wow, y'all have never used a mouse before. So I had to stop and

37:52 teach them how to use a mouse. And even when my youngest at home, whenever he wants to show, it's almost like he's gesturing, like he's wearing the Apple Vision Pro or something 'cause I have them

38:03 remote in my hand and he's like this in the air and he's like, nope, swipe, swipe, That's the one that I

38:09 like. That's the definition of digital nature right there. But I mean, well, take it this way, going back to the original conversation. And it's like, all right. So you're a new graduate who

38:19 came out and you've grown up like, what? You're 21, you've grown up. Facebook's always been there in your life. Google's always been there. Genitive AI's been there for like, good few years of

38:29 your life. And now you come to the GG space and you see this application that's literally rows and rows and rows and rows of damn buttons and then right click in context menus and it's like, oh,

38:40 you want to use the application that's caught in your job? Well, here's a training manual. So don't have fun. In three weeks time, you might be a novice user and it's like, what the? And that's

38:51 kind of a problem across our whole space. I mean, that's why when we design curate, when people look at it, they're like, where are the buttons? I'm like, there aren't buttons. That's the

38:59 whole point, it's like, it's clean. It's an interface we got the guys that designed Facebook to help us with it, which is scientists, we couldn't design a UX for shit. So we paid someone to do

39:10 it for us. And it's like our system compared to those other ones, night and day. But still, where at the level where you still need to enter more buns would love to get it to the stage. Maybe we

39:20 should get your kids to be like testers on it and be like, all right, kind of three year old who literally is born into vision goggles, can they utilize it? If so, great. Maybe that's where we

39:27 need to be gone

39:34 Curious, it's like, it's super slick. And even with me not knowing that space very well, when you guys came in and you explained the problem and you're showing what you guys were working on, it's

39:43 super slick. So I think the name of the game was just showing people, showing people what's even possible. Yeah, 100. I mean, the more day we can get in there, and that's kind of why they all

39:53 made data so important 'cause it's like, all right, once you get the data in there, then you can see the power of it. But with us, it's always like, hey, here's Teapot Dome and everyone's like,

40:02 oh, Teapot Dome again. It's like, I've seen this in like 17, 000 different applications. So that's kind of how I get the data in and they can see it and they can feel it and go, well, that's my

40:12 data. That was quick. I mean, we can load data pretty damn fast. We managed to do a 53 million wells and somewhere like half a million data sets in like two months. And that was involved the

40:23 whole conversation of how do you want the data? Look, how do you want it to go? The actual loading process was super slick. That's where we got to be, just being able to suck in all that data.

40:31 And then once it's in the other end, people being able to go, all right, now I can see it in a super cool interface. I can click and I can access it. I can search it. I can get it to where I

40:39 need it to go. It's where a lot of our new tech is. It's like, it's the integration component. Cause the other thing that's in our space is a cloud and it's like subsurface is like the last

40:48 bastion of on premises because it's like, all right, none of our applications run in the cloud. Our seismic volumes are like terabytes in size. So what's the cost to me? Get it actually in the

40:58 cloud in the first place.

41:01 the rest of the organization is like, Yeah, cloud everything, cloud this, cloud that. And then subsurface is like, No, we're going to have it on-premises. Thank you very much because

41:08 everything I do is on-premises and I need a big beefy like processor that sat underneath my desk and sounds like an aircraft carrier. I'm like, It's just a completely different space. It's just a

41:22 surface. And so that's kind of the next stage is like, Okay, well, if you do want to go on the cloud, how do you get data in there? And then you've got all these applications that are just not

41:31 configured to work with APIs. I mean, APIs aren't new technology, but there's so many, so many geotechnical applications that are built on like 30-year-old infrastructure. It's like, Well, how

41:40 can I get data into that? And that's where we create this thing called VFS, which basically replicates the entire data stored virtually so that an old application can navigate through and find the

41:49 data they need to. But that's kind of where we are in our space. It's like there's a lot of legacy apps Yeah, I like to think one day. I mean, this is probably in the sky, but one day, I think

42:00 data management won't be an issue. I think one day, things will get to the point where ingestion is super easy, cleansing is super easy, enriching and contextualizing data is even super easy.

42:11 What's the time horizon on that? I don't know. But I think with the way things are going with journey of AI, I think there's given me a lot of smart people working on it. And I think that, you

42:19 know, what is single-handedly still the biggest issue of just data quality in our industry I think we'll just, we won't even think about it. I mean, that's, that's our aim. Our aim is ultimately

42:32 that I don't want to be doing data management. I would love to be doing the course if that happens after the data management aspect of it. And so quick, we can get there, happy or I'll be.

42:41 Absolutely, man. This has been great. No, appreciate it. Great conversation. So no, it's been great to get, like I said, getting to know you guys and seeing the technology. And I think you

42:49 guys are going to continue to crush it in this space. So excited to watch. Yeah. All right. Thanks very much Absolutely man All right guys, take two seconds, leave a rating review, share with

42:58 all your colleagues, and we'll catch you guys.