AI After Dark

00;00;00;02 - 00;00;27;29
Unknown
All right, speaking of names, how'd you come up with Smart Chain? Where they come from? I call, well, actually, full credit goes to John Hanson, who's the the, the original smart chain guy. He invented it. Oh, shit. He invented it in, like, 2006. And during the height of the the blockchain. He had this vision that he could use blockchain to automate payments between, vendors and their operator customers.

00;00;28;06 - 00;00;48;15
Unknown
Right. And so he decided that blockchain was the solution. We we saw a lot of value in being able to automate that workflow. But ultimately the blockchain element of it maybe were a little too soon to the game. Right. It's really good for zero trust kind of stuff, like if you're trying to trade some kind of services. I don't give a good on what that is.

00;00;48;18 - 00;01;14;02
Unknown
Well, somewhere across the world that you don't know. Yeah. Cool. Blockchain is good, but when you're doing AI, when you're doing a transaction with a trusted partner, blockchain is superfluous, they say. So essentially, smart chain has turned from smart blockchain into smart supply chain. Right. The transition's quite nicely where we've seen a lot of value with, with the procurement guys looking to bring essentially real time auditing into operations.

00;01;14;04 - 00;01;32;05
Unknown
And so it transitioned over really nicely. We still have the hooks for blockchain. We, we, we tie pretty aggressively into the court of blockchain and Ethereum. But honestly, it just hasn't been a major value driver for our customers. Was going to ask, like, do they even care? It's there's always a technophile in the back of the room that's like, yeah, let's do this.

00;01;32;05 - 00;01;53;17
Unknown
Right. That's it's trade. And and you know, Ethereum or block or Bitcoin. And we're like, yeah, that's good if you really want to. And then we look at everyone else in the room, they're like oh right. So yeah, I imagine is one of the things where it's like, especially if you're not as familiar with it, that you, you, you base your assessment on the hype.

00;01;53;17 - 00;02;14;05
Unknown
And right now. Yeah. Blockchain doesn't have the hype right now. So it's AI. So yeah I mean it paid well for us in the beginning when we saw it as a vision. The technology didn't come through. The value proposition wasn't there. The value proposition for us is just a lot more focused on the nuts and bolts automation of transactions between vendors and operators and oil and gas in the oil and gas space.

00;02;14;08 - 00;02;31;06
Unknown
They just want to save time. They want the numbers to be right. They really don't care about transacting in Bitcoin. So we leaned into the the pragmatic side of it. Yeah that's I could go way I could go I've I've, I could go off the reservation for an hour talking about how you want to structure your value propositions in this market.

00;02;31;07 - 00;02;51;15
Unknown
I do yeah. I mean this is, this is the energy market right. This is a people that need an overwhelming amount of proof points before they adopt any kind of digital innovation. You're 100% right. Like this is the problem that a lot of companies who have come to market like it, particularly in the oil and gas space, with a technology forward value proposition.

00;02;51;17 - 00;03;03;20
Unknown
Right? They tend to get a lot of momentum. They tend to get a lot of interest, but I don't think they get a lot of conversions, right. I really don't want to name names. People can go off and do their own research for these things. But there's there's genuinely a lot of these companies. It happens in other industries as well.

00;03;03;21 - 00;03;22;15
Unknown
But the beauty of the oil and gas industry is they they can they can sniff bullshit out real fast. Right? And so you need to go into the room with a really, really straightforward business objective, right. Nuts and bolts. I will save you X yeah, you pay me y. And as long as that mathematics works, then we can do business.

00;03;22;17 - 00;03;41;27
Unknown
And then they say, how are you going to achieve it? And then you can introduce your technology proposition, but you know, hard value needs to sit on top of the technology. Technology can't lead the lead the meeting. Does that make sense? Yeah, absolutely. And so that is something I think difficult for this industry and a lot of legacy industries is the ROI calculator is kind of broken right.

00;03;41;27 - 00;03;57;15
Unknown
With AI meaning like you can't it's really hard. If it's not I'm going to rip out this software and replace it with this for cheaper. Then they start like saying well okay, time save. What does that mean? What is the value of that? Like people say it was I mean what is the value of that. It's it's been tough.

00;03;57;19 - 00;04;19;28
Unknown
Yeah. You just you can't do that. I mean, we've tried to make. Don't get me wrong. We have like a fair amount of AI creeping into into our organization. And we can talk about how that structured in a minute. But but you're right. If if your AI players simply enabling time savings. I mean, maybe, you know, how does that translate into real dollars and cents?

00;04;19;29 - 00;04;37;12
Unknown
Okay, cool. That guy can go home early. I'm still going to pay him a full paycheck, right? Yeah, it's it's a tough one. You need to. We've found that you need to go in with a much more discernable value proposition. Like, you will be able to reduce your fuel spend by 5%, and they can look at their fuel, spend $100 million.

00;04;37;12 - 00;04;54;12
Unknown
Okay, cool. That's going to save me 5 million bucks. Cool. We're good. Right. That that's arithmetic that they can do. But time savings is a lot more intangible right. The other problem you've got with with an AI forward play is that you can come into your you can come to the room and say, hey, I can do anything.

00;04;54;15 - 00;05;21;28
Unknown
The problem with AI can do anything means it can actually do nothing. You know what I mean, right? I think the key to making I work at least now, is to use it to turbocharge those nuts and bolts value propositions that I was talking about. You know what I mean? Right? Rather than saying, hey, we're going to use AI to solve all of your problems, and we're going to make it, and we can expose that AI to you, the operator, they're like, oh, shoot, what the hell is this?

00;05;21;28 - 00;05;40;05
Unknown
Is this something else I need to learn? Do I need to change? What we're trying to do, at least, is use AI to make our business as efficient as cost effective as fast as possible, and then deliver a more traditional value proposition to the customers. Does that make does that make sense? Okay. Yeah. Yeah, yeah, I think that's the secret to it.

00;05;40;05 - 00;06;07;13
Unknown
That way you can in-house a lot of the risk and a lot of that, that bias that comes along with AI but still leverage most of the value that's created by it, and deliver that value to your customers in a traditional way. Yeah. Would you say, would you say that when you're presenting this? Oh, I'll save you 5% on fuel costs, that there is, a belief that that's that's achievable or is it immediate kind of defensive positioning saying, all right, prove it.

00;06;07;13 - 00;06;27;10
Unknown
And that's when you have to go into hard diligence on how yeah, I mean, no one like any I'd say any oil and gas person or any oil and gas operator, especially in the procurement side, worth a medal is not just going to believe me at face value. Right. But I would say the oil and gas industry right now has most of the major savings have already been squeezed out of the industry.

00;06;27;10 - 00;06;51;20
Unknown
And you've got a lot of procurement staff, a lot of operations guys out there looking not not for the yards, but for the inches, you know what I mean? Those small savings. So I would say that most operators we've worked with have been more than willing to consider, multiple small savings opportunities in aggregate, to come up with a sufficient ROI to engage with contract automation.

00;06;51;23 - 00;07;10;11
Unknown
You know what I mean? Like, we can save you here with this cost automation. We can save you here with this tracking, we can cost you here. And then they say, okay, cool. Prove it. And then you go, you deploy, you prove it right. And maybe you don't hit 5%. Maybe you hit seven, maybe you hit three. But there's a there's a justified path for why it's worth continuing.

00;07;10;18 - 00;07;36;13
Unknown
And then they go, okay, cool. The math works out. You can stick around on that. Yeah. What? So you took out you took a blockchain product and are now are you presenting it as an AI product? As suddenly we presented as AI enabled. Right. So we're a technology forward company. The idea being that and this is something we've done with some of the major operators, minus a lot of the major operators, to my understanding, have internal AI teams.

00;07;36;15 - 00;08;05;09
Unknown
Right. And I don't know if all of those operators are getting the true value out of those AI teams. And so we're seeing some interest from these operators to some sometimes directly compare and contrast, like how we use it versus how their internal teams are using it to create value. Right. So people want to understand that we're leveraging the latest technology to to deliver our, our, our services.

00;08;05;09 - 00;08;25;02
Unknown
But in terms of leading with AI, no, no, no, we we lead with nuts and bolts value and they say, wow, how are you doing this? I don't believe you can do this. And then we make the AI explanation as a component of our larger tech stack. Do you see how I'm going? Yeah. Yeah yeah. And where does that fit I mean to date okay.

00;08;25;02 - 00;08;48;01
Unknown
So this is something we can have a chat about this I'm interested to hear. We'll say, as a startup, I mean, we were we the idea of Smart Chain came out about in 2016, and, and John Hansen's backyard in Cambridge, UK. Right. And then it lay dormant all the way through Covid because just, you know, nothing happened then.

00;08;48;03 - 00;09;21;22
Unknown
And then it started really gaining some traction and like I'd say 2023, 2024. And I mean, I didn't really exist at that point, certainly. Like, am I right, like mainstream. And so our original code base was blockchain adjacent, but I would say just a traditional tech stack. Right. And so we found ourselves in this situation where AI has come on board in the last two years, a year and a half, our tech stack was built, but it was far from set in stone.

00;09;21;22 - 00;09;54;03
Unknown
So we're not like a very large organization that has entrenched practices and lots of code like our our code base was still somewhat, mushy. Right? You know what I mean? Pliable, malleable. How have you look at it? So that actually put us in a really good position. And I think there's a lot of companies out there like this, any sort of startup that came around in that 2023, 2022 kind of space where they've still got some flexibility in their operation to be able to embrace this new technologies aggressively as possible, while still having a clear value proposition that they can leverage this year.

00;09;54;04 - 00;10;19;01
Unknown
I'm going so, a new company leading AI first is going to have the problems we just discussed. An older company is going to have an entrenched code base, so they're not able to take full advantage of of what I can deliver. But you've got this sweet spot in the middle, right where that code base is still flexible enough that you can be like, oh, neat, that agent can do passing way better than the traditional OCR systems.

00;10;19;03 - 00;10;33;02
Unknown
Let's have some of that right. And you can bring it into the system really fast. Right? So we found ourselves in a really good position to be able to essentially pick and choose off our flexible code base. You know, yeah, our pliable code base to be able to say, yep, that looks like it's going to be good.

00;10;33;02 - 00;10;47;15
Unknown
That's going to be good. And we're just I mean, we're following the standard agenda. Agenda structure that. Did you say you're building modular Lee. Yes, as much as possible. Right. So I mean, my theory has always been we look at the roles that could be.

00;10;47;17 - 00;11;06;00
Unknown
We look at the roles that can be executed by human beings. Right. We some of them we were doing systematically, we were just traditional code and other we were just doing with the human being. And we're saying, right, okay, so if you can train a human being to execute this task, then you can build an agent or a series of agents to execute this task as well.

00;11;06;06 - 00;11;27;22
Unknown
And then you teach them. So I mean, back in the day, like way back in the day before we started, if you wanted to be able to read an invoice, I mean, first of all, good luck getting the OCR to actually read it, especially if it's handwritten. But they say you get through that. You had to draw boxes on the on the actual, you know, machine and say, look, when you received a scanned copy of this invoice, look inside this box.

00;11;27;22 - 00;11;46;09
Unknown
But if it's inside that box, that's your, you know, the diesel that was delivered and that was the price. And if something moved or the scanning wasn't right, then the whole thing blew up, right. And so, yeah, I had this companies around here that had to do this manually for the longest time because they weren't able to get that OCR tech to work.

00;11;46;12 - 00;12;09;26
Unknown
Whereas now you can rather than, rather than train this workflow like you would a computer to say, look inside this box, whatever you find inside this box is X, if not Y, if not Z, if not if you know what I mean. You can now essentially put together an agent or a series of agents to that can be trained like a human being.

00;12;09;28 - 00;12;27;25
Unknown
Right? So when we receive a new invoice with a new format, all we do is say, hey, prompt, teach the agent a little bit more, just like you would sit down with a person ten years ago and say, hey, this is a new style. The number is actually down here in the bottom right or the units changed or they structure it this way.

00;12;27;27 - 00;12;49;03
Unknown
Now you can now you can talk to the agent, just like you would talk to the human being you would have ten years ago. Does it. Yeah. Do you feel like you this is always within, like a diesel contract or within. Or can you take any document, any PDF and from there be able to say, parse it and say this is this looks most like a diesel contract or, you know, cement contract.

00;12;49;03 - 00;13;07;15
Unknown
Yeah. So right now our focus is definitely, working on fuel, but like the, the infrastructure that we're building is designed to be able to read any oil and gas specific, transaction document. Right. So the system needs to be able to recognize, hey, is this an invoice? Is it a credit memo? Is it a debit memo? Is it a ticket?

00;13;07;15 - 00;13;30;07
Unknown
You know, all those different things. So you you want to be able to train your model within the context of your operation. If you go to agnostic, then it's going to start making assumptions. Right? But just like you would have trained a person ten years ago to say, hey, this is what these documents look like, you're now trained on how to read these types of documents.

00;13;30;09 - 00;13;53;05
Unknown
You don't need to be as restrictive to, you know, specific digital invoices. I mean, this is the number one problem in the oil and gas industry in my mind, which is like, yes, there's unstructured data, but that unstructured data lives in PDFs and Csvs and Excel files. Right. And and it's just terabytes and terabytes and terabytes of data that are in effect trapped because the parsing of it has been difficult to do at scale.

00;13;53;07 - 00;14;09;27
Unknown
And so I, I mean, every single oil and gas company I've talked to has been like, if you can solve me this, I can I can do the rest. Hey, this I, I don't yeah. Okay. So this is just one example of an agent that we're using. I think if we were trying to come to market with just the ability to do that, yeah, I'm sure someone would like it.

00;14;09;27 - 00;14;27;13
Unknown
But there's lots of players out there that can do that. I mean, yeah, Clyde can do that. I'm sure you know what I mean. The ability to to read a document and understand what you're looking at, provided there's a reasonable amount of training goes into the model shoot. I mean, I could probably do that this afternoon with Claude, you know what I mean?

00;14;27;13 - 00;14;49;11
Unknown
It's it's about taking that that agent or that component and then putting it into a larger value focused organization. Right. I know that there's, you know, a lot of the EDR companies and some of these other companies that have been trying to contextualize data, they've been working on this technology for the longest time. Right. How do you read a bar diagram?

00;14;49;14 - 00;15;13;10
Unknown
Right. And how do you read a bit record? I think what's wild is that we've gone from something where that was genuinely a core competency or a unique competency of an organization, whereas now it's just table stakes, you know what I mean? If you wanted to create a company that could rip apart the these documents that you're talking about, I think a person could put that organization together in like two weeks.

00;15;13;12 - 00;15;32;27
Unknown
And and that's commercial scale two weeks. It's insane how fast you can put the stuff together now. So I think we need to be you need to be able to leverage the technology that AI's got, but it needs to be combined into something that's more capable than just, oh, this is this new parlor trick that I can do.

00;15;33;00 - 00;15;52;15
Unknown
Yeah. This is all these conversations I've been having or very have been somewhat more technical, but the truth is, like, it does feel like every time in this wave of AI, someone figures out a hard problem. Yeah, it becomes immediate table stakes. Yes. There's not there's there's no like like, oh, I get a 6 to 12 month, whatever runway to be able to sell this tool.

00;15;52;15 - 00;16;15;09
Unknown
No, no, no. Let's immediately everybody else is catching up. Yes. It's unbelievable. That's probably the scariest thing. Like how fast AI is eating the world is is probably the thing that would keep me up at night the most. Right? I think that's just the reality of the game. So I find myself as a business owner looking to insulate the organization as best I can, knowing that that's what's coming down the road.

00;16;15;09 - 00;16;37;13
Unknown
Like technology has just been democratized so fast. It's it's unbelievable. Right? So again, this is why you can't come to market with a pure I play. You need to become something with a bit more of a a white glove value proposition that leverages AI. Maybe it leverages AI better than your average organization. That's too structured to be able to take full advantage of it.

00;16;37;16 - 00;16;50;28
Unknown
Right? But you can't just say, oh, I'm just going to AI your stuff and you're going to pay me. Yeah. They're like, okay, I can figure that one out. Yeah, yeah, yeah. So what I mean, as a as a technologist, do you feel like you spend a lot of time on the go to market side and from a defense ability standpoint?

00;16;50;28 - 00;17;11;14
Unknown
Yeah. Yeah. I mean you have to consider that for sure. And so to that end for us, I think making sure that you can you can make the customer's life easier. Right. So again you're not using you're not saying hey, here's my AI tool. You can use it right, to say, I've got this AI tool, I'm going to make your life easier.

00;17;11;16 - 00;17;36;03
Unknown
And and you're just going to transact. Did you see where I'm going? And so to that end, I think the most successful, or at least my, my theory is the most successful companies or startups that are going to leverage AI are the ones that are going to be able to embrace the technology faster than the operators can, but then pair that with more of a white glove objective, solution, which is more of a traditional oilfield services kind of style.

00;17;36;05 - 00;17;56;21
Unknown
Right? Where, hey, I'm going to use AI, but I'm also going to show up at your wreck site, right. And I'm going to install that sensor, or I'm going to shake that company man's hand. Right. There's been more than one occasion when I've walked into a, a customer's office, particularly over a medline with dirty boots on. Right. And I'm sitting there talking to them about, hey, this we're going to use AI to do this, and that's going to be passed over here.

00;17;56;24 - 00;18;15;19
Unknown
And it's very clear that I've just come from a wreck site that morning. I think that's been, that's been really effective for us. Right? The customer doesn't the customer doesn't care what the technology is. Fundamentally, they care that you can deliver the value better than they can. And all they need to do is just take the value and then get on with their day to do something else.

00;18;15;19 - 00;18;35;21
Unknown
Yeah, yeah, yeah. It was it's funny because I think the more I've engage with people, the more I think, or actually the more I've engage with people that have some exposure to cloud code or have tried to build something right and have surpassed peak of Mount Stupid on the Dunning-Kruger, you know, effect. They're, they're realizing, like, you know what, I, I these are just tools.

00;18;35;21 - 00;18;51;03
Unknown
And if I don't know what good looks like and if I can't, if I can't show what it means to actually have domain expertise in what I'm trying to accomplish, then none of this really matters, right? It's going it's going to take it's going to be I'm going to get there faster. But there is not where I want to go.

00;18;51;05 - 00;19;11;22
Unknown
Yeah, I need to know where I need to go. And so so you're right. Like being able to come off a rig site, show that you have an understanding of, you know, even if it's marketed almost right, like it wasn't fun. I'm not I'm not saying I'm not saying that that's the case for you. You're saying like like that may be the case for others where it's like position yourself as an expert.

00;19;11;25 - 00;19;31;21
Unknown
I think it's providing a lot more trust because you're you're validating that you know, what you're trying to do. Yeah, 100% people look at this and say, okay, cool. This guy understands my business Andi understand technology. I think the companies that are going to succeed the most by leveraging AI, at least in the next few years within our industry, and let's be honest, the oil and gas industry.

00;19;31;23 - 00;19;53;20
Unknown
Oh, I'm going to I'm going to some I'm going to regret saying this in the future. But because this is sort of a famous last words kind of thing, but I think the oil and gas industry stands to benefit from the new technologies. Right, guys? Right. But fundamentally, our industry is quite insulated in the scheme of things. We tend to speak a very specific language.

00;19;53;23 - 00;20;14;15
Unknown
Certainly our hardware, our tools are very specific to the industry. Right. And this tool has no application anywhere else in the world. You know, I mean, but it's still big news for us is a lot of investment in that. So to that end, I think you're going to see some of the larger AI houses and some of these larger re distributors of AI technology.

00;20;14;15 - 00;20;38;03
Unknown
They're going to struggle to break into our industry. So I think there's a there's a really good opportunity for smaller companies to be able to leverage AI and create value inside the industry. Do you see what I'm saying? Right. I think it's difficult for a company like anthropic to come along and make a genuine, industry centric, operation centric AI value proposition right out of the gate.

00;20;38;05 - 00;21;02;24
Unknown
Right. And so that's an opportunity for people like us to be able to package industry knowledge with technology. Internally. Do you see I'm going with us, right? Yeah, yeah. But the reason why I'm going to regret that is give it two years and all of a sudden anthropic is going to be cool. It's going to be able to speak oil and gas drilling engineer just as well as anybody else on the street over here.

00;21;02;24 - 00;21;19;15
Unknown
It's terrifying. Yeah. Yeah, but I think I think that's where you get into the intuitive element of this, right? Which is like, can you pick up on the fact that your boots are dirty because you actually be doing something on a rig site versus your boots are dirty because you threw them in the back of the pickup truck on your way over, you know, because Claude told you to make them dirty.

00;21;19;16 - 00;21;39;08
Unknown
Yeah, yeah, yeah, I again, like that's one thing that I think our, our industry is really, really good at. It's smelling bullshit. Right. And I don't know, that's one of my favorite things in this. I remember early on in my career, when I got my first hardhat. Right. And so first of all, the, the first, the first hard hat, I got was like a mill hat, right?

00;21;39;08 - 00;21;51;02
Unknown
So it only had the brim on the front. I was like, dumb to show up at a rig site with those. Yeah, because you need the full, you know, the full rim. So I finally got one of those, and then they were like, yeah, throw it in the back of the pickup truck so it gets knocked around.

00;21;51;05 - 00;22;06;29
Unknown
Yeah. And do they smell bullshit? Yeah. They could tell when it was like, oh, you just threw this in the back of your truck versus like, no, you've been out there, you got dope on you. You got, you know, damn straight I got all kinds of stuff. And then and then the stickers. My God, you know, those are like, freaking currency.

00;22;07;01 - 00;22;30;18
Unknown
Experience counts for a lot in this industry. And that's, again, that's one of my one of my favorite things is that everybody everybody has a war story or to some of the war stories on literal war stories, and they're pretty, pretty interesting. I don't think they have anything quite that good, but I find that most of the people in the oil and gas space have spent at least some time on a, on a Derrick, on a frack site, on a well site, doing something right.

00;22;30;18 - 00;22;44;27
Unknown
And they've seen they know what a bad day is like. They know what it's like to get dirty. They know the smell that, you know, the, the, the, the drilling mud or something to that effect. Yeah. And I don't know, I think that that unites us. And I know that's one of the main reasons I'm proud to be working in this industry.

00;22;44;27 - 00;23;03;10
Unknown
I think it's cool. So going back to your anthropic, point, but also go to market side. So something I have, I feel like I've been noticing in this industry is that you get, larger players are willing to trust a palantir or, you know, Microsoft or whatever, some big logo. It's like the old adage where they're like, oh, no, you don't get fired for hiring IBM or McKinsey.

00;23;03;12 - 00;23;19;24
Unknown
Yeah, it feels like there's a little bit of that. But then they're all super frustrated with the work product that comes of it. Yeah for sure. Exactly. I would I, I really I'm trying not to like name names but you have these. So I was like you know ability I I'm doing the same. Yeah. Mutually assured destruction I'm trying to be I'm trying to be careful.

00;23;19;24 - 00;23;38;03
Unknown
But yeah. Exactly. I've seen more than one occasion where, one of the large accounting firms has come in and offered consultative services. Right. And they've charged a million bonus. And at the end of it, they they don't even know what our OPI stands for. You know what I mean? They're telling you, I'm sure you're, well, better. And you're like, no, you know what I mean?

00;23;38;03 - 00;23;57;25
Unknown
Yeah. So I, I see that as a fantastic opportunity for startups in the oil and gas space. Right? You get a couple of guys who genuinely know what they talk about from an operational standpoint. Stop there. And then you can pick and choose the technology that will make your value proposition shine as much as possible. You really don't want to do it the other way around, right?

00;23;57;25 - 00;24;18;12
Unknown
And I guess a logo gets you in the door. But again, I think people smell bullshit very fast. Right? So the question will be is like if if anthropic or a bounty or whatever, you know, can in two years speak like a petroleum engineer, can startups keep pace, right. Can they take advantage of the fact that, like, maybe they're not there yet to be able to lock.

00;24;18;14 - 00;24;41;08
Unknown
Yeah. Customers in? Yeah. That keeps me up at night for sure. Yeah, yeah. And again, the name of the game there I think is just at least for us is to just go niche on niche. Right. If you can provide a very clear value proposition for a very clear or very cheap price, such that you can, you can still play in a space where maybe Palantir just can't compete economically.

00;24;41;10 - 00;25;02;08
Unknown
That's that's probably the safe space right now. But you're right. I think there's a lot of startups out there, myself included, who need to be cognizant of this, right? Eventually, yeah, I will come to eat everything. It's very it's bleak. But yeah, they'll be I don't know. That's the way it's going. There's just going to be there's going to be one winner, at least one winner in every industry.

00;25;02;08 - 00;25;19;07
Unknown
And that's that's kind of scary. But I think, I think people would have argued that in every industry with every technology cycle. Right. Like and don't get me wrong, I think you could look at Google and be like, look like a winner. But there's also there's others, you know, Microsoft, Apple, Apple, they're still others out there.

00;25;19;07 - 00;25;48;14
Unknown
I hope you're right, man. I genuinely hope you're right. Yeah. That. Yeah. The, So so I guess being a niche of a niche. Right. Again, from a technology building standpoint, how much time do you spend adding features, functionalities, building new things versus just refining and maintaining what you've got? Oh good question. I mean, right now, we're genuinely, trying to, I don't I wouldn't call us bleeding Edge when it comes to AI.

00;25;48;16 - 00;26;08;08
Unknown
We're we're still somewhat discerning. We're not doing it for AI site because that just introduced costs. I we've been very careful to not overrun on credit spend or get government credit. Tokens been. Yeah. So to that end, that's been good. But we've been doing that in order to maintain lower cost so we can pass a reasonable value proposition over to the customer.

00;26;08;11 - 00;26;29;10
Unknown
But that being said, we do have dedicated resources focused on making sure that we're choosing the right technology solutions. And right now, those technology solutions change every month. Right. And so trying to stay on top of that is a real challenge. But we're, we're doing our best to, to, to, to, to replace intelligently as we go.

00;26;29;17 - 00;26;56;01
Unknown
But it is turning into sort of a grandfather's ax kind of situation, you know what I mean? No, I don't I haven't heard the grandfather's story. Well, okay. And there's a, there's another one that's, it's named after sort of a Greek boat, but grandfather's axes. The is the, the allegory for, for Cretans like me. The idea being that you got your grandfather's ax, which he's had for 50 years, and he's replaced the handle three times, and he's replaced the head five times, and they replaced the grip four times.

00;26;56;04 - 00;27;15;01
Unknown
Is it still the same ax? Do you know what I mean? And so what we do is we find ourselves replacing this component, replacing this component or replacing this component. It's still a contract automation platform, but it looks completely different than it did four years ago. And the the name of the game is to build your platform in such a way, such a modular way that you can do that intelligently.

00;27;15;06 - 00;27;31;02
Unknown
Yeah, you know what I mean? Yeah. It makes it makes me think of, I don't remember what I was listening to, but when I was feeling smart, I was it was like about how your body regenerates atoms every, like, every seven years. Like, everybody's completely different. But so are you a different person or is it is it part of your soul?

00;27;31;02 - 00;27;58;00
Unknown
Like how do you assess who makes Kim? That's 100% right? Yeah, that's that's my grandfather's ax. But in a lot more biological. Biological. I wouldn't want to think about that. But yeah, it's true. Yeah. And so whereas I where I think the companies that are going to succeed are the ones that are going to be able to find the technology solutions that work for them, don't change technology for technology's sake, but simultaneously look for these opportunities that will make your product better, right?

00;27;58;02 - 00;28;22;00
Unknown
The worst thing you can do is get in there and just start demanding from your team. I need to I everything now, right. And I think some guys on my team might criticize me for saying that a couple of times, but they're very quick to moderate. But yeah, I'm seeing that. Well, there's, there's lots of famous stories, especially from outside our industry where people are being judged on token utilization metrics.

00;28;22;03 - 00;28;38;21
Unknown
Right. And as soon as you make like a metric tag, it it fails to be a metric, right? Because all of a sudden people just stop burning through tokens, like looking up cat pictures just to make just to meet their numbers. Right? Which is which is asinine. You hear about that in other industries, I don't I haven't heard about that so much here and and and oil and gas.

00;28;38;25 - 00;28;56;23
Unknown
No. In fact people I would argue people are more worried about it, you know, and just listening to to operators when they're talking to AI native technologies, they're like, well, what's going to happen when, you know, anthropic goes public? And now there's scrutiny around what they're charging for their compute. And, and you're going to have to just like, are we going to get charged for that?

00;28;56;24 - 00;29;21;02
Unknown
Are we going to cap like what's you know, and it's just another, another question that gets posed with the rapid advance of technology, to the point that we've argued a lot, you know, amongst portfolio companies and Mercury, which is technology's outpacing the human ability to comprehend its capacity. Yeah, right. And so whether that means the actual ability for the technology to perform, but not just that, like what does it look like from a cost structure.

00;29;21;02 - 00;29;47;29
Unknown
What does that look from a scale structure? What is it? You know, that is is it's just adding to the mushiness. Yeah. Technology adoption right now, I was I was it's funny you mention this because I was sitting in a large, a really large operators office last week, and we were talking about this, and one of the main reasons why they're reticent to, to lean into AI natively.

00;29;48;00 - 00;30;06;10
Unknown
Right. Just like bring it on so you can start running queries in the system is exactly that. They don't know what it's going to cost them. Like they don't know what they're getting for their dollars. You know, I mean, the price of tokens changes, like how much a token cost for compute per query. All these numbers are unpredictable.

00;30;06;13 - 00;30;28;29
Unknown
So you have no idea whether you're whether your investment is actually going to bear fruit because you have no idea how much cost you can run a query. The query can fail miserably, right? But you've paid for it. You know what I mean? And you're thinking, what the hell? You know what I mean? So they're they I think these lies were operate as a happy to embrace AI tools where there's a clear value proposition.

00;30;28;29 - 00;30;50;20
Unknown
So put it into your data teams, right. Because they've got some latitude to play. Right. Let your AI teams play with it. But in terms of just giving clawed to a company man, I mean, like their company man's got enough problems. I said this is the last thing he needs is is that right? And what's the value going to be to him for the amount of credits he burns through?

00;30;50;23 - 00;31;08;14
Unknown
Just asking anything. Right. And so I think these operators are looking at it. It's exciting. They understand. But simultaneously they're saying I need to see a clear ROI. I want to know what this thing is going to do, what value it's going to create for me and how much it's going to cost. Yeah. And that question just cannot be answered.

00;31;08;14 - 00;31;30;12
Unknown
And it's getting it's getting murkier rather than clearer right now. So that makes me think such an it has. That quote is like I, I fear under investing more than I fear over investing in AI. How much of that do you feel like applies in the oil and gas space? I don't buy that at all. I think that's something that a large AI company that's shooting for general intelligence, that they're going to say that.

00;31;30;12 - 00;31;49;29
Unknown
Right. But I don't see any of the major, MPs trying to trying to embrace like, yeah, a sentient AI anytime soon. They're just interested in hearing is I say that the thing you hear about capital, this noise or stuff, they just want to get oil out of the ground and to market as affordably as possible.

00;31;50;06 - 00;32;16;14
Unknown
And I don't see arbitrarily investing in AI as a it's a great play right now, especially when you can embrace it as quickly as you can. These days. Right, I get this this might be one of those situations where if you're if you're fit, that was it that the fist through the wall is the bloodiest. Right. So you've got the situation where it's relatively easy to embrace these AI tools in a piecemeal kind of way.

00;32;16;21 - 00;32;42;23
Unknown
So I think you've got operators waiting a little bit longer. You know, they're bringing it in, they're increasing their understandings, they're creating relationships with these AI houses. But I don't see any of them really leaning into it. Whole hog. Certainly. That's where Palantir fits really nicely because Pelant policy is really cool, right? They can bridge that gap. They're they're the OG of understanding the technology and translating it into a value proposition.

00;32;42;26 - 00;33;05;09
Unknown
And they've done that in every other industry, but they're a unicorn. I don't know if there's many other companies that have really enjoyed the same level of success about taking a, a truly bleeding edge technology, then going into an organization with forward deployed engineers and creating discrete value. I'm sure they've had some failure cases, but I've also had some pretty impressive success cases from that organization.

00;33;05;09 - 00;33;22;12
Unknown
Yeah, yeah, yeah. Well, and that's why I mean, maybe not so much volunteer, but I do I do feel like there are oil and gas companies out there that are saying, hey, look, I don't know, I have not I have a clear value proposition, but I do feel like I need to allocate some budget towards AI trials, AI implementation, whatever.

00;33;22;14 - 00;33;51;03
Unknown
Yeah, I buy that. Yeah. Okay. So to that end they're doing it and I, I think they'd like to think they're doing it in a managed kind of way. Are they able to discern, ROI from it yet? From what I can understand, no. They'll be able to come up with some neat things that they did. But has there been any like, enterprise level ROI from large scale AI deployment and side operators?

00;33;51;05 - 00;34;17;09
Unknown
I'd love to hear of an example. I haven't heard of any yet. Right. Certainly my understanding from talking to some of the largest service companies, on their software development side, they've embraced quite a liberal way of doing it. And I, I'm not speaking which, which service companies. My understanding is most of them do it this way, where they let the discrete product teams choose the AI tools that are relevant to them.

00;34;17;11 - 00;34;32;20
Unknown
So you can say, right, I'm going to use Codex here, I'm going to use Claude here, whatever it is. But they just have to bear the cost of it as well. I don't know if I buy that as a, as a good strategy, but it's certainly a good way of hedging your bets and seeing which tools are best for your organization.

00;34;32;20 - 00;35;02;08
Unknown
And I guess that's reasonable for sort of exploratory purposes right now. But the problem with that is that you're putting these AI tools in the hands of people that might not be equipped to really make them sing. And I see that a lot, right where you say, hey, here's this great tool. It can do so many things, but if that person isn't properly equipped to to really use that technology to, to get the maximum value out of it, they're just as likely to make a mess or run up a big cost.

00;35;02;11 - 00;35;24;12
Unknown
And then tarnish the name of that technology as they are to really extract value from it. I say this like I know what I'm talking about. I don't think there really is that many people out there that are well-equipped to leverage the full power of what these AI tools can do. But certainly you need to make an effort to upskill your team as much as you possibly can before you let them loose on something like Claude or Larry.

00;35;24;12 - 00;35;47;15
Unknown
Right. Well, speaking of people, yeah, yeah. So, mythos, what have you have you built your team differently than what a traditional SAS would have looked like? So certainly one thing we've gotten into, so we were structured as a traditional SAS to begin with. The first thing we did was was to say we, we embrace cloud code, which has been great just from a sheer productivity perspective.

00;35;47;17 - 00;36;07;22
Unknown
And this is something I've heard as well as that smaller companies tend to get larger productivity gains out of embracing these code tools than the larger organizations was. Google said that they got something like a 10% improvement in productivity. We got like 100%, 200% improvement, just the ability to ship code faster. And the quality of the code has been good, right?

00;36;07;22 - 00;36;34;06
Unknown
People say, oh, it's garbage. You all know if you've if you get people who know what they're doing, who know how to write code and can use the the code agents as a sidekick just to accelerate them, the code you push is genuinely pretty good, right? You can tell whether it's cramped or not. Structurally, where we have seen where we have made changes is and QA trying to get as much of the QA automated.

00;36;34;06 - 00;36;52;23
Unknown
Right. Whereas maybe you might have employed a couple of QA engineers. Now we can reduce that headcount. And on design as well. Right. So design has been probably where we, we embraced it. The most. And now you can empower your product manager to be a designer as well. Whereas in the past you had to have like a designer or two.

00;36;52;25 - 00;37;12;11
Unknown
Whereas now the tools that are available, you just don't need that headcount anymore. That's so interesting because I have I have found the opposite a little bit and I don't I don't mean that there's a right way. I just mean that like in the sense that, hey, you made this comment earlier, you whatever you build needs to be easily adoptable, right?

00;37;12;11 - 00;37;35;18
Unknown
Like it's so easy to build right now. The bar is so low that convenience becomes a defensible moat, right? If someone picks up a tool and says, or, you know, a software platform, whatever, and it's a lot easier for them to engage with it and extract value out of it, they're going to keep using it. Yeah. And a lot of that, it can be aligned to the UI, in particular, depending on the type of customer that you have.

00;37;35;20 - 00;37;58;07
Unknown
I'm not talking like dev tools. Right. And, and so when, when, when I've talked to startups that have that are engaging with SMB owners, legacy industries, people that are not, you know, that that on the on the range of like profitability to comfortability would prefer to like like to to have that squeezed where comfortability is also profitable, you know.

00;37;58;12 - 00;38;16;06
Unknown
Okay. Yeah. If that makes sense. And, and so, and, and granted, this is like, I remember having this conversation with killed a Pi this was shoot, 6 or 7 months ago. So I may have changed now where he was like, the tools are so good at writing code. I don't need as much software engineers or programmers.

00;38;16;13 - 00;38;33;17
Unknown
I'm hiring more designers to help me understand how to. But this is pre. Yeah, Clyde coming out with design and all that kind of things have changed. Isn't that crazy? Yeah, we asked best because we wrote. That's really cool. Okay. So there was a there wasn't. And you need to have a follow up conversation with him. Yeah. Yeah okay.

00;38;33;17 - 00;38;54;01
Unknown
So Claude came out with an MCP connector into Figma. Right. Figma being the main or one of the major design tools that were out there. Right. And so we built an agent that you could interact with Figma through their MCP connector, with Claude. So you and you could teach it, you could say, right, these are the common design elements.

00;38;54;01 - 00;39;12;02
Unknown
We use these colors. We use these icons just like you'd have a design book, but better than that. Essentially an interactive version of that, something that at an average person could use, like. And I am very average when it comes to design. And as soon as we built that, we used it for about a month. We high five were like, wow, this is going to save us so much time.

00;39;12;05 - 00;39;31;29
Unknown
And then Claude came out with Claude Design and we just don't use Figma anymore, you know what I mean? Like, because Claude can do it. Yeah, yeah. I would love to hear what your what your colleague says now. I mean, don't get me wrong, it doesn't get you all the way there. And this is the story when it comes to it.

00;39;32;06 - 00;39;48;12
Unknown
It comes to I everywhere, especially when it comes like vibe coding is I is really good at getting you to like 80%. And then you just hit this glass wall and the more you go, the the shit gets worse, you know what I mean, right? So you get to this thing, you're like, wow, I am a coding genius.

00;39;48;12 - 00;40;09;06
Unknown
I'm really to push. And you're like, no, you're actually miles away from production. You know what? I mean? But people don't think about that last 2015 to 20%. But we're getting better at understanding that. And utilizing these tools that are out there, like Claude designed to get us that 80, 90% level and then polishing it using traditional ways.

00;40;09;08 - 00;40;32;28
Unknown
Right. But having that presence of mind saying, okay, I've gotten everything I can out of this, now it's time to switch over to say, maybe traditional might be the wrong word, but certainly, I different process to to bring it home. Yeah. If you were, if you were to teach somebody because I think I would imagine this question would come up, which is like, how do I know when they're, you know, in a law of diminishing return, especially when it's compute cost get elevated?

00;40;32;28 - 00;40;58;10
Unknown
Yeah. When I need to stop using all these tools and start refining with something a little bit more traditional, you're spot on. Actually, diminishing returns is exactly what happens, right? And ironically, simultaneously costs increase, right? Because as you're building out this thing that you're vibrating, right, every, every prompt, every query, every like interaction with that gets fundamentally more token heavy as it gets bigger.

00;40;58;16 - 00;41;15;23
Unknown
Right? So you've got one. You're you're returns are going like this and your costs are going like this. So that genuinely is like an academic point where you should say stop, if you can figure out how to actually say that, then you shouldn't be doing this podcast. You should go just make that program. Yeah. You know what I mean?

00;41;15;23 - 00;41;34;08
Unknown
Yeah. What you say. Okay, cool. You've done enough. Stop. Now it's time to do it this way. I don't have a. We don't right now. It's just feel I haven't got good. I haven't got a good solution for that. We we build it out. We use it. It's great for prototyping, changes, whatever it may be.

00;41;34;10 - 00;41;57;07
Unknown
But it's really good at bringing new interfaces to market fast. We get that in front of a customer, we get the feedback, we can make changes on the fly and then push it. You know what I mean? We can't let you can't let perfect be the enemy of the good, especially when you're using these tools. Yeah, but if I'm if I'm someone working at a corporate and they've decided, hey, we're giving everyone cloud max licenses or whatever.

00;41;57;10 - 00;42;15;22
Unknown
And so you start playing around and you start seeing what's possible. You started pushing what Co-Work doesn't artifacts give you this visual representation of what you dream of, you know? Yeah. Maybe even code gets you, like you said, vibe coded to a point where you've got an application that you're spin up every single time you want to use it, like, yeah, yeah, yeah, but yeah, but it's happening, right?

00;42;15;22 - 00;42;34;07
Unknown
Yeah. And and it's like, how do you, how do I get educated enough to be able to say, okay, I've reached my point where I just don't know what I'm doing anymore. Yeah. Or what are the trigger points? And by the way, I have some ideas in my mind for me personally. Yeah, I don't I don't work at a corporate, you know.

00;42;34;07 - 00;42;50;15
Unknown
What do you got? Yeah. I mean, the corporates are dangerous, right? Because these corporations just need people to get on with the specific job at hand. Right. And they can use you can use AI tools to just accelerate, like, hey, help me read this Excel sheet. Right. Which is very much within. It's a it's a job that you needed to do, right.

00;42;50;19 - 00;43;10;21
Unknown
The AI as your sidekick can help you. I can see that being a very clear application within any corporation. Right? But it gets really dangerous when you get. And I was extremely guilty of this when I worked at Schlumberger. When they give you these tools and then you go off into the weeds and you start creating your own little business systems, right?

00;43;10;21 - 00;43;31;04
Unknown
And then you end up with this just forest of micro apps and that, that creates an absolute mess. And it was bad when all we had was office 365 and you had like Microsoft Flow and things like that. So you could make your own little flows and your own little power. BI I can only imagine how bad it is when you put AI into that.

00;43;31;04 - 00;43;50;09
Unknown
You can start genuinely vibe coding stuff. So I don't know, corporations to that extent need to be a lot more careful when introducing those kinds of tools to their, to their to their teams. Right. You need to make sure, hey, yeah, you have license to do this to, to use AI in these ways. And other people have license to use AI in different ways.

00;43;50;09 - 00;44;13;27
Unknown
Maybe that's the way it has to happen. Maybe you restrict the deployment of AI tools based on job code, right? What? Not not not saying you can have Codex and you can't, but certainly like you can use ChatGPT for these purposes, but you can't use it for these purposes. It's it's a it's a tough road ahead. Right.

00;44;13;27 - 00;44;34;29
Unknown
That. Yeah. Yeah yeah yeah. What what was your what's your I mean I think, I think dashboards are the new spreadsheets. Everybody wants their dashboard. And, and so everybody's spinning up their own dashboard right now because it's, it's not too difficult to get to, and, and to me it was for me personally is like when I do that and if I'm not in the dashboard every day, to me that's pointless.

00;44;34;29 - 00;44;54;25
Unknown
Like what? What was the point of building this or and or especially if I have to re spin it up, like if I have to go back to Claude and say, hey, remember this dashboard we built two weeks ago? Like, can we refresh it then? I haven't created a useful tool. And at that point it's like, I know what I want and I know how far I've gotten to be towards that.

00;44;54;25 - 00;45;15;06
Unknown
And and there's that gap. That's when I would say, hey, I need to call in someone to to kind of help me so that it really understands what's what. We're where we're going with this. Another one. And I won't name him, but is a land man at a at a dump. That was a buddy. And, and so, you know, I was talking to him, and, and this is a company that has one of those AI groups.

00;45;15;09 - 00;45;30;28
Unknown
Yeah, man. And he's like, this guy's having built me for two years. I've been asking for something, and I still can't do it, you know? And so he said, I'm going to just teach myself. And every night he would, you know, go on YouTube and learn and, and try to test things out and build and whatever. And then I, you know, this was maybe a month ago.

00;45;30;28 - 00;45;51;03
Unknown
We chatted. Then yesterday I call him and and for a separate thing, I was like, hey, I'm wondering about this, you know, how does the land man think about this? Right? Because I'm again, I'm thinking about what is what is the adoption from a land man from from a user perspective for a new technology. And like, how do you how is that ROI calculator built for that person.

00;45;51;03 - 00;46;14;22
Unknown
Right. And so we're going through that conversation like, oh, so how's the, you know, cloud training going. He's like, oh I haven't I haven't looked at in two weeks. And so it was like there was clearly a point where he was like, this isn't worth my time. And, and and if you can get to if you can pinpoint that and say, okay, you got to a point where you said I'd would rather spend time doing my day job, whatever that is.

00;46;14;22 - 00;46;42;13
Unknown
Yeah. Then continuing to learn how to build this tool that needs to be done, I think is a very clear indicator, very clear trigger of, okay, I do I didn't, I spent enough time on it. I maybe see enough value in it that I want to get to production ready, but I'm very clearly at a point where my talents, my strengths, my my hand, my, you know, my head space one aren't enough to get me there, and therefore I need to rely on someone else.

00;46;42;16 - 00;46;58;15
Unknown
Yes. No, I yeah, I buy that 100%. Yeah, exactly. And and the tool that you make might not be applicable to somebody else. The data feeds that you use might not be stable. There's so many things that the machine, the tool that you build might be great for that first week. And you're so proud of it. Then exactly what happens.

00;46;58;16 - 00;47;17;10
Unknown
You take the eye off the ball, maybe it runs and it's just costing tokens and you're not using it, which is the worst case, right? Or maybe, you know, maybe it doesn't cost you any money, but either way, it cost you a lot of your own time to build this freaking thing. That's what happens to what 90, 95% of these these things, they're just graveyards of this shit.

00;47;17;12 - 00;47;36;25
Unknown
But that's been happening for years before all AI is done is turbo charged it. You know what I mean? Like, as I said, with office 365, you had the ability to create flows and dashboards. Before then, everybody had their own excel sheet with macros in it. I'm sure that's that's probably where my career started, but I'm sure there's been equivalence of this back to the beginning of time, right?

00;47;36;25 - 00;47;57;00
Unknown
Where you just get, especially in our industry, when you've got so many engineers that love to tinker, they just going to put these things together. For me, I think where where large corporations are going to succeed the most, when it comes to leveraging the latest technologies to create value, discreet, like genuine discreet value inside the organizations at organizational scale.

00;47;57;06 - 00;48;40;10
Unknown
Maybe not like global, but certainly basement level kind of thing is to adopt an entrepreneur model. I'm a huge fan of entrepreneur programs, right. And that's essentially doing what I'm saying earlier about being able to pair genuine operational knowledge with genuine technology knowledge. Right. And so I'd love to see a place where these large corporations encourage these engineers, tinkerers who see a genuine problem and like a path to a solution, to be able to come forward and say, hey, yeah, this is the thing that I want to build, pair me with a resource that knows enough about tech to be able to build it properly.

00;48;40;18 - 00;48;58;14
Unknown
Right? I need to prove to you the ROI. Right. And then if you really do it properly, and this is the part that people don't like doing, pay the guy for it. You know what I mean? Like, oh, okay. You're going to you're going to create $1 million in savings for the company over the next year. Let's decide how we're going to how are we going to do that?

00;48;58;14 - 00;49;18;19
Unknown
How are we going to measure that? And then give the guy 5%, you know what I mean? Like maybe he doesn't have to own it or anything crazy like, oh, give me equity. But certainly having a some kind of share and the value of the tool that you're creating, and then you have to be discerning about what, what, you know, ideas get welcomed into the program and whatnot.

00;49;18;19 - 00;49;40;19
Unknown
And what don't. But you could have a technical resource taking care of 3 or 4 of these things. Back in the day, you used to have a whole tech team building these things. Was it was code. It was a whole undertaking. Whereas that's that's a that's gone. Now. You can have one very competent, they say product manager person who knows what they're talking about from the industry side.

00;49;40;21 - 00;50;00;08
Unknown
And you can have one very competent tech guy who has no idea how oil and gas works. But as I was with with, with coding these tools together, and most importantly, they understand the infrastructure limitations of the organization they're operating. And like, where does the data live? What are my data requirements? What am I allowed to do?

00;50;00;08 - 00;50;30;29
Unknown
What am I not allowed to do? You pair those two guys together, incentivize them appropriately. I think that's how you're going to bring genuine micro tools to the, to the, to the to market or to internal market as fast as possible. Yeah. Having a having a ring fenced technology team that just sort of roves around inside the organization, I don't know, I, I'm sure it's worked in some cases, but I don't I don't see the incentive model there to get these things off the ground.

00;50;31;01 - 00;50;45;28
Unknown
Do you see what I'm saying? Yeah, yeah. In fact, I would even push even beyond that. And maybe this is controversial, but from an organizational design side. Right. It used to be, okay, you had one product manager with a bunch of engineers. Yeah. I could see a world where it's one engineer with a bunch of product. Yeah, man.

00;50;46;03 - 00;51;08;25
Unknown
Yeah, yeah, yeah, exactly. Yeah, I was able to. It's like the pod structure flips and and that crazy. Yeah yeah yeah yeah. But no, you're right. That's exactly what you can do. You can have. But all you need to do is have one. Those technology people need to be ridiculously competent in the structures of the organization. And they can make AI coding tools saying they can get the solution to the point where it's good enough for the internal organization.

00;51;08;28 - 00;51;27;10
Unknown
Maybe it's not something you can sell. That's fine. Maybe it is something you can sell that's even better, but certainly something that can be brought to the entire organization in an efficient kind of way, you know what I mean? And then spearheaded by a guy that knows what he's talking about and then make them responsible for essentially the success of it and then reward them, maybe not punish them.

00;51;27;10 - 00;51;47;26
Unknown
But you see where I'm going with this? Otherwise, these tools, these ideas tend to just languish inside people's heads. And, you know, it's either going to just remain in that person's head or that other guy's going to try and make it do a half hearted job of it. And the situation we just described, and then that value is going to fall away or evaporate, whatever you want to look at it.

00;51;47;26 - 00;52;10;23
Unknown
Right? Yeah. I think IRA entrepreneur programs, I think are definitely the way to go, especially in the oil and gas industry where you have so many motivated engineers that genuinely like creating value. But perhaps they just don't have the appetite to to quit what is probably a very well paid job, right. And go off and try and start a company.

00;52;10;25 - 00;52;30;09
Unknown
Yeah, yeah. Which I don't know if I recommend you do. You have to like, I mean, where I've seen entrepreneurs leave, it's when they, they hit these like internal roadblocks or this red tape and they're like, man, I dealing with this is not worth the pay. I want to go build. I want to have a purpose. I want to, you know, be able to not be limited in what I'm what I'm capable of.

00;52;30;12 - 00;52;47;24
Unknown
And, and when that seesaw is, that's when it's like, right, I'm out. And so that's why I like these skunkworks programs, right? Because it's like, hey, no, no, you have full right to go. Yeah. Create whatever you want and then let's see if we can apply it. You're right. The entrepreneur program used to be right. There's nothing worse than.

00;52;47;24 - 00;53;16;12
Unknown
Oh, man. Okay, now I get we should have a few more beers in there. I really hope they, know that there's. There's nothing worse than coming into a room with a clear technology proposition for, say, a specific basin and then being judged by a manager who doesn't necessarily have the technical competency either. In the basin itself or in the technology that you're you're pitching.

00;53;16;12 - 00;53;40;07
Unknown
Right. But because they have a senior title, they're able to pass judgment on the idea that you've got you know what I mean? That's that that is entrepreneur poison straight away that that's what motivates people out of these large organizations. So if you're going to have an entrepreneur program, you need to have very clear definitions of what ROI is and a very objective value prop.

00;53;40;09 - 00;53;59;10
Unknown
A very objective process for assessing value. And then you need to give these guys latitude to go off and build it the way they want to, exactly the way you just described. Right. You don't have to give them millions of dollars like his 50 grand in credits. Make it shine. You've got you've got, you know, three days or two days a week can be dedicated to this project.

00;53;59;10 - 00;54;18;27
Unknown
We're going to keep paying you. You're going to keep your benefits. If we can see genuine ROI, you're going to see some kind of upside from that. This is how we're going to judge it. Right? And here is a budget, right? It might be magic money because it probably moves from another pocket inside the same organization by giving that person the latitude to go off and build it.

00;54;19;00 - 00;54;42;09
Unknown
Oh man. And then coming back and being like celebrating that success. I think the, the world is getting to the point where we can really make those programs work. And that's going to be there could actually be startup killing material as well. You know what I mean? That's where the a there's I oh, man. Oh, there's a great book called Goliath's Revenge.

00;54;42;09 - 00;55;19;28
Unknown
That's right. Right. Which is about how organizations can take large organizations can, function like small organizations. They can achieve that flexibility that smaller organizations traditionally have a great title, right? Yeah. And so it's a great it's a great book. But like, it's essentially a corollary to, that innovator's dilemma, right? That that's my favorite book. Right. Where basically saying that you've got these large organizations can't do genuine innovation because more often than not, it creates an immune response because someone comes with something that's better, that's going to disrupt the existing workflow.

00;55;20;00 - 00;55;47;10
Unknown
So no, we can't do that. You know what I mean? Whereas you got this, this book, at least Goliath's Revenge. There's a couple of other ones where they're saying, you know, if you can encourage internal innovation and you can incubate that in such a way and you can get the right structure around it, you can have you can you can put in place the right incentive structures and the right tools to, to see that value grow before the, immune response can, can take hold of it.

00;55;47;12 - 00;56;05;19
Unknown
Do you see? I'm going with it now. We're going off the reservation. I'm sorry about that. Oh, that's, Yeah, I, I would love to see these large. What are the successful examples of this? I haven't got any. I haven't had that. Yeah, exactly. Yeah. No, I, I think you're, you're going to find the, your average organization price to keep those things quite internal.

00;56;05;22 - 00;56;34;01
Unknown
Right. But certainly certainly I think that's the way that these large organizations can get truly innovative value from I you either invite small startups with oil and gas experience offering genuine value that's powered by AI, so they can essentially create value at a price point that the organization that the operator never could write. You can go to these large, these large operations like Palantir, and they're going to swoop in and you're going to have the whole that whole show.

00;56;34;03 - 00;56;56;17
Unknown
We can wait to see if anthropic learns how to speak. Yeah. Drilling. Drilling engineer. Right. Or you can embrace, sort of an entrepreneur program where you use the current technology like, say, to a yacht and pair it with people who genuinely know what they're talking about to create value and you create incentive structure for that.

00;56;56;19 - 00;57;17;08
Unknown
I don't I don't think setting up a crack squad of AI guys to just go around and harass operations people is necessarily the right way to, to create value, innovative value with AI. That's it. Giving everybody a ChatGPT license inside your organization and just make their life easier. Like, hey, here's a contract. Help me understand what's inside it.

00;57;17;12 - 00;57;37;19
Unknown
Yeah, interrogate this contract. That's something that that ChatGPT ChatGPT can do natively and reliably. That's genuine value straight off the bat. Like you can save an hour of every one of your workers times just by introducing that. But if you want those people to go further and stop building tools, that's when you can you can make a real mess.

00;57;37;21 - 00;57;50;05
Unknown
Do you say, I'm going with this? Yeah. So where is AI not been helpful to you or to smart chain?

00;57;50;07 - 00;58;16;21
Unknown
I'm I'm running through the the ways we've used it in our head, and it's actually been pretty freaking good on every case. Yeah. Every time we've chose to, we've chosen to employ it. It's been it's been very positive, where we've. Yeah, where we struggle with as these situations where we've actually built out at all. But then the technology has race past it, that our tool is obsolete by the time we build it and we build stuff fast.

00;58;16;24 - 00;58;31;07
Unknown
Right. Like that example of the design tool that I gave you. Like we, we put, we put like a couple of weeks into making that agent for design. And then as soon as we built it we were like, this is great. And then the technology came out and made it obsolete straight away. It was a real bummer, actually.

00;58;31;10 - 00;58;48;10
Unknown
Yeah, right. Yeah, stuff like that. And I mean, if we're if we're experiencing that, I can only imagine that every other day. So yeah, I mean, taking that as a lesson learned, how are you adjusting what to build? Certainly the writing was on the wall. I think if we had dug a little bit deeper, we could have found that Claude was working on that technology.

00;58;48;10 - 00;59;08;22
Unknown
So I think one thing that I'm genuinely trying to do that I think is an impossible task is to keep a is to just do my best to understand what the technology horizon is. So, you know, skate for where the puck's going to be rather than where it is right now. You know what I mean? That's it's insanely hard to do.

00;59;08;22 - 00;59;31;10
Unknown
And if you can figure out how to do it properly, then you should just start investing in the stock market. Right. But, certainly trying to understand, yeah, what the technology is going to be able to do by the time we, by the time we we're finished, rather than just assuming that what's available right now will be the state of the art for any reasonable period of time, like that's, that's gone.

00;59;31;10 - 00;59;49;00
Unknown
That's to the point you were making just before, right? Where you can make a tool and you could leverage that tool for six months before someone caught up to you. Like those days gone. Yeah. Yeah, exactly. Yeah, yeah. Well, cannabis was awesome. I didn't even look at my clock once this is, This is super. Any questions for me?

00;59;49;02 - 01;00;09;19
Unknown
I I'm curious, actually. How are you using AI in your world? Yeah. Yeah, I'm. I'm probably running through these cases like you were before. I, I think, the cloud code changed my life, I guess, in terms of, well, being able to integrate disparate data sources in a way that I never would have taken me, I.

01;00;09;22 - 01;00;38;20
Unknown
I wouldn't have been able to do it right, like, I would have needed, required some third party to help me with, and so, I think I think of like where the opportunities for AI are. Right. Because like, we go through this cycle of decision making as human beings a thousand times a day, something triggers us that causes us to then seek data to understand what is the world of possible decisions you could take, then select the best decision and then act on that decision.

01;00;38;22 - 01;00;59;10
Unknown
And we've been slowly objectifying each stage. Right. There's a trigger. Now, I've gotten to the point where, for certain triggers, I can go out and seek all that data consolidated, provide insights, provide whatever. And most people stop there, right. Insights. That's what everybody is excited about right now. Oh my God. It provides me insights. But it's the next layers that are super interesting to me.

01;00;59;10 - 01;01;20;22
Unknown
Right. Which is all right, now give me all the suite of potential decisions I could make or solutions I could I could take. Right. And and it's and then, then there's some fine tuning and teaching it and contextualizing saying, hey, these are these are the solutions I would do. What what am I missing? Right. Gap analysis. That's another fantastic value add for AI.

01;01;20;25 - 01;01;38;04
Unknown
And then right now I've gotten to the point where it's like I'm still selecting the decision to take, right? I'm still selecting the action. And then of course, acting it, I would like to get to the point where it knows better than I do, which I don't think will take long. What is the best decision to make?

01;01;38;06 - 01;01;53;28
Unknown
Yeah, it knows better than you do most of the time. You know what I mean? But you still haven't got that certainty where it's going to make the right decision enough times. Yeah, well, because it's probabilistic, right. Like it's it's based on the, you know, probability that this will execute in a successful fashion. We're going that drought. Yeah.

01;01;54;01 - 01;02;13;04
Unknown
But it's it's really hard to give it all the nuances that a human has. Right. Like I it's I mean sitting in front of someone and seeing their body and understanding that like what they're saying isn't necessarily what they're meaning. Yeah. It's hard to teach an AI to, to do. I'm not saying it's impossible, because I do think there's a lot of stuff out there that's showing that it can.

01;02;13;07 - 01;02;29;14
Unknown
But but I think there is an intuitive element to this right now, as some people would say, taste or judgment. Yeah, that is hard to have. I really replicate so where is I? And helpful? To me, it's like getting me all the way to this point where all I have to do is judge and act, and sometimes not even act anymore.

01;02;29;14 - 01;02;48;11
Unknown
Right? I just judge and then let the let the agent act, has been incredibly useful in every facet of my life, personal and professional. But you're not giving open claw your credit card number any times. No no no no no. Yeah. No. And other the other thing is like I, I, I do this a bit blindly right.

01;02;48;11 - 01;03;07;03
Unknown
Of like what is my, my token consumption to get here. Yeah. And in fact as we were talking, that's what I was writing down. I was like, I need to go back and review, you know, like, where where's my token spend? Yeah. Like like because I can tell you, like, getting from 0 to 1 or in one being, like a visualization of what's what you want.

01;03;07;03 - 01;03;45;12
Unknown
Yeah. It doesn't take that much right anymore. And but then getting one to be repeatable, that's fucking hard. Yeah. Yeah, yeah. And I and I, for my, I'm willing to bet that like at every step here there's like a token compute that to your point is getting exponentially greater or for diminishing returns. And so I, I would love the data to understand where I am in each of those steps so that I can be a little bit more prescriptive about, okay, when do I stop and start looking for a more traditional paths, or even using open source models, or using a DSP framework that gives me more coverage with less tokens spend, you know,

01;03;45;12 - 01;04;01;16
Unknown
that kind of stuff that I, I, I haven't peeled back the layers enough is as much as I should one of the one of the stories that I keep you tell me for over time. But like one of the stories you can catch. Yeah, one of the stories that I love. Okay. I have a love hate relationship with Elon Musk.

01;04;01;18 - 01;04;18;00
Unknown
A lot of people have hate relationships with him. That's okay. But I still one of the greatest engineers of all time. So I'm going to go ahead and quote him, right. He's famous for his design methodology. I mean, like the guys at work is sick of me telling the story, and I'm sure I've bastardized it now to the point where it's not actually relevant.

01;04;18;00 - 01;04;37;28
Unknown
But anyway, when he was building out Tesla factories in the very beginning, he was obsessed with automation and the new tech and the new efficiencies that technology could bring. So he automated to quickly. Right. So he built this amazing, manufacturing facility that was fully automated, but it didn't bloody work, or it didn't work fast enough. It wasn't achieving the efficiency goals.

01;04;37;29 - 01;04;53;06
Unknown
They wanted to. So they ended up ripping out an awful lot of it. And the lesson that he passes on is that the first thing you need to do is have your requirements right. You really need to be knowing what you're trying to do right, and you need to be able to cut, cut, cut. And then you sort of prototype it.

01;04;53;06 - 01;05;09;28
Unknown
You put, you mock it up, you put it together as best you can, then cut, cut, cut. And only then do you start considering automation. You know what I mean? I think one thing that AI tools are doing right now, as it's bringing automation into the conversation much faster because it's the easiest thing to do. It's the fun part, right?

01;05;09;28 - 01;05;26;15
Unknown
And so you end up with a situation you got right. You've got there where the design is actually, that's the easy part. It's wild to say that like the the spark of genius, whatever. But once you go through that design and the first step is easy, manufacturing is the hard part. 90% of the problem is manufacturing, which is kind of what you're saying here.

01;05;26;15 - 01;05;52;22
Unknown
0 to 1 is the fun part. 1 to 10 is the manufacturing part. Yeah. And you're right that I almost makes that manufacturing pot harder, right? Because it brings the automation up sooner and the opportunity to cut just isn't there. Yeah. Well, if you're anything like me, I my my prompts have gotten lazier and lazier, meaning like and so so of course Claude is going to give it a, you know, a two line question and it's like, boom, here's everything that's possible.

01;05;52;22 - 01;06;18;04
Unknown
Right? And if I'm not doing the effort of trimming it down to the point that it's actually what I want and I know what I want, then yeah, I don't it'll never be useful. The key to success, at least when you're doing one of these exercises, is context. Context is hands down the most important thing. And I think that you across the entire industry and actually data contextualization is is really important independent of AI.

01;06;18;04 - 01;06;37;23
Unknown
But obviously I can use it when we've been doing, we've been using, AI coding tools to help us, code some of the thing where that's going into the hardware that we're deploying and we're deploying genuinely pretty cool telemetry, hardware these days, certainly the sophistication of the firmware is far beyond anything we could have done with the book.

01;06;37;27 - 01;07;01;26
Unknown
Right. But the only way that we made the firmware code base reliable was by spending, like the morning just loading and manuals for every freaking component. Right? Here is the spec sheet for this resistor. Here is the spec sheet for for this power board. All of it. You pilot all up, right? Then you run, then you then you have a build, a version of the firmware.

01;07;01;29 - 01;07;20;13
Unknown
You take that firmware out and then you put it into the board and you run diagnostics against that. You take those diagnostic reports, put it back into the firmware, and then have a conversation with it. That way if you just try the lazy route, which is just chat, right? Hey, make this hey, make that head make this. It's going to make some dangerous assumptions.

01;07;20;15 - 01;07;37;19
Unknown
Some of those assumptions are going to sneak in in the first hour, and they're going to rear their head like two weeks later and you're like, what are you talking about? Right. But again, it's you have to know what good looks like. You have to know what it needs to be doing. Yes, yes. And you have had to go through this like trial and error of understanding what good it really is.

01;07;37;19 - 01;07;47;04
Unknown
Yeah. Otherwise, yeah. It's going to decide for you. Yeah. Yeah. It's not going to be good. Yeah yeah yeah yeah. Dude, no, this is awesome. You're awesome. Thank you so much, cam. Appreciate it. Cheers. My.