Todd Bush
0:18 Hey everybody, welcome to Chuck Yates, got a job, the podcast. If you get a chance this week to watch this podcast and we show them over on Collide, it's kind of cool because this is worthy of
0:31 listening to, but at the end of the day we run through a product demo so you'll get to see the computer screen and the actual Collide AI enterprise system working. It's pretty cool. So if you get a
0:42 chance to watch it, watch it. Thanks for joining All right, I got a text yesterday. It says I can tell Chuck Yates has a job because you haven't dropped a podcast in a while. So thank you, thank
0:55 you
0:57 for coming on, Todd. How did we meet you? I don't know that I've ever actually heard this story. You've been around for a couple of years now. I feel like we're really good friends. Yeah, so
1:07 actually through John Calphine. So John was an early customer of mine back in the day and we met each other through some different oil filled connections and then when we were, when Paulin and John
1:22 started talking about the community side, we actually started getting into detail about, okay, how could you grow with the community and what would that look like and able to come in and help build
1:33 out the community side and then started talking all about AI and workflows and the rest was history. What did you done to make us think that you could actually help build out the community like that?
1:45 It was all farce. Yeah, exactly. Yeah, so I was involved in quite a few of the former kind of oil and gas communities and was able to, was always interested in, and basically this community
2:00 aspect and being kind of in the thinking. I'm right in this verge of kind of a Gen X era where there's maybe a little bit of gap in the oil and gas industry. So the early communities that were out
2:13 there participated in a lot of them. And then really - We said participate, were you like writing code, business, or the above? Yeah, more business and market intelligence and what was happening
2:25 with rigs, frack crews, you know, sand that was being used in the Permian early on, the sand actual mines that were basically starting up back in, you know, 20, we'll call it like 2015, 2016,
2:38 and then helped, you know, thinking through kind of what, what the market was doing back then. And John and I kept in touch all that time as he was, you know, more on the frack side and selling
2:53 kind of, you know, I'd say frack and well tests at different companies. So how do you like do that? 'Cause you started at Chevron, right? Yeah, yeah, so I started at Chevron, basically new to
3:04 the oil and gas industry. And before that, I was kind of traveling back and forth between the coasts working on software and financial banking and basically wanted to be in Houston and had some
3:17 relationships and got into Chevron, where I focused on their digital oil field projects. And so got involved and got to work on water alternating gas and pattern injection and CO2 floods and
3:33 different ER projects. And so that kind of got me indoctrinated into oil and gas. You joined the cult. And joined the cult. And so I've been families, oil and gas as well. So I knew this was
3:46 kind of where I wanted to be and was able to translate that into some software and data and always been kind of interested in kind of data and kind of market intelligence and how that kind of
3:56 influences commercial decisions. It was a good fit for me, and I've been able to be in the industry for now 18 probably years. Wow. Yeah. Yeah. You know, it was funny. I don't know if I've
4:09 ever told you this story, but I was talking with Blair over at Mercury Group, and Colin had gone to Blair and said, Hey, we really need pod to join us full time. I think we need to make him chief
4:19 operating officer or president or whatever your title is, and Blair was
4:29 funny because Blair calls me and goes, What do you think about this? And I go, Hey, Blair, everything you hate about Colin, you're going to like about Todd.
4:40 I try to bring in nice even keel and focus on the kind of workflows and what we're doing, and I feel like there's a natural kind of transition in bringing the community, building out value for the
4:54 community and then also all the AI workflows. It's just It's great to be able to show outcomes and quick wins for operators and the teams that we're working with, so. Yeah, and so maybe just for
5:06 the audience, so just so you know, the last couple of, 'cause you should have better things to do than listen to my podcast, but I had Colin on and he kind of talked about founding digital
5:19 wildcatters, podcast through the community to AI enterprise software And then John came on and said, Hey, before you do that, you actually have to get your data in line. And stuff, tell me what
5:34 we've kind of learned 'cause you and I've been running around so and so far now for about a year. What have we learned? Yeah, so I feel like one of our big learnings would probably be around search
5:46 and the whole value proposition around search. You know, that's something, there are a number of companies that immediately see value in it and want to dig in and understand what the completion
5:56 designs were, what - what's happening at a certain well that they need to investigate. If they have an entire kind of well history where they need to do the research and analysis to understand the
6:08 root cause of a failure, those kind of well search examples are great. 'Cause I mean, well, 'cause I mean, if you step back and you think about it, I mean, some of that stuff happens literally
6:18 going to a banker's box. Oh, yes. I mean, like - Yes, like, and not just one banker's box, like - Well, like, hundreds. Yeah, well, and you do that, and then maybe you get it digitized.
6:30 Maybe it's in one of your legacy software systems you're using that see? Was I did Where. where that WellView? Was that here? Was it exactly remember can't you but, there? And so, yeah, now,
6:35 I think, you
6:46 and I have talked about this, you got Silicon Valley, you told people, we use less than one percent of the data that we've accumulated as an industry, 'cause we just can't find it.
6:57 It's in the ether or on a piece of paper or,
7:00 you know, a PDF somewhere that maybe wasn't related to the well. And so there's a rich set of information when you start looking at the post-job frack reports or maybe a mud report that happened
7:13 years ago that maybe the team acquired the assets and they don't necessarily know exactly what was done and why it was done And so we can uncover those kind of needle in the haystack kind of problems
7:26 as well as then turn around and build kind of workflows on top of it. And so that's an awesome thing. I do think that was the GPT condition, the market. I remember the first time I used chat GPT
7:38 and I went, crap, I'm never going to use Google again, you know, because you actually, you got an answer And then, you know, 20 minutes of using chat GPT, I pulled up perplexity and it's like,
7:50 holy cow, it gives me an answer and it links the source. Yeah, I'll never use Google again. Yeah, and now executives are looking at chat TPT, maybe evaluating contracts, gas purchase agreements,
8:03 anything like that, there's definitely some real simple kind of use cases. I think where we see a tremendous kind of opportunity is, well, now we can take those gas purchase agreements and build a
8:16 workflow around it. So a team knows how to begin renegotiating or looking at the dedications and seeing if there's any overlap in the dedications for different assets. And there's just a number of
8:28 those workflows that provide a whole lot of value to different teams. Yeah, I think one of the things we wind up talking about with clients is it's sequential in nature. You basically, and John
8:42 and I talked a lot about this on the podcast, you got to chunk and bed that data, slapping that data on it, get it in the vectorized database. And when you do that, you can start searching That's
8:52 a byproduct of that. but you have to do that work to then automate these workflows. Yes, yep, and we're seeing, as we get in and talk to many different operators, everyone's a little different,
9:06 but essentially understanding where some of the wells are, some of the well history, we can then take on different regulatory workflows, understand maybe there's root cause analysis that we need to
9:18 look at and build out for teams that are looking at workovers And maybe it's even getting into some of the field operations side where we can start generating JSA's and really understanding not just
9:31 what's happening in the field, but then taking everything, all the internal kind of corporate knowledge and helping the teams in the field actually use that and apply it so you get safer operations.
9:43 Yeah, yeah,
9:47 one of the things Colin says that I like he says is, he says ultimately we have to sell an outcome. You know, so when you automate that workflow, the finance broke in me like that, 'cause you can
10:02 usually quantify it back to money. We spent 275 hours a year doing this, copying and pasting these things. We've automated this totally, 'cause that's what AI does really well, searches through a
10:14 lot of data, grabs what it needs, you can generate documents off it, and being, it now takes 30 minutes So you can actually put a number to what an automated workflow, ie. an outcome is. That
10:32 being said, I wanna make this one case, and you can just roll your eyes at me if you want. I truly believe when you chunk in a bed, get all your data in there, and you turn really smart engineers
10:45 loose on just querying stuff. 'Cause one thing AI does really well is it finds correlation It may not, it doesn't tell you causation. but I think there's gonna be like a million things we're gonna
10:57 find, like stuff that things like, if you have a left-handed pump or your wells fail, 35 more often, there may not be any causation to it, or it may actually be, no, they're using the wrong
11:11 tool or something, you know, who knows what it is, but I actually think they're gonna be big-time game changers for people doing that, but it's hard to quantify. Yeah, absolutely, and we're
11:24 going in with our kind of forward-deployed engineering teams to basically say, here's the value statement, but there's also all those kind of creative, kind of, you know, thought-partner type of
11:35 use cases where maybe it's just a little bit of, like, answering, getting answers about specific areas that kind of generates other questions where you can start seeing some of, maybe I didn't
11:47 really understand what happened 30 years ago on this well, or maybe it's a new well that's been drilled and there's some kind of unique. kind of completion design practices that
11:59 were used for that specific well. And I think there's a lot of just uncovering data to make it a little more transparent across kind of operations. Yeah, I mean, it felt like, 'cause we were
12:09 doing early stage assets at Cane and it felt like we were creating this monster spreadsheet on every play where you literally laid out every well on every variable and you eyeball, regressed all this
12:23 stuff And I remember one play we made a fortune on and what we figured out was the feet of pay you needed was binary. You needed 15 feet or more. It did not mean that 22 feet was better than 17 feet.
12:41 It was solely binary on that. If you were below 15, you did not make a good well. Didn't matter what the frack was. And above 15 or it was 20 or something, But it was literally just binary, and
12:53 if you looked at the wells on a binary basis just on the basis of that, you had to play. And so it became geologists go map it, get enough 20 foot in there. Great, we could draw good wells. But
13:06 AI is going to be a game changer doing that type of stuff. Oh, yeah. And there's so many kind of unique, you know, it's so good at going through large volumes of data versus as, you know, if
13:18 I'm in kind of operations And I have to go back through the paper or I have, you know, 15 wells that I need to review that week, you know, we can actually apply a little AI to that to help
13:30 automate that, but also free up some time for free to do higher value work. Yeah, no, that's, that's, that's interesting So,
13:40 my sense is run around cell and AI like we have, getting applications installed.
13:50 We've done a lot of cool stuff. We've got a company using it on the marketing agreements, like you were kind of talking about earlier. We're building some proof of concepts right now, where we're
14:01 going to look at expenses and funding duplicate bills and a lot of that. But I think, too, have really kind of popped out that we've gotten big, huge wow factors on And the first is the automation
14:16 of the G10s and W10 filings. What are those? Because mom might be listening. Explain it to mom. Yeah, exactly. So every single active producer in Texas has kind of compliance that they have to
14:31 meet. And if you're familiar with the Texas Railroad Commission, that's kind of the regulatory body that manages and monitors kind of all of the Texas operations. And so each year, a couple times
14:45 a year, those operators get a notification letter. from the Texas ROC, where they basically have to report back their testing the wells, they're in good shape, and they basically have some
14:59 indications in there about
15:02 if it's active, if it's shut in. And so they have to do that for every single oil and gas well. And those are the W-10s and G-10s. And that's kind of not the most perfect analogy, but the car
15:14 inspection, right? Yeah. Annual car inspection. Yes To know that it's there, actively producing. And then even after that, there's, you know, several other kind of regulatory processes that
15:27 take place. You know, I think everyone's familiar with kind of the application permit to drill. These are kind of after that, once you're actually producing. And then subsequently, you know,
15:39 after maybe a well is taken offline or it's no longer active, there's some other kind of regulatory processes that we're going to manage as well. Yeah, 'cause the push pull just on wells are, you
15:53 don't want to plug a well prematurely if there are other activities to go do in that well, behind pipe zones to add, et cetera. Can you convert it to an injector, various things, but the railroad
16:10 commission does want you plugging and abandoning those wells according to procedure and code, et cetera, to protect the environment. So that's kind of the push and pull on all of this stuff. It
16:23 was wild back when I started at Cane. I mean, we didn't even talk about PA and wells. It's just like we'd buy all these wells. I would get to it wherever. And then probably after about 10 years,
16:36 we started at least putting a list together that we'd PA. Now they were 40 years from now. Right about the time I got fired, That was the first page in the board book. was the regular PAing of
16:47 Wells. 'Cause I mean, I think Sarah Stodger and others would gripe that the railroad commission's not doing enough about that. But, sure. Yeah, they do push more. So, cool. The, you got the
17:02 computer. Yep. Give us the demo on how. Show me one of these letters. Yeah, perfect. So let me pull up maybe a short one that I can show and what you would basically see in this letter Obviously,
17:16 as an operator, you would get this letter both kind of in the mail and through email, but it's basically going to give you a list of fields and leases and wells that are kind of due for the W10. So
17:31 this is for active oil wells. And you can see I have about, I believe, 14 wells on this particular example. But traditionally, an operator would receive this They have access to a lot of. you
17:44 know, internal data, they're going to possibly create a spreadsheet, begin organizing it, maybe they're going to go and search in there, kind of whatever well database they have, whatever
17:56 production system they have, pull all that information together and then produce that PDF report. You can also have some other options to file it with the ROC, but in general, seeing a lot of kind
18:08 of spreadsheets and, you know, working with the data. So we came in and talked with a couple of different operators that spend, we'll call it kind of 500 hours a year type of effort generating
18:23 these W10s and basically created a way. And this is not value add. I mean, this is literally copy and paste information from here to there. Correct type stuff. This is reporting. Yes, exactly.
18:36 Got it. And so we created the way it's basically take that in our kind of our system right now. So AI is going to read through that letter, pick out all the leases and the wells that are due. And
18:50 then we're going to query kind of a database to get the other well information. And this is like the company's database. Yeah, this would be all the company's database. And then essentially taking
19:00 the well information and then querying a, what is a, essentially a production database that would show all of that well-test production information coming back. And that's like Scout, Merrick,
19:16 and who else has a correction database? Like field direct is one. There's a couple others like Greasebook and the handful of other kind of tools that are out there for different kind of operations.
19:27 And we'll be pretty transparent in showing, all right, here's all the activity, that kind of the AI process to grab that data and enrich the information. And it's going to show kind of the
19:39 districts that we're looking at. So we have wells in District 7B and 8A.
19:43 And the beauty here is that I basically just uploaded that document and then gave it to the system and it created the PDF. So right there, we're taking those hundreds of hours that, you know,
19:55 typically are required to create kind of the W10s and G10s, compress that down to, you know, a couple of minutes. And now I can just verify the data and make sure that, okay, yes I got all my
20:07 leases. I know which ones are shut in. Um, I have the correct kind of oil production, kind of well tests that are there. And there's something, there's something the railroad commit, and this
20:17 makes sense. The railroad commission flags a well. If you show a well test that was greater than a previous one, yes, because, and that makes sense because I want to know what happened. Did you
20:27 work it over? Did you do whatever? And so one of our trust, but verify, uh, things on the thought process. doesn't it say, Hey, we looked for that and we didn't see it? Yes, yeah, so we're
20:39 kind of pulling out the max amounts so that you know, okay, there's no well test that exceeds that kind of max amounts and we're gonna verify those rules. Yeah, yeah, allowables and stuff like
20:53 that too. Yes, exactly, yeah. Okay, exactly. So that's kind of the, that's the just kind of - And you connect this, can't you? When it pops up there If I see something here, it may be for
21:04 the Garza well, if there's something there that I don't like, maybe I think that's eight barrels instead of 10, or I have specific information in here you wanna change, can take that, make those
21:16 edits, and then basically download and send the document to the ROC. So that's cool, not
21:26 that we would do this, but theoretically no human would have to be involved, right? Yeah, so there's already some shows up in an email. Yep, we can grab the email, actually create it all the
21:38 way through and then think we're gonna begin having some other conversations about actively taking it and pushing it to the rubber commission so that if we get to the point where somebody wants to
21:51 step back and let the entire process run, then we'll have that option for them. And the neat thing about this, so our CTO, Kineshius, who will come on the podcast at some point, he and I were
22:04 talking, I mean, he says, you know, we've written this, it's been battle tested now with a couple of people, seems to be working really well. It's like easy to install. This is a couple of
22:15 weeks. We just basically have to wear you keeping the data and make sure we can talk to it. Yeah, and we're building out all these connectors for the different kind of well databases and production
22:26 systems so that we know as we go in, even if it's an on premise. maybe feel direct. We know that we can connect to that through a couple of different ways and even as companies are spending so much
22:38 time kind of on their data efforts and making sure they understand where everything is, I think we can help connect to those individual systems or potentially even those kind of data warehouses where
22:50 they're starting to store more and more well information in production data. So either way is a good option. Cool. The
23:00 other thing I wanted to have you demo real quick is we've got a client that's a minerals company and I probably won't do this justice so correct me when I'm telling the story but basically there's
23:14 been a lot of money each month with, in effect, a sweatshop in India, manually keying in data off revenue statements because. as the owner of the, the mineral or the royalty, you get the, you
23:31 get a wire or an A, C, H payment with your money. Then you get a statement saying, here's how much money you got, here were your volumes, et cetera. And they don't seem to be really nice in
23:44 providing that in a digital format. So most people are out, actually out there keying that stuff in, I think. Yeah. And it sounds like, you know, from all the mineral firms that we talked to
23:56 and even kind of some of the non-op groups within operators, basically getting all that information in, obviously, they're getting it in PDFs and they've scanned it in, but we can actually then
24:04 help them extract that information and use, all right, here's, here's what you're getting paid. Here's what kind of
24:14 the property is showing from a number, like a volume perspective from each individual well, and then, and then, build off of that. So we can take that data and then create, maybe it's just a
24:29 spreadsheet that you want, because you want to verify it. Maybe it's a, what's called a C-DEX kind of export. So the accounting systems, most of those production accounting systems will accept
24:42 kind of an import. And so we're working with a couple operators to basically define that, see how they want to use it. And the one that I can show is, you know, I can pull up a single revenue
24:53 statement. Let's just do that. And, and basically take you through. Okay. Same similar process of upload the file, kick it off. And then we're going to give you kind of this little thought
25:04 process that shows, all right, now I'm going to extract the revenue statement. It's going to show me, okay, how many statements were parsed? What was the total check amount? And, and then give
25:16 you an idea of some of the owners and other information and what that will do is basically give you a In this case, we're exporting to a JSON file, but that's really just for our development. So
25:29 that can be a spreadsheet, that can be an import file format for an accounting system and begin kind of helping the back office to basically use that information in a way that they haven't been able
25:43 to get. Yeah, 'cause I mean, I would imagine a slug of the information where extracting needs to go into accounting because it's money It's revenue, stuff like that. I also think though, the
25:56 volumes, in terms of oil and gas volumes, you're potentially gonna wanna grab that and send it to Aries or PhD Win or Combo Curve, wherever the case may be, 'cause a lot of these mineral funds
26:10 have gotten really sophisticated and are doing their own engineering on stuff. And that's actually been a competitive advantage to certain mineral funds They know the assets a lot of times better
26:22 than other folks do. Yeah, and you get the actual kind of volumes that you can see from both oil, gas, NGLs, all of that information coming through. So it's a rich kind of set that they can
26:34 actually do a number of things with. And
26:39 now we're kind of starting with some of the revenue statement to accounting side, but I feel like it's gonna open up to so many other opportunities once they have a usable kind of data set Yeah, and
26:50 again, I mean, we're talking, I mean, hours worth of keying work that now goes away. Yeah, yeah, and like this example, I think there's six PDFs that I just uploaded. So it just went through
27:03 all six, gave me a little summary, and now we're even expanding that. Obviously this is one that's showing kind of more human in a loop. So somebody actually doing the work versus maybe just
27:15 picking up those files in a drop box in you know, a one drive somewhere where all those files are being loaded and we just automatically run through them. If we're sitting, so these are to use
27:29 Collins terminology, these are outcomes, you know, so we've gone in and we've built outcomes. It's built on at least a slug of the technology that's in effect commoditized across all our products.
27:44 Kind of the AI rag model We've got maybe 20 of it is kind of bespoke code on what kind of outputs you want. How do you get your inputs and stuff?
27:57 Where does this stop? Does it stop? Talk about the future. Let's go like Dreamland. Yeah, so crystal balls. So crystal ball wise, I love the opportunity to come in from an operator perspective
28:12 and look at an entire kind of in the end workover solution. so that you have something that's, you're looking at the failures, you're understanding the root causes for those wells that are going
28:23 offline, all the way to scheduling jobs and scheduling kind of the work, to getting the work done kind of from the field operation side. So that's one that's like, it's a great kind of workflow
28:34 where we can interact with different groups and different teams within one operator, as well as kind of vendors and contractors. Well, you gotta think when you can be more holistic on the view of
28:46 everything going on with that well,
28:49 that you're ultimately gonna be better. I mean, that used to just drive me crazy that operations would never even talk to Reservoir. And you're like, we have a type curve based on a certain type
29:01 of frack. You can't go pump less volumes and expect to get those reserves. Do you all not talk? Yeah, right. And then even getting to the point where, even completely different workflow, but
29:12 one that comes up a lot from the kind of oil field service side is. you know, interacting with sales and the sales team, having to answer questions about maybe its specific products or even take
29:25 kind of a frat company, for example, where they have an operator that they're working with and they wanna understand the best completion design for the Haynesville. Well, there's no reason why we
29:36 can't work with the frat company and get all of that post-job reports in those completion designs Creative way where not only they have access, but also an operator has access so that they, I mean,
29:50 they're basically have access to a lot of that information anyway. And they could then start interacting together within the same platform. So within Collide, there's so many unique opportunities
30:04 where we can drive workflows that are just internal at operators, but then touch on others where we're actually interacting with vendors, other contractors as well as kind of the. entire supply
30:18 chain. Yeah, I'm going to be disappointed if in five years we're not sitting around crowing talking about how we made the CapEx spend among our clients at least 10 better, maybe 15 better. You
30:33 know, I mean, they're just, there's, there's, there's so much if you just, like I said, take a more holistic approach on things, factor it through just tiny little bit of lessons there. As,
30:48 as well as, you know, we haven't really talked about risk management, but I mean, we as an industry spend so much purely in just insurance premiums, as well as then, you know, real life costs
31:02 when accidents happen and, and being able to reduce that substantially, I think is going to be important too. Yep. Yeah. And we've, we've talked a lot about this one too, but it's the companies
31:12 where they still have kind of a growth mindset. Obviously the market's a little challenging right now, but from an operator perspective with a team in place and they're trying to grow
31:25 their production base, not having to hire in the same way that they had to five years ago, where we can help them kind of grow production with maintaining the team and maintaining kind of the
31:38 engineering discipline and operations teams in place. And so you can see this place where we might be in five years where we have some case studies where operators went from maybe 500 to 2, 000
31:51 wells and it's the same engineering team and we have collide workflows running both from
32:02 the safety side, environmental, regulatory, workovers and even getting a little bit into more of the kind of production operation side. Yeah, the more time I spend with this, the more dismissive
32:10 I get of, Hey, I was gonna steal my job. Somebody that knows how to do your job with AI will take your job if you don't learn it, but it really is a huge enhancement tool. And I tell my kids
32:23 every day, please be playing with this, figure out how to use it, because at the end of the day, it's really that. Yeah, yeah. And absolutely thinking the regulatory teams that we're working
32:33 with, they've been so welcoming and just their eyes light up when they see the W's energy team work work. They do this shit Exactly. It's like who likes to copy and paste? Exactly. So I feel like
32:46 there's that angle of, we will not only help kind of automate the workflows but help teams enjoy what they're doing on a daily basis and take some of that root kind of
32:59 non-value added work and actually free up some of their time so they can spend on more strategic things. Exactly, exactly. Dodd, you're cool to come on. Awesome. Thanks for having me.
