Can the machine really understand leases?
Join Mark LaCour and Tony Benson in this all new oil and gas podcast for a discussion on artificial intelligence and processing oil and gas leases with Grooper.
Listen to the podcast in its entirety here: https://oilandgastechpodcast.com/tony-benson/
And here's your TL:DR (listen!) guide:
- 1:40: What does Grooper do?
- 2:33 How many people have to write code to use Grooper?
- 2:54 Unstructured data - how to deal with varying source data?
- 3:57 Making an engineer's jobs easier and providing insight
- 5:39 You don't need IT, add-on APIs or a data scientist to get results
- 8:26 The BIS | Grooper story
- 11:07 Preserving tribal knowledge
- 13:14 Preserving and creating jobs
- 14:33 What is Grooper's artificial intelligence?
- 17:04 The value of Grooper - make higher quality decisions with more quality data
Mark: So Tony, what does Grooper do?
Tony: That's a great question.You know, for this conversation - for this context, I think what Grooper does for oil and gas companies and in the energy sector is to be able to process data for usable information. It leverages artificial intelligence and machine learning to give quality data to make virtually any type of business decision. Whether it be on the drilling side, completions, on land acquisition / divestitures, you can only make a good decision, the most informed decision based off quality data. And that's what Grooper does. I don't want to get into the jargon of it just yet, but in the end that's what Grooper does, is provide the quality data to make better business decisions across the enterprise.
Mark: Now, that's the marketing speak which I totally get. But there's a bunch of stuff that you do that I think is really cool. So, how many people have to write code to get something like this to work?
Tony: Absolutely zero.
Mark: Did you hear that audience? There's nobody slinging code. This is a totally different way to integrate machine learning, natural language processing, all that sort of stuff. Tony, ya'll have done some really cool stuff with what I would consider non-structured data. Right? So literally printed out PDFs and stuff like that. Let's talk about that a little bit. Because that's huge in this industry.
Tony: Yeah, very huge. There's a lot of unstructured data sets because, you know, the bottom line guys, is you cannot control the source data. Especially in this industry. You need something that will understand the intent of paragraphs and the intent of what the document is supposed to be to a lease analyst or a drill tech or what-have-you is looking at it. So, that's what it's able to do.
Mark: So it's literally able to understand what it is. So if it's reading a mineral lease it knows it's reading a mineral lease. If it's reading drilling logs, it knows it's reading drilling logs. I mean, this is enormous - this is huge. And the other thing that you do that I think is cool, is you're actually out there making engineers' jobs easier. And we've talked about this on this show and other shows, but in oil and gas a lot of an engineer's job is combing through massive amounts of Excel spreadsheets trying to get the information they need so that they can actually do their job as an engineer. And you eliminate all of that and you hand the engineer exactly what they need, whether they're a structural engineer, a petroleum engineer - they get to do what they're good at, not combing through a bunch of Excel spreadsheets.
Tony: Yeah, exactly. Let's talk about a real live scenario. Let's say you're a small / medium / large oil and gas company and you're getting hit with all these operational reports coming in, right? Anything from nonproductive time, the bottom hole assembly, what serial number was on the mud report, why was this information put in when there wasn't supposed to be NPT and there was. You know, those are the things people manually enter in, trying to understand versus "How much money did we pay for this particular surface casing run? What vendor did we use? Why did we not save three hours here?" Like I said, it gives the engineer, the lease analyst, the people allocating those AFE's the insight to be able to make those decisions faster. And that just one example on the operations side of things. So that's a real life scenario of Grooper, the software taking in all that information, and the end user or the analyst or engineer is able to make that precise decision based on where they should drill, where not to drill, what bit to use, the list goes on and on. This automates that process for faster, and quite frankly more accurate decisions.
Mark: Yeah, which is awesome and it almost sounds like science fiction except I've actually seen it so I know it actually works! One of the cool things, and please no hate mail from I.T., you don't need I.T. This isn't a bunch of stuff that has to be integrated. Ya'll don't need APIs, you don't need add-ons, you just layer Grooper over and even if that data is disparate - different silos, different databases, Grooper doesn't care.
Tony: No it doesn't, and in fact that's why I mentioned that specific operations use case. Going back to the unstructured data - what if you're in the land department and you just had a huge acquisition. I know the land guys out there are tired of not knowing what the heck they're buying or selling. It's maybe 25%, I've heard. It's just not acceptable. Because that huge amount of information is coming at you it's really hard to get to. There are a lot of data scientists who are required - a lot of template building and a lot coding being built, and what this software allows is a unified platform, able to go ahead and look at this data without having to spend so much time on the engineering of the software itself to get a result. We're just interested in putting the software in and getting the result without all of that extra effort. Which in the end, it enables to scale across the board instead of having singular modules or singular points of contact for a specific solution, that's not the way to go. There's too much information, too many departments, too many people to be able to have that scale where you could afford it, quite frankly and we've broken the mold on that.
Mark: Even if a company tried to do this themselves, if they build their own in-house system, the moment they turn it on, it's already antiquated. Where Grooper is constantly changing, constantly learning, constantly being updated. So instead of the oil and gas company being worried about I.T. types of stuff they only have to be worried about drilling for oil, or pipelines, or refining, or whatever and ya'll take care of the business use of the technology.
Tony: Well, yeah, and it's two-fold. I'll be very clear, BIS software, publisher of Grooper is not the one drilling the wells. The subject matter experts and the people that are there drilling the wells, that is their job. To get that oil out of the ground, to process it, to send it wherever it needs to go. That's not our forte. Our forte is dealing with the data, or allowing software automation, machine learning to be able to give all that data to the people who then make those decisions based off that data. That is our expertise and you know, we've done that for over 30 years. With a multitude and a variety of data sets. And one of the biggest pain points that we've experienced was as a reseller of a lot of this technology which exists today, until we created Grooper in-house over a fifteen year period. What was that problem? Well, when data changes, we have a problem. Which means you're consistently having a problem. Especially and it's been exacerbated in the oil and gas industry. So, it's really taken off the last two, two-and-a-half years, and it's been very, very exciting.
Mark: Yeah, and so we mentioned earlier how Grooper can actually read stuff and I wanted to be very clear to the audience, I really mean read it and understand it. Which is - it's easy to have optical character recognition - that's been around forever, but to understand the intent of what this document is - is incredible. It frees up the people to actually do what they're really good at. Now, you mentioned that Grooper has been a fifteen year process - what's the backstory to that? What was the impetus that started BIS to looking to do something like this?
Tony: Yeah, so we were a reseller of a lot of products - various capture platforms, we used to process a lot of lease data from various counties and parishes from around the United States, a lot of hospital records, and we would stand up these systems both in the front end and the back end. And we saw a problem from within as we processed these documents for companies, and as a reseller there was a huge gap. One of the first gaps we noticed was classification. Could the software reliably read the classification type? Because if you can't classify a particular document, whether it be a drilling report, a completion report, a particular type of lease, it doesn't really matter what the extraction points are that come after that. So, it really developed on the classification side. And then we figured, "These are some ugly looking documents, is there a way to have some NLP if there's a smudge in the document, if this sentence structure breaks down or if the page is literally 180, if it was written and authored 100 years ago and there's certain sections of it missing?" We take all of this in a platform scenario of image processing, OCR, classification, extraction, normalization, standardization, as an entire platform and that's what we wanted. That's what we needed to develop, which is exactly what we did. And now, we've found our home here in oil and gas.
Mark: I think that's really cool. So ya'll had the need internally dealing with your clients, so you just built it yourselves.
Tony: That's exactly right. I'll be very clear, there was only one Einstein who ever lived in this world and I'm pretty sure he's gone. So we don't have 3,500 Einsteins running around our building, but what we do have is fifteen to twenty years of some failures, of not being able to stand up those systems. When people say it's able to understand intent, we didn't feel comfortable making that statement anymore. Based off of those failures we developed our learned and applied knowledge, what exactly we should be doing. Think if you're a drilling operation, a lot of that knowledge is based off of experience. We shouldn't be using this particular bit, I don't know if we should be drilling here in this particular area... a lot of that knowledge is from mistakes made in the past. It's the same scenario here at BIS. We developed a software based off of the mistakes and the gaps from within it and built a platform around our client's needs and our needs. And to this day, we use Grooper in-house. So whatever company, whatever major upstart, whatever - is using Grooper, I have good news; we're using Grooper too, so it's consistently being developed.
Mark: You brought up a good point as far as a lot of stuff that's done in this industry is based on experience and knowledge. And we're facing a talent shortage of epic proportions. All of that knowledge, or a lot of that knowledge is disappearing as people leave, as people retire, and so with the ability for Grooper to come in and look at all the different data silos and learn how the business is done - it's also to help the new young workforce who comes in to keep from making the mistakes of the past. They don't necessarily have to learn from experience, they can learn from the data that, up until know you couldn't really get your hands on because it was all over the place - some of it was structured, some of it wasn't, some of it had taxonomy, some of it was in this type of database or that type of database, on somebody's hard drive or laptop, now all that's gone.
Tony: Mark, you bring up a really good point that I should have mentioned earlier. The tribal knowledge aspect of this. I worked as a consultant at Apache, BHP / Petrohawk, and as everyone knows in the industry there's a lot of consultants who come in and out. A lot of awesome knowledge comes in and out and there's a lot of smart people in this industry. A lot of experience. And they come in and bring that experience with them. But when they leave, that experience is gone. That tribal knowledge is gone. So with going back to the idea of new people coming in and the old people leaving, I mention our artificial intelligence and machine learning - that knowledge that the software is able to obtain from leases in the Balkan, the Utica - that knowledge from various operators around the world and around the United States, it's learning on those documents. So Grooper, if you think about it - is the employee you train once and never asks for a raise. Because it's consistently getting smarter, consistently learning, and you bring up a really good point with a new employee coming in. Those people would be able to say "Oh, here's the data in front of me," instead of having to gather it all up or have to take that time.
Mark: I like how you phrased that - the employee you train once and never asks for a raise.
Tony: Yeah, and you don't want to say that too many times. And for the record, just to be clear. Artificial intelligence kind of gets a bad rep. It's a buzz word nowadays. We start having these conversations and people wonder "Is this going to take my job?" That is a completely understandable and normal reaction, but the truth is, is that I have found personally that there are ups and downs in this oil and gas industry. This allows us to be able to streamline, to save jobs. This technology is coming so if you get ahead of it, and understand how to use it, it's overall better for your career, for your company, etc because it is coming. And I think that's an important distinction with this type of AI. And we can get into what our AI actually means, but I just wanted to make that point very clear. It's been my experiene that it's actually creating jobs because of so much information coming at us. Everyone knows in the oil and gas industry that there's so much data we don't see that's several - lot's of companies paying millions and millions of dollars for - they never even use the data. But what if they could? Would that allow for more jobs? I believe it would.
Mark: And you're also driving efficiencies. Efficiencies drive cost down which means you can afford to pay people more or you can keep the people that normally you might have had to lay off. It makes total sense to me. It is interesting you brought up "What does AI mean?" From your point of view, what is artificial intelligence?
Tony: Well, from our point of view, from Grooper, before I came here I would have guessed the Matrix, but simply put, everyone's definition of AI is different. And it should be. It's a very broad topic. Hence the buzzword. But simply put, our AI is synthetic understanding of data. Let's take it down a little bid deeper than that. So what does that mean? It means that it's able to understand data the way the business would like it understand it. In a form of what we call a content model. What's a content model? Let's dig a little deeper into that. A content model that we create is a gathering up of information. So let's just say you have hundreds of email coming to you. Let's just say you have a file share or SharePoint or whatever. It gathers in the information from these sources, both digital or paper. It then cleans up the image for better OCR. It then classifies that document - is it a drilling report, accounts payable, finances, what vendor, what do we name the document - the extraction piece. I think the oil and gas industry might have coined the term 'acronym' so there's a lot same things that people call differently. So when we extract information, we normalize and standardize that data. And then we output it to whatever source it needs to go to. So that - what it is - synthetic understanding of data. What it is? A content model. And what it does? What I just described. That is the AI that is in Grooper. We've found it to be a very effective tool, helping people understand what the heck is coming out of these various departments.
Mark: It is literally an amazing to watch this industry change almost beneath my feet. Just ten years ago, you and I would not be talking about this on the microphone. It just didn't exist. And now it's all over the place. When people think of AI, a lot of them think about things like Siri or Cortana, or Amazon Echo where you actually get to talk to the machine, the machine gives you good information, but the thing that Grooper does, that I think is incredible is if you ask Siri "Where is a pair?" Siri may say your pair of shoes is over there. Or you can go buy pears at the grocery store. Siri doesn't really understand. Siri is trying to put words together to try and make sense of it. Grooper would understand exactly what I was talking about. Right? Because it has that AI component fully baked in.
Tony: Exactly. And once again, going back to that insight that you need. I guess a good analogy would be, if were to give you ten thousand dollars and say "Invest in this." The first thing you would do, is you would gather data to make the most informed decision possible. Now how you do that - there's a variety of ways. But to make the most informed decisions, in anything, you need data. You need access to data. And is it quality data? Etc. So to make those informed decision, the most informed, you need that data brought to you and that's what Grooper allows for. That's what the software and the AI allows for. It's very simple. When you break it down into the simplest of terms, once again - it just allows for a better quality business decision. Whether that be drilling, land acquisitions, etc.
Mark: And there's a flip side to all that, Tony. That flip side is, if there's bad data in there, that human may not catch as being bad, Grooper will catch that it's bad and keep you from making bad decisions based on bad data. That happens in this industry all the time.
Tony: Absolutely. It's able to flag certain things that were being missed because other things were more important. As everyone knows, those little mistakes add up. And they add up very very large. Once again, it goes back to the engineers doing engineering work, the people in the field or the ones allocated the AFEs, which vendors to use... it just allows for them to make those high value decisions and high value actions versus lower, less valuable work with respect to manual data entry and the mundane-ness of that type of work. So once again, Grooper allows for a higher value decision making process.
Listen to the podcast in its entirety here: https://oilandgastechpodcast.com/tony-benson/