← Oil and Gas Upstream
Loading summary
Transcript42 lines
- [00:04]
Jag Gattu
Oil and gas production is the union
- [00:06]
Elena Melkert
of natural systems with advanced science and complex engineering.
- [00:10]
Jag Gattu
Smart people across the globe create this
- [00:12]
Elena Melkert
remarkable place we call Upstream.
- [00:15]
Jag Gattu
And each day brings a new challenge.
- [00:18]
Elena Melkert
This is the Oil and Gas Upstream
- [00:20]
Jag Gattu
podcast where we look at how these systems come together and learn from the people who make it happen.
- [00:30]
Elena Melkert
Welcome to Oil and Gas Upstream. I'm Elena Melkert, your host. Some of you know me as the former Director for Oil and Gas Upstream research at the US Department of Energy. I retired from the doe, founded Energia Consulting, and joined the Oil and Gas Global Network as a podcast host. My sponsor is ifs. Through one comprehensive platform, IFS supports the unique needs of the oil and gas industry, resulting in streamlined workflows, automated business processes, and lower operating expenses. For more information, visit ifs.com and now today I'd like to welcome my guest, Jag Gattu. He is the founder and CEO of Uptime AI. So, Jag, thank you so much for joining us today.
- [01:19]
Jag Gattu
Thank you, Elena. Great to be here. Looking forward to the conversation.
- [01:22]
Elena Melkert
Absolutely, absolutely. We were trying to connect at SARAH Week and that didn't work out, but it was pretty hectic and pretty exciting. But I'm glad that we have this more quiet time for me to understand. And for me, one of the themes at SARAH Week was this notion of AI, from the reservoir to the boardroom and everything in between. So one of the things I've been thinking about is that for when we have this podcast that's focused on upstream, that's my subject matter expertise. We have broken it up into these three sections because it's easier for us to process with our brain processing speed, but with AI, we can actually elevate our thinking and think about larger questions. And that's pretty exciting. That's pretty exciting. I'm getting ahead of it a little bit though. Tell us about yourself and then you can tell us about Uptime AI.
- [02:16]
Jag Gattu
Absolutely. I am Jag Gattu. My background, I've spent about close to a decade doing hands on industrial data analytics. I work with pretty much every automotive company around the globe. Then defense companies like U.S. air Force, NASA, Boeing. These were back in the day when people didn't use the term deep learning. Right. And then I also worked with oil and gas and energy companies in mostly doing optimization data analysis, have a bunch of patents in the space as well. So that was my first half of my career. After that I spent another 8, 9 years working at companies like GE, oil and gas, Baker Hughes, leading their product teams both in terms of Smart instrumentation, as well as most the asset monitoring and software solutions to do with machine learning and deep learning at the time. And during that was essentially during that time, I worked with pretty much most of the process super majors, as you can think of in this world today. And one of the common points that came up was, hey, we have all these tools that tell us that there is a problem here. There is essentially, my energy consumption is increasing, my equipment is tripping and all that, showing problems. But then they had a challenge in terms of figuring out how to resolve those problems. And every time we had these conversations, people were always looking towards subject matter experts. They were asking them, hey, what would you do? How would you solve it? And that was essentially the genesis of Uptime AI. We started after my time with GE and Baker Hughes. I left and then I started Uptime AI really with the single focus and a mission to have solutions that can mimic what experts do, to be able to solve problems and guide operations teams to improve their operational profitability. That's how it all started.
- [04:12]
Elena Melkert
Yeah. Yeah. No, that's exciting. I want to paint a cartoon for myself just to make sure that I have captured it all. We didn't always have computers, and this is just strictly oil and gas point of view. So we had a paper spreadsheet and really good erasers, and we used to do these equations and put the answer in a column and then do the next equation and then the next answer, and then you kind of did it. Okay. Then we had access to mainframe terminal. So there was some very primitive spreadsheets, electronic spreadsheets that we could use there. Okay. And then after that, we started being able to model reservoirs and production and just all the aspects of upstream. And we would actually have more need for more space and more speed. We would try to solve those questions. And then after that, help me, what was the next kind of level of capability that we had such that we have here? Because all of a sudden, in my mind, it boomed to big data and then data analytics and machine learning and whatever. Did I miss the step in there or did it go just that quickly?
- [05:20]
Jag Gattu
No, certainly it's fascinating to think back and see how the technology and the operations in general in the oil and gas industry been evolving. Right. The way I look at those step phases are the first phase was, like you mentioned, you know, it's all about automation. We brought in the control systems, the PLCs, and then we started automating a lot of the tasks. Right. And that's how we got the DCS and all of that. And then after that we decided that's when, you know, the mainframes and other computing generally became more available. And we said, how about we monitor or we look for trends or thresholds and in general alerts. Right. And that's when the whole alert started flowing and that became a thing. And then we said, okay, you know what, this is good, this is working. But it would be great if we had more data so that we can process more information and make smarter decisions. And that was the era of the last 10 years, I would say, or so, where we started digitizing a lot of things. Right. We started adding more sensors because sensors became a lot cheaper than what they used to be.
- [06:26]
Elena Melkert
That's right, sensors. That's right, yeah.
- [06:29]
Jag Gattu
And then we also started getting these IT tools. We brought in saps, we brought in CMMS systems, the historians and all the dashboarding, all of that. As an industry, we have done amazingly well in terms of increasing the data, I would say. I was reading some articles and I think in the last few years we've been able to grow the data almost 11x in the industry. And the scary part is actually not that in the next two years, probably we're going to increase it by another 9x.
- [07:01]
Elena Melkert
Oh my God.
- [07:03]
Jag Gattu
So now the data is actually we've solved that problem. But then what we are trying to do now is what do we do with all the data? Back when we started, the data was less and the expertise was more because we had people. Because we started with, remember, we started with people doing all the work and now what's happening is while the data is increasing, people are going away. The other day I saw this statistic where it said the number of people enrolling into petroleum engineering has dropped by almost 70% or so. And these days most of the oil and gas majors are telling us that they can't find the right talent. And even if they find the right talent, they can't retain them for 20 years like they used to in the past. Right. And that's what is essentially driving the need for AI in, specifically in the oil and gas industry now. It's not about getting the data, it's about, about getting intelligence. It's about being able to make decisions. And that's what AI is really bringing to the table.
- [08:05]
Elena Melkert
Okay, excellent, excellent. So how does that happen? How does Elena's oil and gas company, with my 100 wells, perhaps take advantage of all of this and optimize the profit? I mean, bottom line, that's what I care about.
- [08:22]
Jag Gattu
Yeah, absolutely. What used to happen in the past is we had all these sensors, right? I mean, we brought in the sensors and the sensors, let's say you take vibration sensors, right? They measure vibration. And that data used to come into some software. And the software used to tell the engineer, hey, look, your vibration is increasing and you have a problem or you see it in a dashboard. But the next question is, why is it happening? What do I need to do to make sure that it doesn't happen and how can I fix it? How did we fix it in the past, right? All these are questions that people used to resolve. Now when people do or try to figure out the solution, they don't just look at the vibration data. Just like if I have fever and I go to the doctor, doctor is not going to prescribe me a solution based on my body temperature alone, right? They're going to look at a lot of other things. Same thing happens in oil and gas as well. And now with AI, what we are able to do is we can now connect all the dots, not just with the vibration, not just look at the vibration. That's what people are currently doing with existing tools, but really with AI, you can connect the vibration to your process, your flow data, your pressure data, your environmental data. Even, interestingly, I can have the AI look at what is the equipment before this equipment. I can take it even further and further up and understand that, hey, this is vibrating because my motor in front of the pump has a problem. It's all about connecting those dots. And once you connect the dots between the different information, the AI has to reason. And that's why we call our agents as reasoning agents, because that really is the differentiation between AI and just the machine learning and the deep learning. Being able to reason and give you a decision point which essentially is driving towards an optimal goal.
- [10:35]
Marc Lacour
Hey, it's Marc lacour, editor in chief of Oil and Gas Global Network. We want to give you something free for being a listener to one of our podcasts. It's very simple. Go to YouTube, search for Oggn, then comment on any video, send us a screenshot of your comment to marketing g g n.com and we will sen a free Oggn hardtop laptop sticker and a temporary tattoo. Once again, super easy. Just go to YouTube.com, search for oggn, comment on any video, send us a screenshot of your comment to marketing g g n.com we'll put some free swag in your mailbox right away.
- [11:07]
Elena Melkert
Yeah, the doctor example was very good. I mean, I mean, I'm a mom, not A doctor. But when your child's crying, crying comes from many things. Are they hungry? Did they get hurt? You know, whatever. And being able to look and assess at all those things. And of course it organically, we've got these other senses that we can't measure or whatever, but moms know, that's the thing. So you're trans. It sounds to me as if I got it right. You're transferring that MOM capability to assess a situation and come up with the right response in a timely manner, which is quick, if you're talking about children, into the AI agent, where the 8i agent doesn't have this sixth sense, but it has actual world data, which means that the AI agent, if I'm still getting you, can actually come up with more elegant and sophisticated solutions that will be probably more correct than what the subject matter expert might have come up with, having their experience limited to this particular field or that particular field, or quality of fluids or whatever, that kind of thing. So am I on the right track there?
- [12:15]
Jag Gattu
Yeah, 100%. I mean, it's interesting because, you know, a lot of this type of knowledge, like when you say the mom experience or the mom sixth sense. Right. In a lot of ways, humans are very good at building that capability over a period of time because we are really good at understanding the patterns. I've seen very senior engineers go to the field, touch the pump, touch the turbine, or listen to the sound and tell what's going on and what's the problem, what needs to be fixed. Right. It's essentially they're using their senses and their experience to really make those connections. And what we are able to do now with AI is similarly taking the information and being able to reason. We have so many examples like this where when you talk to an engineer, an engineer, generally, if they are a mechanical engineer, they know how to analyze a mechanical system, but they may not know the chemistry of the process. And that's why we have a chemical engineer. And many times the problems, when you look deeply, the problems could be coming in from a different area. And the engineer may not be even able to assess all the complexities and all the interconnections. But with AI, you can. Just because you see vibration doesn't necessarily mean it's a bearing problem. We've seen many cases where you use a DM water system which has. Has some contamination in it, and that could essentially lead to the vibration increase over a period of time. Right. Similar, like you mentioned, if a child is crying, it may not just be that there is an immediate problem or you have to many times to make the optimal decision, you have to understand what's causing that and you have to go to the root cause. And that's essentially what the AI can really make those decisions in a very quick and efficient way.
- [14:10]
Elena Melkert
Absolutely. And so many times we've run into the situation where you have a solution in the field, you go and you fix it and it helped, but it didn't solve and then you have to start over or whatever. Or the solution wasn't exactly a perfect fit, it was a very good fit. So it didn't fix it or whatever. But here when you're working on data and measurements and things like that, so it seems to me that this capability is growing in many directions in the sense of more sensors and more interfaces with the sensor. And what does that whole landscape look like? And gosh, there's no end to it, right?
- [14:50]
Jag Gattu
I mean, it is amazing the possibilities. I have to tell like, you know, at Sarah week we were talking to all these senior leaders in the industry who are trying to solve problems for their businesses and the challenges that they are facing today is decision latency. Businesses want to move faster, they want to improve their profitability. And for oil and gas industry, I mean, you know, our production and generation part of the business represents a majority alliance share in terms of where the money could be made for the oil and gas business. And we're seeing opportunities where you have a million equipment and you have maintenance plans which are hundreds of millions of dollars annually. And how can you optimize the plan for every one of those equipment? Today people do that maybe once every several years. But we have seen cases where companies have not optimized their equipment maintenance plans for five years, ten years as well. But with an agent, you don't have to sit with a person and go through each one manually, which would in two years. Instead the agent could guide the maintenance teams. Hey, are you trying to change this filter? Guess what? You don't need to do that because it's not necessary. And these are the reasons why. And this is how much you can save if you do that. So the guidance can be much more real time. You don't have to wait for years and years to optimize. I'll give you one more example. We have a customer who upstream oil and gas customer and who spends about roughly you or say 200 to $300 million in spares every year.
- [16:36]
Elena Melkert
Wow.
- [16:37]
Jag Gattu
And they said we want to be more efficient and let's optimize this. What they did was they tried to go into the market and see can we find someone or can we get teams internally and do that. And they got a quote which was several tens of millions of dollars estimate and it would take them one and a half to two years to do that. And now after we showed our agents, they are now actually using the agent which could completely transform, which could make it, you know, 100x faster and do it in real time, not just a one time activity. And that's essentially the potential. And there are many, many use cases like this all the way from capital planning. So many different areas we're just barely touching. I would say right now what we're using in AI for oil and gas, it's probably less than 5%. It's miniscul.
- [17:28]
Elena Melkert
Yeah, yeah. So is it fair to say that it's all about the data? The more data you have or the different kinds of data, or your ability to measure something you couldn't measure before, whatever that adds to the value of it and speed. I mean, what would be the limit? What would be the diminishing returns on having like too much data or can you not have too much daily? Just like you can't have too much garlic.
- [17:53]
Jag Gattu
Garlic. The interesting thing is back in the day when we were having tools which would only look at a certain type of data. So for example, let's say you have a software that looks at energy data and tells you whether your energy consumption is increasing or decreasing. Now if you don't have any energy measurements, that software can't really do much. Now on the other hand, think about a human. So let's say the company has found out that we're paying too much bills on energy, our energy bill is increasing or our steam consumption is increasing. You go to the engineer and you tell them, look, our steam consumption is increasing, we're spending too much on energy. Let's figure out, typically the engineer wouldn't say, oh, I don't have the power measurement data, so guess what, I'm not going to do anything. As humans, we're really good at connecting the dots again, right? And there is always enough information because people are looking at, there's a lot of data right now in the operations. Is it enough? There's always opportunities to add more. But I would say there is enough data that we are still not able to process all of that to extract the juice out of it. And that is different about AI because in the past if you try to do a machine learning or just essentially a regular monitoring software or IT type of software, they are just Taking one type of data, they're not really reasoning. But with AI, even if you have imperfect data, you can actually connect the dots. Because when you connect the dots, you can make the vision or you can make the view holistic and thereby you can actually extract insights that were otherwise not possible. And that's why I tell people that when you think about AI, you have to ask the question, can a human do this work? Can a human add more value for this problem? If a human can do it, AI can also add the same level of value, but make it much faster. This was not the case. When you look at just data processing back in the day, does that make sense? The difference between traditional machine learning and how AI is actually changing that and the data requirements around that?
- [20:08]
Elena Melkert
Right, right. I'm always talking about how the human brain is the best computer ever. And what you're saying is that if you have the data, and it doesn't matter how much data you have, the more the better, but doesn't matter, the AI agent can fill in whatever blanks probably more effectively than a human can, because there are limits to what we can do. But there's a creativity element that humans can look at the question in a different way, perhaps.
- [20:35]
Jag Gattu
Exactly.
- [20:36]
Elena Melkert
And that kind of frees you to do the fun stuff as opposed to the stuff that you can do, because that's all you can do. Because that's the limits of what we know.
- [20:44]
Jag Gattu
Oh yeah. I mean, the interesting thing in oil and gas is that, number one, the industry has been doing a great job in collecting data. I mean, you talk to any oil and gas company, they have tons and tons of data. It's not necessarily the case in all industries. We do work with predominantly process industries and process industries like oil and gas or chemicals or cement or power generation, utilities. Right. Even metals and mining. They're all generally highly automated facilities. And generally speaking, they are more streamlined and they have good amounts of data. The other thing to also think about is within oil and gas, Right. As I was talking initially, unlike other industries, the amount of expertise leaving the industry is actually scary in some ways.
- [21:41]
Elena Melkert
You mean subject matter experts?
- [21:42]
Jag Gattu
Subject matter experts, right, gotcha. Like you mentioned, with tools like these, what they're doing is they're actually acting as an assistant to the real, the human experts. Right. Instead of the human expert saying, hey, I'm going to do one root cause analysis or I'm going to do one hazops every two months or every 10 days, right now they can say, you know what? My agent is going to do the analysis for me, and it's going to present a result and a case. I'm just going to review that in an hour and tell the agent, did you do it well or did you do it wrong? And the agent will keep learning. And you're basically training an agent to reduce all the manual work so that you can make the decisions and you can make the judgment.
- [22:32]
Elena Melkert
Yeah, yeah. Oh, this is fascinating because when I first started recording this series of podcasts at Sarah Week and I would talk about AI, I was very, very uncomfortable with it based on the early failures of not just the learning curve of early modeling, whatever I said, you're making an assumption that the core did not reveal when we looked at it, you know, something like that. But now I'm understanding that there's a deterministic element and then that there is an agent. And that makes me feel much, much better. So now I'm kind of going, well, what can the AI not do for me in the oil gas business? Help me with personnel issues, can they help me with hiring and finding the right people and comments that I want to make to the government with respect to policy or whatever? It really is about framing the right question and then teeing up all of the factors that have an influence on that question, because that's what humans do, and then come up with the best decision. So it's a lot easier, if I can say that, to play these what if games with an agent.
- [23:40]
Jag Gattu
Yeah. See, today, what we're seeing around in the society or in the market from AI use cases, we see a lot of chatbots, right? ChatGPT and Claude and Gemini and many more. And they have made AI more prevalent everywhere. Right. Many people are now using it on a daily basis. However, that may not necessarily be the most highest value use case for each industry. Just because there is a conversational assistant, which adds so much value for a call center business or a customer support business, doesn't necessarily mean that's the same technology that's going to add the highest value to oil and gas. Every industry will evolve into the point of what is my biggest problem and where can AI bring value to me today? What's happening is most of the companies, they know that AI has tremendous value and they also know that if they don't embrace it, they will get behind and it's not going to be an easy one to catch up to. All the thought leaders in the space who we're working with, they're all looking five, ten years ahead of time, but they're also looking at what are the biggest value drivers for my business, for my operations, where AI could add value. And what we should also think about are what kind of skills do the agents need to apply to solve those problems. Remember when we talked about what a human is doing or what a subject matter expert is doing? The agent should be able to do that. But a human or an expert is able to do several different things and some of them are very complex things. They can look at a P and ID and they can say, what does this mean? An agent should also be able to do that. What is happening is for each industry, their agents are going to be having or bringing those specialized skills as part of that agency. And that's how they're going to be able to build to solve the problems that are most important for them. Oil and gas for power. It's not going to be the same agent, the same skills that are going to be useful in oil and gas versus a retail store. They're going to be different.
- [26:06]
Elena Melkert
Very good. Oh my gosh. I could talk to you forever. We are almost out of time. Is there anything else you wanted to share that we didn't get to? This is fascinating. I just love it. I'm getting so much more comfortable, so more creative in the kinds of questions I want to ask in terms of, of an AI agent helping me.
- [26:23]
Jag Gattu
What I would say is, in the industry, when people are thinking about how do we apply AI in this space, definitely start with what is my biggest value driver? Because if you start with things which are just because some other industry or someone else is doing that, they're good for showing that you're using AI. But ultimately the value is not going to come. So pick the big problems, then work with solutions or pick up solutions that are specifically bringing the right skills to solve those problems. Using AI, that is important to have that experience and domain knowledge because oil and gas is specialized. Make sure that is happening and then certainly have the support from the top leadership to be able to encourage people. And we're seeing as that encouragement comes, people will pick them up because that's what is going to really make their lives easier and save a lot of money for the business. I'll stop with that.
- [27:22]
Elena Melkert
Yeah, that's fabulous. That's fabulous. Well, Jag Gattu, founder and CEO of Uptime AI, thank you so much for joining us today.
- [27:32]
Jag Gattu
Thank you, Elena. It's a pleasure.
- [27:34]
Elena Melkert
Thank you. And thank you everyone for listening. This is Elena Melkert, your host for Oil and Gas Upstream. More next time.
- [27:42]
Marc Lacour
Thanks for listening to Oggn the world's largest and most listened to podcast network for the oil and energy industry. If you like this show, leave us a review and then go to oggn.com to learn about all our other shows. And don't forget to sign up for our weekly newsletter. This show has been a production of the oil and gas.