← Oil and Gas Upstream
Oil and Gas Upstream
From Field Signals to Financial Decisions: Interview Sonali Singh, SVP Quorum Software | Ep 341
00:29:13
Loading summary
Transcript105 lines
- [00:04]
A
Oil and gas production is the union of natural systems with advanced science and complex engineering. Smart people across the globe create this remarkable place we call Upstream. And each day brings a new challenge. This is the Oil and Gas Upstream podcast where we look at how these systems come together and learn from the people who make it happen.
- [00:30]
B
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. And I'd like to make a shout out to my sponsor, ifs who maximizes efficiency, reduces costs and enhance asset reliability through the use of their software designed to support the unique needs of the oil and gas industry. For more information, go to ifs.com so we're recording here in person at Sarah Week and Cyrilk has been so exciting. We're day four and just so many smart people, creative people, all about energy, all aspects. It's really been, you know, really exciting. And if you can ever come, please try. And the Agora Innovations is also just so fabulous. So this is, I don't know, I've been here a few years now, but I'm so happy to be able to do it and meeting so many wonderful people like my guest here. My guest today is Sonali Singh. She's the senior Vice President of Product for Quorum Software. Well, thank you, Sonali. Thank you for joining us today.
- [01:49]
C
Thank you, Elena. It's great to be here.
- [01:51]
B
Excellent. Tell us a little bit about yourself and what, what you do and why we should. Love you.
- [01:57]
C
Yeah, So I am a chemical engineer, but pretty soon after I graduated I jumped right into software and software development. So I like to think of my 25 plus years in the industry journey in kind of three phases. So my first phase was I was really a hands on builder. So I built some software for engineering, procurement and construction companies.
- [02:24]
B
So when you say build, you're meaning coding.
- [02:27]
C
Coding.
- [02:27]
B
Okay, good. Yeah, I just want to make sure I get, I'm with it because that's not my area.
- [02:31]
C
Yes, yes. So designing and coding software.
- [02:35]
B
Okay, yeah, let's build.
- [02:36]
C
Yeah, yeah. And you know, engineers spend a lot of time in software tools to try and design and build whether it's chemical plants or refineries or LNG facilities, et cetera. So I spent that first phase of my career really watching them and seeing all the challenges they had and then tried to solve them with technology and building. Software products that. So how is an energy always an energy? Always an energy period. Yeah. And then the second phase, I was a co founder of a product and then we launched a startup. So my second phase was sort of startup entrepreneur.
- [03:18]
B
Wow.
- [03:19]
C
And I was CEO of that company for seven years and that was really exciting. So we came up with a AI driven product to automatically build plants in 3D so it could be again, LNG facility, chemicals, upstream, offshore platforms. So the product would automatically build it in 3D and leverage a lot of AI to automatically kind of hook everything up, place the equipment, auto wrap, pipe, things like that.
- [03:51]
B
Yeah, yeah. Well, so we are still early stages for AI, Right. As everyday users, I guess is a way to put it. And AI gets it wrong. So how did you bridge that? I mean, how'd you fix that, make sure that it was right?
- [04:06]
C
Yeah, I've been, you know, half of SARAH week. Feels like an AI conference. There's so many AI sessions here. And then one of the things in a panel yesterday, I said, you know, I'm actually traditional AI. Like I've been working with AI back in the late 90s. And so that was before ChatGPT and Gen AI. So there's been these traditional AI tools. They're called like expert systems or neural networks. They're different because they actually solve to like a final problem. Like now if you use ChatGPT, you could get a different answer depending on how you prompt it. Right, right. So we call that like non deterministic. And my. When I built this first product, it was very deterministic. Like there was really only one answer after it considered all of the engineering principles.
- [04:55]
B
Okay, now see, that is something I didn't realize.
- [04:59]
C
Yeah.
- [04:59]
B
All one category. Because I know, I feel as though we've been using AI and oil and gas forever. We have with the data though, and we just have so much data, there's no other way to do it.
- [05:08]
C
Correct.
- [05:09]
B
But you did have to have an answer and had to answer all of the science and physics rules.
- [05:13]
C
Right? Exactly.
- [05:14]
B
And then this ChatGPT being AI, I thought that was. Oh, good. I feel so much so non deterministic that I don't like.
- [05:23]
C
Well, as an engineer you don't want that.
- [05:25]
B
Right.
- [05:26]
C
You want those math formulas to come up with a single answer and have a good degree of accuracy.
- [05:33]
B
I want it to be about the assumptions I put into my model, not the things that I don't know. Exactly the things I do know. Okay, very good. Well, see, I've learned something already. So.
- [05:41]
C
Yeah.
- [05:42]
B
Okay, so. So the Transfer of from or the evolution from deterministic to non deterministic. So pick it up there again for me.
- [05:51]
C
Okay, well, and we're going to come back to that because I'm going to talk about that more with Quorum Software.
- [05:56]
B
Okay, great. Already?
- [05:59]
C
Yeah. But just in terms of my journey. So then I ran that company for about seven years and then I sold it to a company called Aspen Technology or Emerson. And so then I went to Aspen Technology and I had a similar role there as vice president of product where I was managing all of the process simulation tools for upstream and downstream. So really again focused on innovation and growth in software and helping process engineers and chemical engineers with what they do day to day. And then while I was doing that, Quorum found me with this role here at Quorum. And I'm really excited by what Quorum Software is doing in terms of AI and data moving forward. So that's what brought me here about four or five months ago.
- [06:49]
B
Oh my gosh. So you've been with Quorum about four or five months?
- [06:52]
C
Correct.
- [06:52]
B
Okay, so when you say you're senior vice president of product, that product means.
- [06:57]
C
Yes. So Quorum has many different software products. And the Quorum software really sits at the center of like where upstream operators can plan, produce, move, as you said, and account for any kind of hydrocarbon. So I manage, or I oversee, if you will, the segment of products that really go from field operations, so like SCADA systems, well, production systems, all the way to the back office for accounting, land management, and then into the main midstream suite as well, where we're looking at pipelines, transportation, scheduling, and then we also have measurement solutions that help with that. That handover from upstream to midstream. Yeah, so that's, that's the segment that I oversee. And one of the exciting parts of this role is now taking all these different products that we have and really putting this data and AI capability to bring operators even more value than what they've been getting.
- [08:07]
B
Yeah, yeah, no, that's. That is so important. Every time I work with groups and of course it's, you know, still early days for AI, but especially people who are hesitant or afraid. I guess I'm just afraid of it because I don't understand a lot about it yet and it's not my favorite thing either. So I'm a very hands on person. But the point is that every time I work with groups, things go well. When everybody is kind of has the same vocabulary, the same background, whatever. It's when you interface with another group has its own tribe. That interface is a weakness.
- [08:41]
C
Yes.
- [08:42]
B
And so I find that doesn't matter whether it's, you know, some church group or daughter's school or whatever, or work. That cultural difference between the different areas of expertise is always a challenge. And you always have to be. You have to work extra hard there. Everybody does. But the point is you have to. And what I'm finding at this conference with the AI being a key theme on every aspect of it, is that it kind of smooths over some of that. Can you expand on that a little bit for us? Because you're nodding your head. So I'm onto something here.
- [09:16]
C
Yeah, no, no, for sure. I think I've been in several sessions where, you know, especially here at SARAH Week, where they've started with AI, and then everyone said, well, hold up, let's. Let's define what AI we're talking about. Right. And I think today most of the conversations are around what we'll call generative AI or like the large language models. ChatGPT. Right. The things now people have started to try to use. But then there's like, I'll say more, the traditional AI, which are things like I mentioned, like neural networks, expert systems, technology, deterministic.
- [09:52]
B
You gave me some really good understanding in five minutes, so thank you.
- [09:56]
C
Yeah, those were deterministic. This geni is mostly non deterministic. And then now we've got this new emergence of like agentic AI.
- [10:07]
B
Oh my gosh. Yeah, agentic.
- [10:09]
C
A G, E, N. Like an agent.
- [10:13]
B
Ah, agent, I understand.
- [10:14]
C
Okay, so agents like you think chatbots, right? Yes. Like the little chatbots you communicate with on many websites and things. But agentic workflows now means there's multiple little agents working together to help you maybe make a decision or to automate something that you've been doing.
- [10:36]
B
Yeah, yeah, yeah.
- [10:38]
C
So maybe we can we explore an example. Oh yeah, this is an example where exploring with Quorum software. Okay. Because right now we're kind of that unified end to end workspace, right? From the field to finance. But let's say I'm a VP of operations at an upstream operator and my SCADA system tells me that five of my wells are down. Okay. I've only got one or two crews. How do I decide where to send that crew? Which well to send that crew to? Right? So today for that vp, they've got to talk to multiple people. They've got to figure out each well's production, how long it's been down. Right. And you would think maybe, well, the highest producing well Gets priority. But then the other thing that comes into play is profitability of that well. Which well is most profitable? Do we know why the well is down and can we even repair it with the crew that we have? Right, for that day? And then you have to also look at things like what are your lease terms on that well and what are penalties you may need to pay based on how long you've been down. So this is a decision, right? This VP makes by talking to multiple people, collecting multiple data. So an agentic workflow that we are creating at Quorum will go into all the systems and give him that data or her give her that data.
- [12:07]
B
There you go.
- [12:08]
C
And say of your five wells, like wells one and two are the highest production, but well three is your highest profitability. And look at, you know, what's the price of oil per barrel that day or that week and then kind of help calculate what profitability impact there is. And then they may look into like, well, what's the cause? Is it a compressor? Do I know why the well is down or not? Can that crew fix it? And then really just come back to her with, here's the well, you should prioritize for these reasons. And then it's up to her to make the decision, like kind of agree with it. Or maybe she'll go back to the agents and say, well, what if I, you know, what if we can manage the lease terms? Or what if I have a crew that can repair that compressor? How does that change it? Right? So now something that would have taken her probably like the full day to figure out and you're not comfortable maybe that you've made the best decision now she's going to get that information within a minute, right, and be able to kind of go through that what if scenario and then just make better decisions.
- [13:22]
B
Now that's the AI I want.
- [13:24]
C
Yes, exactly. It's prescriptive. It's prescriptive. It's. It's prescriptive and it's like a deterministic answer.
- [13:32]
B
Yes, yes.
- [13:33]
C
And the way we can do that is you take those like ChatGPT models and then you kind of put some like guardrails on it, right? Like you say, here's physics, here's physical constraints, right? Like there's only certain values that pressure, temperature, what the volume calculation. Here's the, you know, here you've got to understand that we can also put in guardrails, like around regulations and regulatory frameworks and like economics, right? Like, what is the price per barrel? What are your terms for how you're selling that or procuring it. So you put those guardrails in there. Now that AI can really get to a single answer. And you also know where that answer came from.
- [14:18]
B
Yeah, that is huge on a couple of levels. Number one, I think that the human brain is still smarter than any computer.
- [14:25]
C
You put it.
- [14:25]
B
Not fast, yes, but smarter.
- [14:28]
C
Right.
- [14:28]
B
And so I want to be free of some of these little questions, these little calculations. Physics, math, I want to be free from that and really think about, you know, the intersection of one assumption with another assumption and whatever, and the limits of what I actually know.
- [14:44]
C
Right.
- [14:45]
B
Which informs what. Or I should be putting my money to find out and learn because it has such an impact perhaps on my models and those kinds of things that I feel. That's freedom. That is what makes me really excited
- [14:57]
C
about that, because we want our engineers to be thinking critically. Right. Like, that's the value that they bring. And then this sort of releases them from kind of the mundane data entry or data calculation, and they can really get the information and then go through that process to make better decisions.
- [15:19]
B
Well, time is the only thing that's limited. And so we can't make more of that. We can make more money, but we can't make more time. And so spending time on the less important questions.
- [15:30]
C
Right.
- [15:31]
B
When they can be delegated to an agent, perhaps.
- [15:33]
C
Yes.
- [15:34]
B
See, I'm getting good already.
- [15:35]
C
There you go. Got my own agent. Yeah, yeah.
- [15:39]
B
The other thing is that I came into the business, you know, many, many years ago, over 40 years, and we didn't have personal computers.
- [15:47]
C
Okay.
- [15:47]
B
We didn't have this, you know, I mean, we had. There were computers, obviously, but it was a shared terminal with a big, you know, I don't even know what to call it. Computer. And so when modeling first came into play, you know, reservoir modeling, there was digitized maps or whatever, and translate the various logs, all the information you knew about the reservoir, into this model that was limited in size. I mean, the cells had to be really big still, whatever. And every time we would get an answer, it had, to me, it had to tie back to the core. We had to touch something that we knew because just organically, we could process a lot about what we were looking at in the core, ignoring that the core was just a small sample of a huge space. But the point is we had it and we could relate to it. And choring is still at the heart of what we do to validate what we think we know.
- [16:45]
C
And those technology has changed so much, right. Over the years since Then. And I think there continues to be, like. I think there's a lot of opportunity with seismic interpretation with AI that can help those engineers.
- [17:01]
B
Yeah, yeah. Okay. Now you're making me feel really good about AI as long as it's deterministic. Deterministic. And it's because of the agentic. Oh, my gosh. Two new words.
- [17:14]
C
Okay. Yes.
- [17:15]
B
Deterministic and agentic. Okay, so where are we? I mean, tell me about the current state of the art with agentic and what our little feature is here. Obviously, with Quorum definitely leading the way there.
- [17:29]
C
Yeah. So the other thing that helps AI agents of any kind be more accurate is the data that's underneath. Right. So that's the value that we have at Quorum is because our software is the system of record for upstream operators and for midstream operators. So that's where all the land docks are. That's where all the revenue accounting is done. Right. All the well data is there. So now what we can do is we can take all of that data, and there's like. What you want to do is give it context, give it meaning. So the more meaning that you can give to the data, the better the AI is. Right. Because you're really helping those AI agents now understand that those five wells we talked about, it's the same well that's in this lease. That's, you know, because they're not all called the same thing. And then it's the same well that's in this accounting system. It's the same well that, you know, so. And a SCADA system might be reporting on. So I think what we don't realize today is we have a lot of great systems of records, and they're super important for what we do. But, like, if you think of it as a system, they don't necessarily all understand each other. So what we're providing is context. And we're providing context across all of that data. And that context that then makes those AI agents much more accurate and then helps us really build out this sort of network of agents or an agentic workflow that will understand in that same example, we talked about that for these five wells, here's that same well in the land Doc. Here's the same well in the accounting system. So it can really kind of correlate and understand that whole network better.
- [19:23]
B
Yeah, yeah. So a robot still can't go out into the field and weld. You can take one out there and park it, and it can do it. But there is a human factor in the sense that There are a lot of jobs that are never going to go away because you just need to have a human in charge of the agent. Whether it's software or whether it's a physical robot or whatever, you're always going to have to sew. But safety is something that can be enhanced. You don't have to send a human into a tank.
- [19:57]
C
Right. There's a lot of environments that are hazardous. And if you could have the human maybe guiding that robot or guiding that, it could even be a drone or something that they're guiding in there that can do some of the work, inspect, look at maintenance opportunities, whatever might be needed. I think there is a lot of mistrust of AI out there, maybe some fear. Right. But I guess being somebody who's traditional AI and I've been working with it for so many decades now, I would say it's an opportunity. Right. I think we all need to look at it as an opportunity to how can it maybe help me in my personal life, but how can it help me in my work life and help me develop and grow professionally as well?
- [20:52]
B
Right. Well, and I think that is a danger for some people to say, oh, let the agent handle it. But no, we have to have me even more be more attentive to some critical analysis of something and keep that agent growing, I guess is a way to put it, because they'll learn from those things as well. But there's probably never going to be a time where human will not need to be a guide and those kinds of things. And of course, we will learn from the agents as well, and our critical analyses will become more acute. We will look at what are the better questions to be asking. And thanks to checking.
- [21:31]
C
Yeah, I mean, it's a silly example, I'll say. But a lot of people now use AI to remodel their bathrooms or remodel their kitchens and things like that. Right. So I think what happens, and you could associate that to drilling a well and doing completions and all of that. Right. Is it just comes up with maybe some crazy designs, but maybe some designs you never thought of.
- [21:57]
B
You never thought of. Exactly.
- [21:58]
C
And just because it's. And it's not that it thought of it, it's just that it's running so many cases, it's probabilistic. It's just running every probable case. So then maybe out of that, you're like, oh, there's a case I never thought of. There's a recipe for, well, completion I didn't think of that would, you know, give this benefit So I think those are the opportunities that we have to kind of, you know, stay open to.
- [22:21]
B
Absolutely. So when are we going to get to the point where I can have my own agent, pool of agents that are specialists and, you know, keeping my closet if. Because I can never find anything in my closet.
- [22:34]
C
Well, the good news for upstream and midstream operators who are Quorum Software customers is they're going to start seeing that this year. Right? Yeah. So we are very actively working through this in this year and through next year. And so we're starting to build out that framework, that example that I just shared on critical, well, analysis. That's something we're working through this year and we're going to get out into the market later this year. So we're really excited to get it into the hands of our customers. And one of the things we're focused on also is really collaborating and co innovating with our customers. So they're also coming to us to say, hey, you know, I've got operators out there maybe that are looking at, you know, ethane recovery or methane. And we want to understand the impact of that. Right. On the economic side, can you help us with an agent that'll kind of tie together that data and that analysis? And so it's. This is all, I think, moving very quickly. And you know, at Quorum, we're really focused on building this out. We also have coming out even sooner actually, is on the planning space side. So, you know, traditionally corporate planning and asset planning planning has been kind of siloed. Like there's boardroom talks about where should capital go based on what they own. And then there's like, at an asset level, they're looking at what they should build out. Now there's going to be this agentic workflow that helps across those two groups. So that from the well to the boardroom, they're looking at the same data and making decisions on like, where do we want to invest in as we go forward?
- [24:24]
B
Right, right. And predictive analytics for asset maintenance.
- [24:29]
C
Yes, exactly. So predictive analytics actually to some degree has been around for a little bit of time.
- [24:35]
B
Yeah.
- [24:36]
C
And then so what?
- [24:36]
B
We want it to be really good.
- [24:38]
C
Exactly, exactly. So that's the opportunity. Right. Like, it'll get even more accurate, it'll get even better. And then from prediction, you want to get to prescription. Right, Right. Like, okay, you're predicting it now tell me what to do do. Help me make that decision so that, you know, we're making the best decision to really prevent that from happening. Right.
- [25:01]
B
Right. And I don't know if it's further upstream or downstream. What am I doing? What can I do better to not even go into that arena. Yeah, things like that.
- [25:11]
C
Exactly. Now, I think the challenging part, and I've heard some of this here at Sarah Week, is we also have operators driving towards the AUTONOMY autonomous operations. Right. And so I think. And I'm still learning this as well, I think there's still a human in the loop there. Right. I think it's just talking about, like, make the decision, you know, go to the next step. If it's requiring maintenance or repair, like, that kind of automatically happens. But there will be, like, control points at touch points, you know, where at any point the real operator. Right. The human operator can kind of ensure that things are working correctly.
- [25:55]
B
Yeah, yeah. I think it's. You have to. What they do is they set up the guardrails, I guess, is a way to put it. And if they hit the guardrails, then the human has to intervene. But if you don't hit any guardrails, then that's the benefit, right?
- [26:10]
C
Exactly. Exactly. Yeah. I don't know. Are you familiar with, like, the Waymo autonomous cars?
- [26:17]
B
Yes and no. Autonomous cars, yes. I'm not comfortable with that.
- [26:21]
C
Yes.
- [26:21]
B
I don't know about Waymo specifically, so.
- [26:23]
C
Yeah, you tell me. So Waymo is. Is one of the autonomous car companies out there, and I'm from San Francisco, and they're prolific in San Francisco. So you can't go without taking one at one point. So it's proof, you know, that autonomous vehicles will work in a city in a contained environment, and people take it all the time. But there are times, cases that come up where, like, a Waymo car will be at an intersection, and if it's a. Like, if the lights are out and it's a very crowded intersection, it has a fail safe in there that says, nope, I need a human to help me safely get through this situation.
- [27:04]
B
Oh, it does call for a human. What, a smart.
- [27:08]
C
Exactly. Well, so. And I think that's. That's my example when people get scared about autonomous operations, you know, And I think that's a good example. Like, that's an autonomous car, but there's still fail safes in there. Where. And it can be a remote human. Like, it'll call back to a human in a different location and say, this is my situation. You know, you've got the cameras. Help me out of this. Out of this situation. And the camera, the person, I'm sorry, will guide the car through that situation. So.
- [27:39]
B
Oh, that's great. That's great. Well, we are almost out of time. Did I get to ask you what you didn't get to say or.
- [27:47]
C
Yeah, I think, you know, I just want to say that at Quorum, we're also adopting AI internally and really trying to make the most use of it as an organization and as a company. And then of course, more importantly, to bring value to upstream and midstream operators. So it's an exciting journey and we're kind of excited to be at the forefront of it.
- [28:11]
B
Yeah. Yeah. Well, Sonali Singh, senior Vice President of Product at Quorum Software, thank you so much for joining us today.
- [28:21]
C
Thank you so much, Elena. And great to be here.
- [28:23]
B
Absolutely great to be here, having a great time, as you are too. And you're definitely making an important contribution to the conference because AI is everywhere. So I'm so happy that you're here and that you're here with us on today. And thank you everyone for listening. This is Elena Melkert, your host for Oil and Gas Upstream. More next time.
- [28:44]
D
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. Don't forget to sign up for our weekly newsletter. This show has been a production of the Oil and Gas Global Network.