Building AI-Ready Enterprises: From Data Silos to Unified Platforms

Summary

Leaders from BP and Oxy, moderated by Infosys, discuss how large energy companies break data silos and build AI-ready lakehouse platforms to move AI from pilot to production across complex, federated environments with thousands of users.
Patrick from Oxy, the company's chief AI and digital officer, explains how Occidental Petroleum builds programs to scale AI adoption across an upstream oil and gas company, describing the experiential learnings from moving beyond proof-of-concept projects to systems with measurable revenue impact.
Practical strategies discussed include lakehouse architecture for unifying diverse data sources, design thinking for identifying high-value use cases, and the critical importance of leadership visibility and role-modeling in driving organizational AI adoption at scale.

Building AI-Ready Enterprises: From Data Silos to Unified Platforms

Enterprise modernization requires more than new technology. It demands unified data infrastructure, governance frameworks, and organizational alignment to enable safe AI deployment at scale. Hear from leaders at BP and Oxy on how they're breaking data silos and building AI-ready platforms across complex, federated environments.
Learn practical approaches to data foundation building, enterprise governance, and managing change across thousands of users. Explore lakehouse architecture for unified analytics, strategies for handling diverse data sources, and the critical role of design thinking and leadership visibility in scaling AI adoption. This panel covers modernization catalysts, migration patterns, automation, and how companies measure success with measurable business impact.
🤝

Chapters

FAQs

How are oil and gas companies like BP and Oxy using AI on Databricks?

BP and Oxy are building unified data foundations on lakehouse architectures to scale AI beyond isolated pilot projects and into production systems that generate measurable business and revenue impact. Both companies operate in complex, federated environments where data is distributed across many legacy systems, making the data foundation work the necessary prerequisite for any AI initiative.

What is the role of Oxy's chief AI officer in scaling AI adoption?

Patrick, Oxy's chief AI and digital officer, leads company-wide programs for AI adoption at Occidental Petroleum, covering both building out AI systems and creating the organizational conditions needed to scale them across the enterprise. His scope encompasses everything related to AI within an upstream oil and gas company whose core operations involve exploration and oil production.

What challenges do energy companies face when trying to scale AI initiatives?

Energy companies face the fundamental challenge that AI pilots proliferate but scaling them requires unified data infrastructure, consistent governance, and organizational alignment that most companies are still building. Panelists from BP and Oxy describe how the experiential learning from moving proof-of-concept projects to production-grade AI systems reveals gaps in data quality, governance, and change management that must be addressed first.

Why is design thinking important for enterprise AI adoption in oil and gas?

This video's panel covers how design thinking helps leaders identify the highest-value AI use cases and build solutions that change how decisions are actually made rather than just demonstrating technical capability. Leadership visibility—where executives visibly use and champion AI tools—is described as equally critical for driving adoption across large organizations with diverse teams and thousands of users.

Full transcript

[00:08] Good afternoon everyone. Um my name is Shri Ram. I'm from Inhysus. Uh I managed the energy business for Infosys. Um and I also manage AI across multiple sectors of Infosys. Today we're going to be talking about um very
[00:24] fundamental problem that we are seeing. We're seeing so many AI pilots in the industry. We're seeing so many data plus AI pilots across the industry across every industry. One of the biggest challenges that we have been having is how are we scaling these initiatives of
[00:42] modernization as it applies to data as it applies to AI how industry leaders are accelerating adoption of uh AI ready lakehouse that's the topic we're going to speak about today uh and we are going to take a very
[00:57] industry operator centric view of things today you're going to see two est clients of us joining us in that panel. Uh one is BP, other is Oxy. Both of them are tremendous technology
[01:13] operators. They are at the forefront of technology innovation as it applies to AI, as it applies to data. Um and my co-panelists are going to be talking about their experiential learnings of what they have done with data, what they
[01:29] have done with AI in the industry. Um and we'll have five 10 minutes of question answers at the end of the session. Uh it's a 30 minute panel 30 35 minutes panel. Um so we'll have some uh four or five questions uh which will go around the panelists and then we'll have
[01:47] uh time for the audience to ask questions. So please hold your thought in terms of the questions that you have. Um with that uh let me invite Patrick Banger from Oxy. He's the chief AI and digital officer at Oxy.
[02:02] Patrick, please. And from BP, we have Julie Inguan joining us. Thank you, Julie. Thank you.
[02:17] Thanks, Patrick. And I'm going to put their faces, all our faces, so that you can remember our names as we are talking through this today. So I get the center stage. Wonderful. Uh so Patrick just for the
[02:34] audience um if you could give a brief introduction of yourself, what you do uh what you drive in Oxy uh your role please. Yeah. Hi, good to uh be here. Thanks for having me and thank you for spending some uh 40 minutes with us here. Um my
[02:50] name is Patrick. I work for accidental petroleum, short oxy, which is an upstream oil and gas company. That means we drill holes in the ground and bring the oil to the surface. Um, everything else is taken care of by by others. Um, and within that company, um, I deal with
[03:06] everything to do with AI. So, I'm I'm the chief AI officer. We build out the programs to both create our own AI models to create the software that deploys them to do the change management that ultimately sees the value delivery um of that and we are primarily looking
[03:23] to create additional revenue. So our source of value is principally more barrels um not so much uh costsaving or workforce efficiency because as a as a physical company that where most of the costs and the productivity are in machines and assets that that's where
[03:39] we're headed. Thanks Patrick. Uh Julie over to you. Um my name is Julian W. I am with uh BP and we are a global energy company. Uh which what does that mean? It mean that
[03:55] we also drill the hole. We got to dig up the oil but then after that uh the oil and gas and after that it's go through the refine uh the refining process and it's also go through uh the midstream and then we also have trading and retail
[04:11] business. Uh for us uh AI and now data is everything right. So for us the the um it it's not about if we miss something it's not about miss a quarter but it's more than you know uh an
[04:27] incident that we have to take care of. It's about um uh it's about the the miscalculation of the allocation of our resources. It's so AI now is on top of our agenda. Great. So with that I'll uh begin with
[04:44] my first question. Uh this is for both the panelists. Um so I'll begin with you Julie for this one. Right. Um oil and gas as a sector is typically thought of as a very legacy industrial sector. Right. Things change much slowly
[05:01] compared to a digitally native sector like a healthcare or financial services. So what is making in your opinion modernization urgent in the oil and gas industry? So if you're looking at the oil and gas
[05:18] industry right for us the system that we have right the data that we have is is is always silo if you're looking at from the the moment that we try to find the oil on the ground or in the ocean right
[05:34] all the way until we can produce that oil and and put that into the end users everything is very silo okay so you have the drill ing you have the the the uh the production you have the reservoir you have the maintenance management and
[05:51] later on the logistic all that stuff is is we looking at quite silo and for that the model modernization of the organization become very hard how we put everything together right because AI is not about one silo thing AI is about
[06:09] your data in your organization looking across organization Yeah. Okay. Thank you, Patrick. Your thoughts? Um, I I would push back a little bit on the premise of the question if I maybe a contrarian. Um, you got to know what the
[06:25] core of your business is, right? So, as an upstream oil and gas company, the core of your business is to drill wells and bring oil to the surface. We're not a software company. We're not a digital company. Um, so if you come to an oil company with a new turbine or a compressor or a pump or a new steel for
[06:41] better pipelines, it'll be adopted almost immediately. Um, if you come with new software, that's not the core of the business. Um, and so that will be a little bit slowed down. And yes, absolutely, our data is siloed. Um, for
[06:56] good reason because we slowly adopted a software in the 1990s and we will diligently operate that till today. Um and therefore the silos um have have appeared um and now need to be solved. Okay. A follow up to that. Uh Patrick um
[07:15] is is there then a need for the industry. Um yeah you're focusing on barrels cost per barrel. uh you're not focused on the below the line things like efficiency, productivity because
[07:31] those are very small percentage for a company like Oxy or BP right so if that is going to be the case is AI going to generate more above the line opportunities for you what I mean by above the line is revenue generation opportunities new business development
[07:48] opportunities reducing the cost per barrel is going to is AI going to generate will that modernization has to happen for In my opinion, yes. Um, so a AI I think now is in a position where it can be a revenue generator or for Oxy it already
[08:05] is. Um, so it actually produces incrementally more oil than without AI. And in that sense, it reduces the cost per barrel because you're dividing more barrels by the same cost. And so it, you know, it looks like it's reducing it,
[08:22] but we're not a cost reduction agent, right? We're a barrel creation agent. So it's a very different lens on the same thing. But to make AI operational at a company like this, yes, you have to break the silos. Um, you have to modernize your data infrastructure,
[08:38] basically put it all in one place, um, some sort of cloud managed by some sort of a one pane of glass to look at that. And that could be a certain company starting with a D I think it is. Um so to to be able to have that visibility um
[08:55] that that is a necessary precondition but as I'm sure we'll get to it is not enough yet right you then have to actually do the AI bit and then the most important is the change management to get your field personnel to do it. Yeah I think um Patrick when we were discussing with him offline he was
[09:11] sharing a very interesting bit of information. We have always seen our customers talk about AI in the context of how it improved productivity, efficiency, uh operating margin, things like that, below the line as they call it. Patrick
[09:26] was mentioning a figure uh I'll let him take that thunder on that one on what he has been able to drive above the line for Oxy. Please go ahead. Yeah, I I don't want to mention actual numbers in public. I'm I'm sorry. Um but it's it's it's it's a significant amount
[09:42] of revenue that AI is able to to generate for us. Yeah, it's so there are actual use cases that are being implemented in AI in the data space which has driven um across industries. We have seen that in
[09:58] financial services. We have seen in enterprise clients like BP, Oxy when we work with other big clients like Exon Mobile. the because of the quantum of the revenue that is being spoken about here. These are all hundreds of billions of dollars companies, right? So even uh
[10:17] a 1% improvement is a billion dollar addition, right? Imagine the quantum of that. So that's the kind of scale that Patrick was talking about. So it's extremely important to apply AI in the context of above the line opportunities than just focusing on workforce and
[10:33] productivity. Uh back to you um Julie in your context in BP's context what does modernization really require to practice it at enterprise scales right at the enterprise case right for us for
[10:49] BP specifically um back to the silo right we we talking about the AI and for AI to be good your data has to be the foundation of everything if you have your data that
[11:05] not together right your data is not quality data your data that don't have the context and then the outcome of your AI or the result of the AI will not there so for us and for me as a um uh
[11:21] data engineer right I I care very very much about the data foundation and I see that from the um from the enterprise level right before you can do any um uh scale right right it's it's it's good
[11:36] that you do a PC here and a PC there right but talking about the at the scale what we need to do first is always take the data governance in consideration right so you have to put in a process you have to make your data talk to each
[11:54] other you have to develop a context of your data and then we can talk about AI yes okay So you have to build the data foundations. Yes. First. Yes. Um what next? Uh the once you have the data foundation
[12:11] and the data foundation will start not with the not with the data engineer. It's a it's a whole company together with the data foundation. We're talking about the the the the structure, right? How do you make all your silo data talk
[12:27] together, right? Pull that into the foundation. have the governance process on top of it. Okay, build your data context. Build what we call the AI data readiness. uh make sure that you you look at it not
[12:43] only in a snapshot but it's also a continuously right because if you looking it from the uh just a snapshot your data quality change your data change right your data change um every second right and then you you have to
[12:59] continues to to to look at the data foundation build your data foundation and then on top of that will be your genie is will be your AI yes Wonderful. Thanks, Julie. Uh Patrick, as a follow-up to you, um uh Julie spoke
[13:15] about the data foundation elements of it, the data engineering elements of it. How are you approaching this problem? Um at Oxy because in oil and gas, we all understand there are different types of data, right? There are data that is
[13:30] residing in your relational databases, transactional data which resides, operating systems data, all of that resides. Then over and above that you have seismic data, you have well logs, uh you have with ML, DLS, last data,
[13:46] different types of data sources. So when you have those kind of different dimensions of data, how do you prepare that data foundation and eliminate those data islands that Julie was talking about? How are you approaching that problem? I mean ultimately that's the purpose of a data lake or take a lakehouse um is to
[14:04] be able to have multiple kinds of data that reside in even more systems um ready for you. So for instance uh we have physical sensors in the field that measure you know pressure and flow rate and temperature and things like that.
[14:20] Those would be measured um every so often. Those that I mentioned maybe every couple of minutes. Um then vibrations might be measured many times per second. Um things like the weather might only be measured once an hour or
[14:35] once every couple of hours. So you have a very different time cadence. So the things that you measure seldom go into a regular database. The things you measure a lot go into a special dedicated time series database. They live in different areas um because some of the things are
[14:52] out of normal network range. So you you have you know different databases that are on premise that sometimes get updated via satellite. There's a net that networking involved. So you have multiple databases just to cover those industrial sensors.
[15:07] Um and then we have we have drones that have cameras on them. We have field cameras that stream live many many of them right. And then of course the you you can't retain a 247 streamed video from tens of thousands of cameras
[15:22] anywhere and and you pay for the storage bill. So that's basically a throughput and lost kind of a situation. So you have AI in the middle to analyze that. So all these things are happening and that's what a data lakehouse is there for you to resolve is you have all this stuff and how do you put that into again
[15:39] one place uh that has one pane of glass. So you still have the silos, but it's like they're glass silos, right? So you can look at you can look through all of them and and that's what the lakehouse is there there to do and it does that
[15:54] wonderfully. Does it help with your unstructured data as well? Documents that could be PDFs, scan documents. Yes, absolutely. We have lots of documents. Um they're also stored um as as files in a system. Um but at that
[16:10] point it's probably good to mention metadata right so everything has certain metadata so for instance for a time series I have to know which sensor it it was that gave rise to this thing where it's located um maybe you know there's a latitude longitude coordinate where this thing is
[16:27] for a document like who wrote it uh when it was written is there a version two three of this document with red lines or whatever um so that metadata is is essential and that allows us to get some of this context that that was mentioned, right? So, I have to be able to place
[16:42] that in context and then I can say, okay, I have a time series measurement, I have a picture, I have a document and they actually belong to the same thing. Um, and and now I I can analyze that whole thing as a holistic entity and
[16:58] come to some sort of an AI conclusion. Wow, that's a very interesting problem to solve. Um following following up on that Julie um see historically we have seen there is seismic data there is data from geoscience from geology then you
[17:15] have uh uh production systems data the drilling data that you have available each of them are sitting in different silos right so when you're building the data foundation and when you have the data lake data lakehouse um estate as well are the
[17:33] silos becoming getting better with advent of platforms like data bricks and what the lakehouse and uh unity and genie and some of the newer things technology that is emerging is that making the industry's problem much more
[17:49] manageable in terms of the quantum the volume of data that you're managing definitely definitely as you see with the the the lick house right we we don't really care where the data is stored you could have your data in S3
[18:05] buckets or with your a in Azure ADS gen 2 or you can leave your data in the system of record with unity catalog you can pull everything together right you can have your uh delta share uh if you
[18:21] you can have federated catalog you can uh you can pull all the sources together so um in BP what we have is uh since we are global company So you name something I'm pretty sure somewhere in BP we we
[18:37] have that tool set we have that technology but that's why the silo happened in the past it's really hard for us to pull all that data that you talk to get that you're talking about right from geospatial to our map data to our drilling data to patrol to tech
[18:53] locks is quite difficult but with unity catalog right with the lakehouse we now have all the sources is is still not speaking the same language with though, right? But at least we have everything together and now we do the next step. Yes. Okay. How do you see this play out? It's
[19:10] it's the same answer that Julie had. Yeah. What she said. Okay. That's perfect. That makes my job easier. Yeah. Right. So, see what are the biggest challenges um that you face in getting
[19:25] your organization ready for these kind of modernization data AI initiatives. Uh I'll start with you Patrick. What are the biggest challenges from organizational perspective that you see? Because it's never about the technology right uh technology is there all the
[19:41] answers are out there. It's about what questions and the human elements that are involved. So your thoughts on that? Yeah. So I mean the the first problem for companies like us is that this is not the core of the business, right? So um the the vast majority of people
[19:56] around an oil company do not come from a software background, a computer science background. I'm a mathematician. There are very few of us. Um they typically come from an engineering background, a phys a physical engineering background, a mechanical chemical reservoir engineering background, right? So that's
[20:11] that's the first thing to realize. Um so when you come along with all sorts of um IT acronyms, nobody understands them, right? So when when I first arrived um and I talked about AI, I got questions about acoustic impedance and that that's
[20:29] not even a joke, right? So I had to make clear that when I'm talking AI means artificial intelligence and not acoustic impedance or asset integrity. Uh right. So uh that's the kind of you have to level set um right and again speak their language um not teach them your language
[20:46] but you have to learn their language and adopt that. Um and also for example again field engineers they they will not care whether your your data storage is like in a tier one or in glacial storage or an S3. They're like what is that?
[21:02] Yeah just it's a hard drive right? Yes it's a hard drive. It's okay. Um um so you have to meet them where they're at, which is the use case that they ultimately want to put this to like, okay, so there's this complicated IT stuff happening and what I need here
[21:19] is I need a notification on my phone that I can understand and drive to coordinate and act on it. That's that's what you need to deliver and that that's important. So meet them where they're at, I think, is is the key challenge for us. Now for them the key challenge is
[21:37] changing workflows. Um so really it's re-engineering the corporation right? Um they've been doing something in a certain way for a long long time and a new tool will change that way of working. The question is how? Um and the
[21:54] problem the challenge is that most of the people just don't want to right they've been doing it that way for a long time. It feels comfortable. It feels safe. You know we humans hate change. uh for good reason. Um so we have to persuade them why the change is good for
[22:11] them. So that that's a big lesson I've learned over these last few years is telling people why it's good for the company to do this doesn't matter. You have to explain to these people why it's good for them individually. Yeah. How is your life better with this
[22:29] AI thing, right? They don't care what's better for the company, doesn't Yeah. Um, and if if if you have that constructive conversation with people, then then you're you're on to something. And then of course the the third challenge is all of this lumped together
[22:45] over a large group of people is called change management. You know, so you're effectively having that conversation a lot, which means the AI program has to have staff and budget to have that conversation with the many many users across the company. and that usually
[23:00] gets ignored. Um, so you know, as a as a rough rule of thumb, I would say whatever you spend on creating the tech, that amount of money, you need that again, the same order of magnitude money for change management. Wow, that's a big reveal. Thank you for
[23:18] that. So Julie, how do you see this? How do you manage this at a company of BP scale with your complexities? uh and and you at least they are just on the upstream side of things. You are upstream, midstream, you have uh
[23:33] downstream as well and you have a commercial trading function as well. Yes, it is a journey. Yes, it's is a journey and we just at the start of it. Yeah, I totally agree with Patrick. Change management is everything and and
[23:48] uh education and and teaching and knowledge. uh um we start small right in BP we start small um we first start with uh uh the introduction right we uh we roll out
[24:05] co-pilot so people can see with their dayto-day right from email from uh presentation creation from uh from your excel spreadsheets they can see the power of AI that help in their day-to-day and then we roll out
[24:20] something small for the business processes to a small group of people and see how the reaction right how uh the team take it how the engineers that out on the uh out on the field because you know believe it or not they are not like
[24:37] us they not sit in the office and then they can go on uh uh five different AI tool right together at one time right some don't have laptop okay some don't work with a a tablet either they just only have their on. So how do you even
[24:53] roll out uh your AI solution to them? So it is a journey. Totally. Yes. Wow. Yeah. Okay. Thank you. Um so Julie, this question again comes to you. What do you think leaders should do
[25:08] first in modernizing their enterprise for a AI ready era? Right. What do leaders need to do first? um in BP I think I I I see our leader
[25:24] they set example right they set example in the um in the technologies organizations right our CIO first when we start talking about AI about claw code right about uh uh a cursor about
[25:40] that right he actually the one who use that right he's one of the first one who actually use that in is personal and also in the uh in the business context, right? And then when your leader sets example, I can see a change in our
[25:58] workforce in our workforce from a business standpoint, right? when you can see a few use cases, right? A few team that roll out their solution and you can see the optimization, you can see the dollar amount like what Patrick
[26:15] mentioned in the beginning, you can see the change in people but again it is a journey and we're just at the beginning of that journey. So can I phrase it as leadership by example? Yes. From your senior leaders? Yes. And some initial quick wins.
[26:31] Initial quickly. Yes. Okay. So um Patrick, you may have something to comment on these two dimensions. So what should leaders do? Um I know uh you're a contrarian when it comes to quick wins. Uh just so that you guys are aware. So I I'll let uh I I'll
[26:49] let Patrick take that. Go ahead. Yes, I'm a big fan of design thinking um that I would I would recommend. Um so Amazon calls it working backwards, which gives you kind of the theme, right? So the idea is you start by thinking what will happen when I'm done. Um what what
[27:06] does it look like this beautiful new world when this initiative is is finished or mature? Um and then work to the present from that future state. What do I need to do now? What do I need to do in a year in two years to ultimately achieve that big dream that we just came
[27:21] up with? Right? Um and dreaming big is is one of the things. So I I like to start big. Sorry. Um the other thing is though that you need to align that vision with everybody involved which is the hard part right
[27:37] dream big that can happen in minutes in your own mind right but then to align that vision across all the sea level leaders and and VP leaders and so on at the company and get them to sign off on a common vision that's the hard part um and but that that's what design
[27:53] thinking involves right you run workshops with these people to discover everybody's vision look for the commonalities ities argue out the differences until you have a common shared and agreed vision. So that that's the first challenge. The second challenge is when everybody starts changing their mind about that vision.
[28:10] Um but uh that that would be my starting point. Um but leading by example to use the tools is is definitely also uh has been shown to be super super helpful. Um and especially where Genai is concerned, we we've done the same thing, right?
[28:26] We've gotten all the senior leaders to vibe code and to say so in public. Um so to encourage everybody else in the company to do the same. Sure. Um this is my last question to both of you. Um what do you need from a technology
[28:43] partner like data bricks and a consulting services partner like Infosys. In this journey as you go through this what do you need from us? what do we need to be uh cognizant of? What should we be working towards in making your
[29:00] life easier? Um, exactly that, the life easier part, right? So, I I like to refer to as a as a no worries solution, right? Um, we need the tech to just work. Okay. So, again, we're not a software company. We're not a tech company. We want to use
[29:17] AI to solve problems. um data and everything connected to data is a little bit like the crude oil, right? We're like the refinery for that crude oil. But what we actually want is we want to use the stuff to achieve real things. I
[29:34] mean, again, in the oil context, it' be called propulsion, right? You want to move something from A to B. I want to move things from A to B. I don't want to worry about the the road surface uh that I'm moving on, right? So, I need that to just work. Make it easier for you. Yeah, Julie,
[29:50] your thoughts. I only I I am not only needed to work, but I needed to work in the long run as well. Okay. So, yes. Right. And um and I I also need the operational and the
[30:07] support of the role that you have, right? Because the role like Patrick said, right? Sometime you see pothole, right? Sometime there are uh the road that in front of BP uh in Houston we you know it's is it took more than 10 years
[30:24] and they still working on it right right it's it's a long journey we want it to be faster right we want it to be uh um uh supportable and we want it to be I know that now a day every day we wake up
[30:39] there is a new solution there is a new AI right we want what we have is a low stable to support a a old business like us, right? Right. We don't like to change, right? We know that we have to change, but it's so fast every day. Something we
[30:55] haven't even done the solution is new thing out there, right? That faster, that quicker and then make us especially uh engineer, software engineer, data engineer, right? We always want to chasing what's the new thing, what the next thing, what the next thing. But our
[31:12] business is not operate like that, right? So things that work stable, right, and cost less. Thank you very much for being here. Thank you everyone. Thank you.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.