C3.ai, Inc. (AI) Earnings Call Transcript & Summary

Hello, everyone. Good morning, and we appreciate everyone joining us for our Second Annual BofA Digital Energy Conference. I'm Chase Mulvehill, the Senior Oilfield Service and Midstream Analyst here at Bank of America. I'm also joined by Brad Sills, who heads up the coverage of the enterprise software sector here at Bank of America. To kick off this year's Digital Energy Conference, we're fortunate to be joined by Tom Siebel. Tom is the founder and CEO of C3.ai. C3.ai is a pioneer in the AI and enterprise AI space. Many of you may remember Tom from this. Tom is Chairman and CEO of Siebel Systems, which ultimately merged with Oracle back in 2006. Tom, thanks for joining us today, and we look forward to hearing insights into this ever-evolving digital industry and the impact digital transformation is having on the oil and gas sector today. So with that intro, why don't we just jump right into the Q&A session.

So Tom, let's kick it off by taking a few minutes to hear about your past and how you ultimately decided to pivot from CRM, where you founded Siebel CRM and sold it back to Oracle in 2006 to starting C3 to focus on AI.

Well, thank you, Chase, and it's great to be here with you. This is my fourth decade in the information technology industry. I did my graduate work in Relational Database Theory at the University of Illinois and went to work for a small startup company in the early '80s by the name of Oracle Corporation. And they had this idea that they were going to commercialize relational database technology, and they might have been a $2 million business at the time. I think they have about 20 employees. And they had, I think, about 12,000 feet at 2710 Sand Hill Road in Menlo Park, and that was the professional experience of a lifetime. So I worked with the people at Oracle for about a decade and was involved in the commercialization of relational database technology, and that was quite a professional experience. And we established and maintained a global leadership position in relational database systems and then in application software. And then I went on to found a company called Siebel Systems, where we embedded -- and then Siebel Systems was where we embedded the CRM market as you know of it today. So this was about the application of information technology and communication technology, the sales, marketing, customer service, sales automation, call center and Internet cell service, product configuration, marketing automation, what have you. And so Siebel was a rapidly growing enterprise application software company, I believe, today, the fastest-growing enterprise application software company in history. And we started in about mid-1993. By 2000, we were doing about $2 billion in revenue at, I believe, 8,000 employees in 29 countries and established a clear market leadership position globally in CRM, in call center, in field service, in sales automation, in market automation across all verticals and all geographies. I think when we sold the company to Oracle in January of 2006, I think we had about 75% market share globally in CRM. Today, CRM has grown to be a -- this year, it's a $60 billion software business and a $60 billion services business. So people will spend $120 billion on CRM this year, so the installed base of CRM software has to be something in excess of $1 trillion. When we -- after selling Siebel Systems to Oracle, we began -- I got together a group of people, and we started thinking about what was next in information technology. And we met in the spring, summer and fall of 2007 -- well, 2007, and then more actively in the spring, summer and fall of 2008. And it occurred to us that there was a step function in information technology that was going to change everything about computing, and these vectors included elastic cloud computing, the Internet of Things, big data and predictive analytics. So in January 2009, we set about building a platform that would allow organizations to a software platform. We've provided all the services necessary and sufficient for companies to design, develop, provision, operate these predictive analytics or enterprise AI applications that took advantage of those technologies. And so now fast forward 10, 11 years or so, we are a public company. I believe we're the world's leading provider of enterprise AI applications. We provide our core application called the C3 AI Suite, which companies use to design, develop, provision, operate the enterprise AI applications of their choice. And then we have 40 enterprise AI applications and product lines for oil and gas, for utilities, for manufacturing, for aerospace, telecommunications and manufacturing. And the -- so this has been quite a professional experience. It's a very rapidly growing market. This market for enterprise AI applications is expected to be a $300 billion software market in a few years, which would be larger than the entire information technology market globally when I went to work at Oracle. So that's what we've been up to, and it's in so far the professional experience of a lifetime.

Yes. And so I guess, Brad, did you have anything to add?

No, no. Go ahead, Chase.

Okay, okay. Sorry for telling you to jump in. So some -- obviously, you have a storied history as a pioneer in the software enterprise space, so probably who better to ask this question. And I think you actually wrote a book. I'm actually looking at it right here on this. But -- so what is really meant by digital transformation? What did it mean 10 years ago? What does it mean today? And what's it going to mean in the future?

That's a good question, Chase, because as I started visiting customers and boardrooms in 2009, '10 and '11 and '12, whether it was Shanghai or New York or Paris or Rome, every CEO was talking about digital transformation. And yet when you poke at it a little bit, they had no common concept of what it was. And candidly, I found the term a little bit perplexing because -- and I was thinking like digital transformation as opposed to what. I mean, analog transformation. I mean, what -- so it's confusing. And so I gave it a lot of thought over about 7 years and talked to a lot of people. And I believe what digital transformation is all about, it's about the application of this new step function of technologies that we talked about, elastic cloud computing, big data, the Internet of Things and predictive analytics to deploy these enterprise AI applications. So these idea of applications that you can build applications that are predictive is a whole new idea, okay? And it fundamentally changes the way that we manufacture products, the way that we service customers, the way that we drill for oil, the way that we explore, the way that we reinvent companies. Like -- companies like Shell are reinventing themselves to be zero-net carbon footprint companies. So I think digital transformation is about adopting this new set of technologies that I think will probably be necessary to survive in the 21st century economy.

So maybe it's also kind of expand on that a little bit. We've got a lot of oil and gas investors watching today that may not be so familiar with C3.ai. And so maybe just take a moment explaining your AI platform and the applications that you offer and really kind of what's the value proposition to the oil sector -- the oil and gas sector. I know we're going to step through here in a minute and really dive into some of the things that you're doing with Baker and Shell and some other companies, but really just kind of simplify it for a bunch of oil and gas investors on kind of exactly what it is, this AI platform and the value that you can bring to the oil and gas sector.

Okay. So when we're looking at applying predictive analytics or enterprise AI to the value chain of oil and gas and renewables and increasingly, renewables that are coming, what this is all about, it's about aggregating all of their data sources, subsurface data, ERP data, OSI PI tags from all of their equipment, telemetry from all of their devices and valves and ESPs and rod pumps and offshore oil rigs, okay, and refineries and pipelines, weather, terrain, social media, commodity prices, GDP growth rates, what have you, aggregating all of those data. And these are massive data sets of structured data, unstructured data and sensor data, aggregating those data into unified federating image and then processing those data as they arrive and processing them through machine learning models, typically supervised learning and unsupervised learning models, sometimes deep learning models to basically bring a normal social and economic benefit to the value chain. And the examples would include things like AI-based predictive maintenance for devices on offshore oil rigs, okay? So let's take a low pressure compressor, for example, where which is one of hundreds of devices on offshore oil rigs. So we can recover when this device fails undetected. The results can be catastrophic, environmentally catastrophic, okay, and economically catastrophic for the production company. With using this AI-based predictive maintenance to predict device failure with very, very high levels of accuracy, say, many hours in advance of failure, so that the production operation can avoid the failure and either turn off the equipment or replace the piece of equipment, so that no environmental hazard occur. We -- a less dramatic example might be a submersible pump. We use AI-based predictive maintenance on submersible pumps or rod pumps or equipment and refinery, so that we can identify a piece of equipment that's going to fail 50 or 100 or hours or a week in advance of failure, you can replace it and then you don't have all the downtime and then, okay, and the rebuilding time and the lost production cost. Hydrocarbon lost accounting is another example. AI-based predictive maintenance for wind generation. Well placement analytics. A well production optimization analytics. Process optimization in refining. Integration of renewables, distributed energy resource management, AI-based prohibitive maintenance for wind turbines. So the net of these, when you apply these AI applications with large across the enterprise say the size of Shell, the economic benefit that you're looking at is in the order of EUR 3 billion, EUR 4 billion, EUR 5 billion a year in recurring economic benefit, and the result is safer energy, cleaner energy, a less environmental impact, and that's the -- that's kind of what this is all about. So that's what AI does to the oil and gas value chain.

Great. Kind of if I may jump here for a second here. Great to see you again. Thanks for doing this, and thanks, Chase, for having me. What you're describing are real business metrics, real business problems that you're solving for. But when you peel back the onion, you're really -- there's a lot behind the technology that enables this. My question is are you selling more to IT buyer? Are you selling more to the business owner? What's the typical sales audience that you're kind of going in? And what have they started with? AI concept has been around for many years, but you guys are bringing to market a platform that really encompasses the whole food chain here end to end. So have they started with these AI initiatives? And where have they failed? And how does C3 come in and solve for that? So I guess, a 2-part question, the one would just be on your typical sales audience, second would be where are they coming from.

Our type of call, the customer to whom we are speaking is in the C-suite, okay, the Chief Executive Officer, the person refining for -- the person in charge of drilling, upstream, downstream, retail, okay, midstream, what -- that's the person we're usually talking with, okay? The -- every one of our customers, okay, will have spent sometimes 2, 3, 4, 5 years trying to build these applications internally, either using the primitives that are providing -- provided by the hyperscale providers, like AWS or Azure or the Google Cloud, and frequently in combination with these many bright shiny objects out there that are kind of confusing with the IT community, and I'd say confusing to the investor community as well, where you have all of these components out there, like Databricks, like DataRobot, like H2O, like a Snowflake. By the way, all products that offer, I think, value, okay? I mean, these are good products that offer value, various database systems from Oracle and Amazon and utilities like Bigtable and BigQuery from Google. And some almost invariably, our large customers, like [ in now ], like [indiscernible], like LyondellBasell, like NG, like Shell, like Baker Hughes, will have spent some years trying to assemble scores to hundreds of these components into a unified cohesive platform that does something useful. Well, people have found these propositions to be minimally complex, exceptionally expensive and almost invariably collapse of their own way. So after they've tried that 2 or 3x, they will come to us and we provide all of the services, all of the software services in one package that are necessary and sufficient to design, develop, provision, operate these types of applications. And then for oil and gas, with our partners at Baker Hughes, we have built an entire family of oil and gas applications, enterprise AI application software that runs on top of that platform, that does things like readiness and predictive maintenance and hydrocarbon loss accounting and, okay, and production optimization and refining. And so that's who we talk to. That's the journey our customers go through, and some of these failures are pretty spectacular. I think GE might have spent $6 billion, I think, in the decade trying to build this thing called Predix, which was coupling together all this componentry. IBM had a project that I believe is now canceled. They still use the brand a little bit called Watson, where they clearly spent scores, that billions of dollars trying to cobble together componentry. And we have a unique technology built on something called a model-driven architecture, where we built this thing from whole cloth with the idea that it would be a cohesive unified solution, and that's what's unique in the market, and it's unique in the oil and gas market, and it's very unique in our partnership with Baker Hughes that, as you know, we sell actively in the market under the brand, Baker Hughes C3.ai.

That's great.

And in a moment, I want to pivot over and kind of really dig into the partnership with Baker. But since you mentioned this, I want to go ahead and ask it. Predix obviously was a big failure. And so I guess the first question is, what was it about Predix that made it fail? And then number two is what is it about your platform -- your AI platform and applications that avoids the pitfalls of Predix?

Well, Predix is architecturally identical to what companies either have tried or are trying to build themselves. And this is where you take a number, you take the solutions provided by one or more hyperscalers, like AWS or Azure. And then scores have open source and proprietary components from companies like Databricks and Alteryx and others and try to cobble them together with DataRobot, H2O, SparkCognition, Cassandra. And you try to cobble these things together using structured programming. And you identify cohesive [indiscernible] that works. But when you consider the number of things that you need to connect with structured programming, the number of data sources that you need to connect, the number of machine learning models that you need to connect and the number of fields and the data sets, what you're dealing with, I mean, a large implementation might be 100 million or 200 trillion rows of data from 100 unique sources, and we're dealing with hundreds of petabytes of data that all need to be connected together just to unwind the -- kind of the data lake hair ball. Well, there are a number of things that you need to connect there. It Is in the order of 10 to the 13th. Well, this is a nontrivial problem, okay, 10 to the 13th. And even better, you try to solve the problem of that complexity using maybe 2,000 to 3,000 programmers in Fort Knox who are communicating by conference call across 12 time zones. Now it's -- just it's mind numbingly difficult, okay? Now what -- and to our knowledge, everybody who has attempted that has failed, and no one has ever succeeded. Now what we did was something very unique is we built -- we use something called a model-driven architecture, and for anybody who's interested, you can look that up on Wikipedia, which is a very -- it is a different model than structured programming. It is a logical extension of an idea that was first developed by the object management group about 2000, this thing that some of you will recall called a service-oriented architecture, where everything is an abstraction, and it reduces the complexity. It's a model-driven architecture. We own all the intellectual property, and we have the patents on the use of a model-driven architecture for data aggregation, for enterprise AI, for analytics, for data presentation, what have you. And this model-driven architecture, which is at the heart of the C3 AI Suite and all of our C3 AI applications, reduces the level of complexity for, say, order of 10 to the 13th or 10 to the 3rd to design, develop, provision and operate these applications. So it reduces the code that somebody needs to write by order of 3 -- maybe 3 orders of magnitude and reduces the cost by 2 to 3 orders of magnitude. So we're able to deploy very, very large-scale complex applications, the largest AI applications in deployment on earth in relatively short amounts of time, enabling organizations to realize economic and social benefit in -- of great magnitude.

And so -- sorry. Go ahead, Brad. Sorry.

Yes. If I just double-click on that real quickly, Tom. I mean, the model-driven architecture also enables that ability for you to swap in or swap out some of the components that maybe a customer has already built out. There are a lot -- there's a lot behind the team on the platform. There's the development framework and standard management. There's data integration. There's an ML library in there. And in many cases, companies have already made investments in one or more of those components. So how does that architecture enable you to come to market with a product that you can say, "Hey. We'll let you leverage what you've already invested in so far into our platform," that ability to swap in or out with that may already have been built?

Thank you for asking, Brad. Yes. So our -- the C3 AI Suite consists of hundreds of microservices that do things like data connection, data persistence, authentication, access control and factor authentication, queueing, ETL, machine learning services, event processing, map produce, continuous analytics processing, data visualization and then application development tools. Now -- and out of the box -- and when you build an application on the C3 AI platform or of any application that we provided from C3, we will run without modification on any hyperscaler environment or on any standalone device. So this could be on, say, an NVIDIA device on the edge. It could be on the Azure Cloud. We can move it to any application from the Azure cloud to the AWS cloud, to the Google Cloud, to the IBM Cloud, to Bare Metal, to NVIDIA GPU device on the edge. And these applications will run with that modification. Now virtually every one of our customers has made technology decisions for one reason or another that they prefer Databricks for virtualization. Well, we often use the open-source version of Databricks. If they want to use Databricks in that slot, put it in there. Or maybe, they want to use Alteryx as a data science tool instead of our -- a citizen data science tool or they want to use DataRobot for AutoML or H2O for AutoML. Or they want to use Aurora -- Amazon Aurora or Snowflake as a relational database system. And so you can plug any one of those components into our architecture, and it continues to run. And so that's just fine. It's not a problem. It also future proofs our architectures. So as people come out with better virtualization technologies, better machine learning libraries, better machine learning services, more efficacious platform-independent persistence technologies, whatever it may be, you can plug it into our architecture and works just fine. And virtually every one of our customers does this. They do this at Air Force. They do it at Shell. They do it at LyondellBasell. They do it at Coke. They do it -- and now ENGIE, you name it.

Yes. So if we can kind of just pivot a little bit and look at your competitors out there. Some of them are actually speaking at the conference today. But you've got Schlumberger. They talk about their DELFI platform. You've got Halliburton that talks about their iEnergy platform. It sounds like yours is a lot more comprehensive. It's kind of an enterprise AI platform, where theirs is probably a little bit more siloed. But maybe, I don't know if you can like explain the differences of kind of your offering and how that compares to kind of what you see out in the oil and gas sector that's more kind of specific to just kind of oil and gas versus something where yours is you can do it across different verticals.

Well, I think, where Baker Hughes is unique in the oil services provider marketplace is they have this model-driven architecture where we're able to deploy these applications very quickly into production. And what's very interesting is when we deploy, say, AI-based predictive maintenance for offshore oil rigs, production optimization for refinery, hydrocarbon loss accounting or retail fuel station analytics, 95% of the code is exactly the same. It's that same code for all these applications. I mean it's miraculous. They're sharing the same data. They're sharing the same objects. They're sharing the same machine learning libraries. So this application that we've built for refining has utility for predictive maintenance for valves, has utility for hydrocarbon loss accounting, has utility for retail fuel station analytics. All of these assets are reusable. If we look at the work that's being done at Halliburton and Schlumberger, it's very similar to the kind of work that GE was trying to do with Predix, where they're building application-specific software stacks using structured programming that are highly complex, where the work that you do for upstream is not reusable downstream, is not reusable for refining, is not reusable for distributed energy resource management. So we have this kind of this common resource that allows a Baker Hughes C3.ai that allows our customers to very rapidly move from 1 application to maybe 100 applications. And the largest use case we have in oil and gas would be Shell. And we have -- if you look up shell.ai, that is Baker Hughes C3.ai running on top of Azure, and we're very tightly coupled with Azure there. And then as Brad mentioned before, they use some third-party utilities. So I think they use Docker for containerization, Kubernetes. I think they use DataRobot for AutoML. I think they might use Alteryx today for their citizen data science. But that's the difference between -- so there's really a huge difference in the solutions that Baker Hughes is offering in the market versus their -- versus Schlumberger and Halliburton, they're -- where they're currently using structured program and -- to try to build these systems.

Yes. I appreciate the color there. And kind of talking about Baker Hughes, I guess you're a few years into this joint venture, this partnership with Baker Hughes. So looking back, like, why did you ultimately choose Baker Hughes versus some of their competitors? And then talk about how important Baker Hughes is in your go-to-market strategy.

Well, Baker Hughes chose us, and the -- after -- and Baker Hughes, as you will recall, was the primary distribution channel for GE Predix. And I -- so I remember it was a few years ago that we were visited by Lorenzo Simonelli and some of his people that wanted to approach the market with a -- they did their research and what was available in the market. And they decided after doing that research, they wanted to form a strategic relationship with C3.ai. And so we agreed to enter into a strategic alignment, something we referred to as a joint venture, where, I'll call it C3bakerhughes.ai, where we have agreed to partner with Baker Hughes to the exclusion of all other oil and gas service providers. And then Baker Hughes agreed to standardize on C3.ai to the exclusion of all other predictive analytics software solutions. And so we jointly design, productize, market, sell and service these solutions globally, and it's a wonderful partnership. They're a great partner. It's -- we're accomplishing -- I think we're just getting started. It's early days of what we will accomplish, but I think it's -- we've made a lot of progress. And we both spent a lot of time and energy on it, and we're -- it's a great partnership.

Yes. And so you talked about some of the successes you've had a little bit. But Shell, you've got a good partnership there across both Baker and yourself. And if we think about the successes that you've had at Shell, could you maybe just point to some of the value that you've been able to drive? I think in one of these conferences, I think you mentioned a number that maybe they put in their 10-K or something. So just remind me of what that number is, if you remember. And then how do you -- how do they actually measure the success of digital adoption for them? Are there certain KPIs or anything like that they look at and they say, "Okay. Well, here's what kind of value you're adding for us." And so yes, I mean, you just -- I would imagine some deep -- some big thorough discussion was had when you signed -- when you renewed the contract, I guess, was it last year or earlier this year with Shell. So...

We've renewed the contract with Shell multiple times. The first use cases were AI-based predictive maintenance for offshore oil rigs. The second was, I think, production optimization for LNG operations in Australia. And then we've moved beyond that, beyond that, beyond that. But one application they have, they're doing AI-based predictive maintenance for 100 -- and descriptive maintenance for 130,000 valves, which is how many valves they have in Shell. That involves 2 million machine learning models in the pipeline in production in concurrently 2 million machine learning models. So these are -- I mean, they must have 100 applications in flight at Shell today, upstream, downstream, midstream, and they're using it to -- this is very much part of the strategy to reinvent Shell. This is a massive digital transformation to reinvent themselves as a zero-net carbon footprint company. And C3.ai is -- I'm sorry, Shell.ai is at the leading edge. And when you look beyond underneath the covers of C3.ai, you find -- I'm sorry, Shell.ai, you'll find C3.ai and Azure.

Right, right. Great. And if we step back, and we'll ask [indiscernible] probably this a little later today, but digital adoption across the oil and gas industry space is slow. And I'm not quite sure why if it's oil and gas guys are stubborn or if there's -- It just takes a long time for you to actually understand the value or it takes a long time to implement the infrastructure to be able to extract the value. But what is it about the oil and gas sector that has made it slow to adopt digital?

Well, I think the oil and gas sector has just been kind of thrashed around with oil and gas prices in the last year. And so it's tough. And these guys, as you go from, what, $115 a barrel or whatever it is, Chase, you're an expert on this and not I, to negative $37 a barrel, okay, and then $35 a barrel, I mean, these guys are just getting thrashed around where they're in survival mode. And so it's -- I mean, as it goes from $115 to negative $37, you're really not thinking about which information technology project you're going to deploy. Unfortunately, you're thinking about how many people you're going to lay off. And so these poor guys have just been getting thrashed by economic events, okay, pandemic events, geopolitical events in the Middle East and Russia. So it's -- I think that's both kind of slowed it down in the short run, but I think accelerated -- accelerating it very quickly. I think as companies stabilize, and these people realize that they need to fundamentally reinvent themselves or they're going out of business. And the -- and so we have leaders like Shell, who understand that. I think Aramco understands that. And it's -- but it's been -- the last 3 years, it's been a rough road in the oil and gas business. I'm glad I'm not in that business, but it's -- that's affected this kind of innovation when you're in just kind of survival mode.

Yes, yes. I feel the pain sitting here covering the sector. So they've now got me covering 2 sectors instead of 1 because the oilfield service sector has shrunk so much. But we've only got a couple more minutes here. So one question I did want to ask, and if you kind of look out and think about the future of the oil and gas sector, and you think about where Baker Hughes C3 will have the best adoption rates, do you think it's going to be kind of more in the traditional upstream space, more in the midstream space, more in the downstream space? Like where do you feel like the value proposition for your digital offering will drive the most value for your customers?

Well, I think that, first of all, Baker Hughes has unusual strength, I think, in the LNG market. So there's no [indiscernible]. I think as it relates to the integration of renewables, that's a common case across all the utilities. But I think honestly, upstream, midstream, downstream, I mean, it's all -- whether it's well placement analytics, predictive analytics for refining operations, hydrocarbon loss accounting, retail fuel station analytics, integration of renewables, I mean, we address the entire value chain together with the -- and we have the deep, deep, deep domain expertise of Baker Hughes, coupled with the enterprise application software and AI expertise at C3 AI, and it's a pretty powerful combination. And then we partner with our -- we're very closely aligned with the hyperscaler providers, most importantly, Azure and Google Cloud, so we can prevent -- we can present together a very, very compelling solution to a medium or large oil and gas company for all aspects of their business.

Awesome. Got a great partner there with Baker Hughes for sure, and we look forward to a lot more news on the oil and gas side from C3.ai and wish you all this success, Tom. And with that, I'll turn it over to you for any closing remarks.

Well, Chase, it was great to talk with you. I wish you -- it was an honor to be included in your discussion today, and I look forward to continuing the dialogue going forward. So Chase and Brad, thank you, guys. It was nice to talk with you, and I hope you have a very successful conference.

Thanks, Tom.