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

All right. I think we're ready to get going with the next session. I'm DJ Hynes. I'm the senior software analyst here at Canaccord. Thanks, everyone, who's tuning in. I've said it a bunch of times. I'll repeat it. This is the 41st year that Canaccord's had this event. We couldn't do it without the support of all the clients that dial in and participate in these sessions and certainly without the corporates who are the stars of the show. So we're delighted to have C3 here. We have CEO, Tom Siebel. We're going to do this as a fireside chat. I have a list of questions that we'll run through. It's intended to be interactive, though. So for the folks who are dialed in, please submit your questions. I'm going to open up the dashboard here. I can integrate them into the conversation, but don't hesitate to participate in the session. So with that, Tom, thank you for doing this.

Look, my guess is most folks in the audience are familiar with you if they're old enough to have been investors back with your first success. But maybe just give us a bit of your background and then back and dovetail into the idea behind C3 and kind of the quick story of how we got to where we are today.

Okay. This is my fourth decade in the industry, DJ. I'm a computer scientist from the University of Illinois, did my graduate work in relational database theory, went to work with a little start-up company. We've had about $2 million in revenue in Menlo Park called Oracle Corporation. And I think we had about 20 employees at the time, and that turned out to be a pretty good idea. And I worked with Larry and that crew for about a decade. I was ultimately one of the guys who ran that business. And then spun out of Oracle in -- to start a company called Siebel Systems. That was about the application of information technology and communication technology to sales, marketing, customer service. We invented this market that you know of as CRM. Today, I think, an order of $90 billion market. And we -- that was a very rapidly growing company. I think by 2000 -- we started in '93. By 2000, we're doing about $2 billion in revenue, had 8,000 employees in 40 countries. That company, as you know, was merged with Oracle Corporation in January 2006. And so after that, we thought about what was happening next, what would happen next in information technology and spent a few years planning through 2006, '07 and '08, and occurred to us that back when I got involved in the information technology business, it might have been a $200 billion business globally. Today, it's, what, order of $3 trillion. And it occurred to us that the next step function in information technology was going to be about elastic cloud computing, big data, the Internet of Things and predictive analytics. So in 2009, we started C3. And the big idea was to build a software platform, a set of software services that would enable companies to design, develop, provision and operate enterprise-class AI applications that took advantage of this new step function in technology. And while elastic cloud computing and big data and predictive analytics were a -- it was a -- these were nascent markets in 2008, by 2021, they are no longer nascent. And so it turned out to be a pretty good bet. And so it looks like we -- after 10 years of software development, it looks like we ended up in the right place at the right time with the right product. And so now we are in the business of providing the tools and applications to people, some of the world's largest corporations to design, develop, provision and operate enterprise AI applications. And so we're kind of the really, as far as I know, the only kind of pure-play enterprise AI software company, and this is what we're doing.

Yes. It's a perfect backdrop. A few good bets there in your career. It's funny. All right. Look, so C3 is -- it's engineered with what we've talked about being a model-driven architecture. Just talk a little bit about what that is, right? I think at its core, it's that all the components of AI and data management can kind of be assembled modularly, right? And you guys act as like the fabric that holds everything together. Just talk about like why that's the right model, how it differentiates C3 and what do you can do for your customers in terms of time to market and cost savings?

With the traditional way that software is built in the 21st century, so the state-of-the-art is something called structured programming. And what organizations will do when they want to build enterprise AI applications or platforms like GE with Predix or IBM with Watson or these large projects that took place in companies like Maersk or a Shell or Enel or Bank of America is they will go out and either use open source componentry. They'll use services, software services that are provided by the hyperscalers. And they will use proprietary technologies, things like AWS, Lambda, Snowflake, Databricks, Data Robot, Amazon Aurora. So they choose all of this componentry that handles things like data aggregation, MuleSoft; data connectors, MuleSoft; data persistence technologies, Cassandra, Oracle, Amazon Aurora, ETL systems, queuing systems, auto ML systems, graphic display systems. And they will use structured programming, and they'll choose order of 100 of these things and then attempt to use structured programming to couple these things together into a uniform, seamless whole that works for building general purpose enterprise AI applications. And GE spent about $6 billion and a decade on this and before that collapsed. IBM spent scores of billions on IBM Watson before that collapsed. And virtually every one of our customers, be it Enel, Shell, Coke, the Department of the Army, has had multiple attempts to try to build these things from scratch. They invariably fail. So what we have the idea of a model-driven architecture, and anybody who's interested can look it up on Wikipedia. It's a pretty elegant idea. It's basically, they're taking the logical extension of what the object management group used to call it a service-oriented architecture and build an integrated set of services that were all designed to work together that handle things like persistence, connection, ETL, queuing, MapReduce, machine learning services, data visualization, what have you. And so this is an abstraction layer that reduces the complexity of the problem from the data scientist or programmer perspective from, say, order of 10^13 or 10^3. And net-net, it enables us to rapidly design, develop, provision and operate a very large-scale enterprise AI applications. Now this is a technology on which we own all the patents. So if somebody wants to apply this idea, basically the idea of using a service-oriented architecture for data aggregation, for machine learning services, for data presentation, for the purposes of building an enterprise AI application, we have been granted all the intellectual property on that. So that is the secret sauce. It reduces the amount of code that needs to be developed by order of 1,000, okay? It reduces the order of 1,000, 3 orders of magnitude. It reduces the cost and time to develop by order of 100, 2 orders of magnitude. And it is the long-term strategic competitive advantage that we have at C3.ai.

Yes. Yes. Yes, it's important to understand. So I appreciate all the detail there. I'm going to dig into some of the stuff that you're doing with the product in a minute, but let's talk about just kind of common use cases that you're seeing in your core verticals, right? Like help someone who is maybe new to the story understand what are the key issues that you're solving for your core customers? And you can talk about some of the new use cases that are surfacing. But I think understanding the core is important.

Well, I'll talk about the business problems rather than the technical issues. So the business problems, killer apps are things like AI-based predictive maintenance, okay? So why is that important for a place for a company like Shell, okay? We can use AI-based predictive maintenance to predict device failure on an offshore oil rig before that device fails. We do not have an environmental disaster. CEO doesn't go to jail, okay? And there's not billions of dollars of fines. So -- and we do that at Shell. We do that for all sorts of assets across the upstream, downstream, midstream, okay, in the oil and gas industries in companies like Shell. Another example would be smart grid analytics for grid operators like Duke, New York Power Authority, Con Ed, ENGIE in France and now in Enel in Rome. Enel is the largest utility in the free world. They have 60 million meters in 40 countries. Now putting that in perspective, there are 100 million meters in the United States. This is a pretty big grid.

Yes.

And we use AI and smart grid analytics to do everything from distributed energy resource management, very difficult problem associated with integrated renewables, an application called volt-VAR and by bringing voltage in the grid, it's the largest and most complex machine ever built. In Enel, we've aggregated 100 trillion rows of data, 100 trillion rows of data from 50 enterprise information systems, 18 instances of SAP, Siebel, Salesforce, Dynamo, 2 different SCADA systems, Atlas. We go out to the extranet for weather, terrain and social media. We update that 62 billion times a day. We've integrated data from 47 million sensors. This is the largest AI application deployed on earth in nonclassified space. We aggregate this data into unified, federated virtual image. We did not even move any of these data. We process the data at the rate that they arrive and what -- where's the payoff, okay? AI-based predictive maintenance for power transmission and distribution assets. We can identify a device failure before it happens. So the transformer doesn't explode like it does in Northern California every day and kill thousands of people, okay? And the power doesn't go out in Rome. It's an application called volt-VAR, we bring a voltage, an imaginary voltage in the phase, reducing the amount of power that we need to fuel the grid by 10%. Energy efficiency programs. I talked about distributed energy resource management. The economic benefit of what we're delivering at Enel, and it's in their annual report, is order of $5 billion a year in economic benefit.

Yes.

United States Air Force, AI-based predictive maintenance. This is predicting device failure or system failure before it happens in aircraft like E-3 Sentry, that's the AWACS, C-5 Galaxy, F-15, F-16, F-18, F-35 Joint Strike Fighter. Some of these airplanes cost you $100 million a copy, okay? And so if you can, in F-35, I think it's roughly $100 million a copy. So we increase the availability of these aircraft by 10%, 20%, 30%, 40%. The implications on cost and, more importantly, readiness, are very significant. So these are either fraud detection, whether we're doing anti money laundering for banks, okay? Whether we're doing -- or whether we're doing fraud, for example, in energy. Who's stealing energy? How much are they stealing? So these are typical kind of bread-and-butter applications of AI of enormous economic benefit. If we look at Shell, Shell's looking for order of EUR 4 billion a year in economic benefit for the projects that they're deploying with us upstream, downstream, midstream at Shell. They call it Shell.ai, which is basically the combination of C3.ai and Azure services.

Yes. Yes. Look, I mean, it's one thing for me to say, like, hey, C3 is solving important problems. But then when you go into the details and you lay out just exactly what you're doing, I think it hammers it home. The follow-up question to that is, you talk about saving tens of millions of dollars, hundreds of millions of dollars. How do you think about pricing software that can deliver that type of savings? Like how do you make sure that you're getting your share of the economic value that you're delivering?

Well, in some of these cases, we're talking about billions of dollars in annual kind of benefit. So this is new, and I've been in the information technology business for a while, okay? And when you're in the boardroom, talking to the Board about $2 billion, $3 billion, $4 billion of economic benefit, and they believe that could be true, imagine when you get to a place like the Department of the Army or Health and Human Services. You think they have fraud in Health and Human Services, logistic issues in the Army, I mean the cost issues, numbers get staggering. The -- I'm sorry, I got so caught up in establishing the economic benefit, DJ. I forgot the fundamental question. What was...

Just about like how do you make sure that you're pricing the software right to extract the value that you're delivering?

Well, we -- okay, our pricing model is basically based -- at a per application development -- price for an application, it's based upon a per developer, cost for the developer. And then they pay us per CPU second on a usage model, which is very common in cloud computing. I think that we -- I believe that we charge a fair price. As you know, we get -- historically, our average transaction values have been extraordinarily high, and they have been coming down over time. But we're -- we believe we charge a fair price, but it is -- and it is a fraction of the economic benefit that accrues to our customers.

Right. Right. A question that came in from the folks tuned in online was around, is this just for like the biggest of the big customers, the Fortune 100, the Fortune 500s? Or can it democratize down into the mid-market and smaller businesses? And maybe that's a good bridge to talk about Ex Machina.