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

Thank you for joining us here. We have with us Mr. Tom Siebel and the management team from C3 AI. Again, thank you for joining us at the Annual Needham Growth Conference and joining us for this fireside chat here later in the day. Tom, thank you very much for joining us and giving us this opportunity.

Thank you, Chris. Nice to see you.

So I guess one of the things that I just want to jump right into and I know we're a little short on time here, but in my discussions with investors, people are still trying to figure out what C3 AI actually does? How you fit into a broad, very fragmented analytics landscape? On the most recent earnings call, I know that you guys said that you are "selling vehicles when everyone else is selling ball bearings." I thought that was a good analogy and a good place to start. But can you talk to the core value prop for C3? Why is it different than others in the marketplace?

Great question, and thanks for asking. So if you think of the C3 AI suite, this has always been a decade in almost $1 billion building, okay? And this provides all of the services necessary and sufficient to design, develop, provision operate an enterprise AI application. So you can think of this as the union of all of the other AI capabilities that you know of in the market. So all of those capabilities are features. So let's take Palantir. Palantir does data aggregation takes -- allows you to aggregate structured data, nonstructured data, telemetry, what have you, into a unified federated image, okay, and represent those relationships visually. That's our data integration capability. Let's take DataRobot. DataRobot and H2O are companies that do AutoML. We have -- AutoML is within capability that we provide. Let's take a platform independent persistence technologies like Snowflake would be a platform independent persistence technology, and we -- basically a Databricks. And so we have that. What would be other examples. Alteryx is for citizen data scientists, and we provide that capability and the function that we call Ex Machina. So if you took all of these bright shiny things, Databricks, DataRobot, Alteryx, Lambda, TensorFlow, all of those capabilities and put them in a cohesive package that all work together out of the box. So you don't have to spend hundreds of millions of dollars trying to couple those pieces together, that's what we offer to the market. okay? That's the C3 AI suite. On top of that, we built 42, I think, 40 or 42, enterprise AI applications that our turnkey applications built with that platform that address the value chains of banking, financial services, manufacturing, telecommunications, defense and intelligence, oil and gas, utilities, what have you. And about half of our revenue accrues today from these applications and half of the revenue accrues from the platform itself.

Great. Thank you for that backdrop. And perhaps we can discuss a real-world example, right? So you're typically dealing with very large enterprises who have already purchased a number of these data analytics solutions and have built processes around. How is it C3's architecture fits within this landscape? And are you partnering? Are you integrating with some of these vendors? Are you replacing others? How does all this hold together?

Okay. And I would think, historically, it's true that our customer base was almost exclusively very, very large organizations. Today, I would say that's changed over the last couple of years. And today, we have relationships with large organizations, with medium-sized organizations. We have small relationships with large organizations, and we have relationships where we're selling to products for $400 a month to very small organizations, let's take San Mateo County, okay, would be a relatively small customer. Now one of the larger use cases we have out there would be Shell. Shell is the, I think, fifth largest company in the world. Largest -- one of the largest oil companies in the world, maybe the second or third. And there, we're using enterprise AI across their entire value chain to build applications upstream, downstream, midstream, integration of renewables with whom do we partner. We partner with Microsoft Azure, okay, that provides the cloud platform. We partner with NVIDIA that provides kind of very powerful GPU technology. We partner with Baker Hughes that has deep expertise in oil and gas. We partner with Databricks that provides virtualization technology. We partner with Alteryx that they're using for their citizen data scientists. We partner with Microsoft that's using Power BI for reporting. And what we are -- and this is -- I think they're using -- they might be using DataRobot for AutoML so that would be one of the partners. So by the way, when I said we provide all of the capabilities necessary and sufficient for these applications, some companies will want to use their technology, their favorite virtualization technology, say, Databricks instead of ours. We use the open source version of Databricks or Spark or they might want to use -- these guys are using DataRobot for AutoML instead of -- what we use out of the box is the open version -- open source version of H2O. So that's fine. Companies can kind of mix and match whatever the solutions they want, but we're building them applications for upstream, downstream, midstream, everything from AI-based predictive maintenance for devices on offshore oil rigs, predictive and descriptive analytics for, I think, 0.5 million valves or we have 2 million machine learning models in production, hydrocarbon loss accounting, process optimization and refining and cracking production optimization in oil production facilities, fuel station analytics. They have 5,000 fuel stations. They sell more coffee in their gas stations than Starbucks sells in any given day, integration of renewables, distributed energy resource management. So these are the classes of problems that we're solving at Shell as Shell reinvents itself into a net-zero carbon footprint company by 2050, we're very much part of that. So it's us, Microsoft, Baker Hughes and these other companies working together to get Shell where they want to go.

Understood. And thank you for that example. If I'm thinking about the core of your platform, right, can you talk to the changes from customer to customer across -- and I know you're working in a number of different verticals like telecommunications, financial services, just the process that you guys have built in place as far as connecting to these different data sources and the user interface involved with that. Can you help us think about that as well?

Yes. One of the requirements of any application of this nature, be it in financial services or utilities or precision health or whatever it might be is you need to -- if you want to buy enterprise AI to optimize any process, supply chain, demand chain, production optimization, demand forecasting, customer churn, whatever it might be. We need to aggregate very large -- a large set of disparate data sources from within the enterprise and from outside the enterprise and from sensor networks. A particularly large example would be Enel. Enel is the largest utility in the fleet world. They're based in Rome. They have about 60 million meters in 40 countries. Putting that in perspective, there are about 100 million meters in the United States. So this is a pretty big grid. At Enel we've aggregated 150 trillion rows of data from I believe 18 instances of SAP, 12 instances of Salesforce, some instances of Siebel systems that are still there, 2 different SCADA systems, Maximo, Atlas. We go out to the Extranet for weather trains, social media, that gets updated 62 billion times a day. And then we are connected to a constellation of I think, 47 million sensors, 42 million smart meters. Some of these are emitting signals at 90 hertz cycles, that would be 90x a second, okay. And we've aggregated this data into a unified federated virtual image where we process these data at the rate they arrive. So our data integration and visualization technology is one of the really keys to the success of this company. And whether we're dealing with the Missile Defense Agency, whether we're dealing with the intelligence community, whether we're dealing with readiness for aircraft, manufacturing, production optimization at places like Phillips, process optimization at LyondellBasell, this is the aggregating data into unified, federated image is kind of a level 0 requirement. It's something we're very, very good at. And one of the reasons why we're being achieving the success that we are achieving. The other aspect that I will point out that is counterintuitive. But as a result of this model-driven architecture, whether we're doing anti-money laundering at a bank, whether we're doing process optimization for polyethylene cracking units or whether we're doing predictive maintenance for devices on an offshore oil rig, 97% of the code in all those applications is the same. It's the same code that's running in every one of these customer solutions. All that changes are the data sources, the machine learning models and the user experience expression. So this is -- we're able to scale this so rapidly across verticals and from vertical to vertical.

That's great. That is a great background. And I did want to touch on it because I know you had made mention but thinking about the machine learning models, how long does it take to train these new models and develop new use cases and absorb or ingest new data types? How is it you guys are managing to all these different problems in the increasingly complex world we find ourselves in?

Let's tease that apart a little bit because there's a number of questions there. I think the one you're really getting at is training the machine learning model, okay? Now we're -- we can -- when we were doing the first application we did at Shell was AI-based predictive maintenance for low-pressure compressors and offshore oil rigs, okay, where we had to aggregate massive amounts of data from SAP, PI tags, other things. We aggregated those data in a unified, federated and built the user experience, built the machine learning models that would predict failure with I think about 85% precision, 80% recall giving them 18 hours notice, which doesn't sound like a lot, but it's enough on an offshore oil rig to prevent a problem. And these guys get real twitchy about problems, failures on offshore oil rigs, as you can imagine. It took us 4 weeks to do that from beginning to end. So that would be an example. I think that the first management application that we built for cash management at Bank of America, it took us -- 6 and Bank of America is a pretty complex organization. The data is particularly complex because it is -- as you know, the aggregation of a very large number of banks. And you get into Charlotte where the data centers are, I mean, these are data centers from many, many organizations in entirely incompatible formats. And we aggregated that -- those data into a unified, federated image, built the machine learning models, trained the machine learning models and stripped the user interface in 16 weeks at the scale of Bank of America. So that will give you a feel for the speed at which companies are able to realize the benefit from what we do.

That's great. And again, forcing or splitting up to different questions. I kind of bubble them all into 1. But if I'm thinking about the new use cases that you guys continue to address, I'm curious how many of these are customers coming back to you saying, "Hey, we have a need for X, Y, Z as a use case versus you guys actually developing these use cases and then bringing them out to market? Does that make sense?

100% of the cases, the customer is coming to us presenting us with a problem, presenting us with a road map and if it's somebody who's in the aerospace business and they have a road map, the person who runs the business is in the room, they have the budget and the will to do it, we're in the aerospace business and we're building it. And then we -- but we do it, Mike, in a way that it is a utility, not only to that aerospace organization, say, United States Air Force, but we'll have the same utility for a Boeing or an Airbus. So then we're in -- but it's very -- in the long run, we will address all verticals. The sequencing in which we address specific verticals is very much coin-operated and demand-driven.

Understood. Okay. And another thing that's come up on earnings calls recently is this advantage you guys talked to with respect to human capital, right? The recruiting process, the hiring that you've done, can you talk about what goes into that and the culture that you guys are building in C3? How sustainably can you continue to ramp at these levels?

I think that is a -- human capital at C3 is a very distinctive and unique competitive advantage. To give you a feel, if I'm not mistaken, last quarter, we had 18,600 job applicants from around the world at C3. We're 700 people, 18,600 on fast path, I think that annualizes to something like 72,000 a year of people, okay, in banking, people from MIT, okay, data scientists from Princeton, computer scientist from name the companies, salespeople, finance people, what have you, who are applying and saying, "Hey, I want to work at C3." I think of those 18,000, I believe that we interviewed 6,000 of them and hired 200. So this would be in order of magnitude more selective than Princeton. If I'm not mistaken, almost 70% of our people have advanced degrees, I believe, 9% have PhDs. The people who work here are tried, tested, experienced professionals, extraordinarily well educated and who are on the top of their game. People we tend to self-select for people who like to work, for people who like to work in teams, for people who like to work with customers, who are people who are challenged by very difficult problems. And I think if you go look at glassdoor.com, anybody who's interested, and you'll see the comments that are -- anonymous comments that are probably less. Some of them are very damning. People who didn't like it here, I mean, they throw us into the bus. But the great majority of them, you will see are we're consistently ranked amongst the top companies in the world to work for, and it is a very high-performance culture. And I believe that, that is -- it is absolutely sustainable. It is probably the promulgation and communication of corporate culture is my important -- my most important job. It's what I do in every communication that I do, whether it's to the markets or to customers or to investors, who really, I'm talking to my employees. And so I believe it's sustainable. And I believe to the extent that we have succeeded in the market, that is the reason why, and I believe to the extent that we will continue to succeed that will be one.

And one of the things that we hear from people in the industry talking to maybe the best way to gauge differentiation is looking at data scientists right, which obviously can't scale. It's a different way to look at the industry. What, in your view, is the best way to gauge talent differentiation of the workforce from an outside perspective like ours?

Would that be our workforce or -- so this issue that I think the question you're asking is can you gauge the competitive advantage of a company by the number of data scientists that work there? I think in the long term, that's true. I mean it's true for Bank of America. It's true for Needham. It's true for Goldman Sachs. It's true for General Motors. It's probably less true for us. I think we have about 90 people in our data science organization. I suspect we have probably 180 data scientists in the company, but it's less true for us because what we're really not doing -- we're less about doing data science than we are about providing a set of utilities that allow data scientists and all these other companies to be enormously productive at applying data science to their financial services company or their automotive company or their agricultural company or whatever it might be to achieve their digital transformation objectives.

Got it. And I know one of the main reasons for the IPO was to build out this sales motion with multiple different channels, right? How is that you are thinking about those sales channels today? How are they scaling versus your expectations?

I think this is going great. This is where I spend the bulk of my effort on developing a large, powerful global ecosystem. We have tens of thousands of professionals selling with us around the world today. I have 12,000 people selling with us in oil gas from Baker Hughes. I have tens of thousands of people selling with us around the market today at -- with the Microsoft Azure team. I have 4,000 professionals selling with us at Google Cloud and Thomas Kurian. I have thousands of people selling with us in financial services at FIS, thousands of people selling with us in the defense and intelligence community at Raytheon. So you can expect to see us continue to invest in this partner ecosystem to take the success that we have achieved. And I believe today, okay, I believe we are the world's leading provider of enterprise AI applications. And if I am able to leverage this through a large and powerful partner ecosystem, including the hyperscalers, okay? And then specialists, you say oil and gas with Baker Hughes, the utility market and ESG and sustainability would be NG. Financial services would be FIS, defense and intelligence could be Raytheon, more to follow, think telecom communications, think precision health, think ag, whatever it might be, that is going to be one of the -- we'll prove when the book is -- when the story is written, we'll have proven to be one of the critical success factors that contributed to the success of the company.

And maybe just taking a step back, when you -- how isn't your first going about identifying these partners and then working with them to announce like a more formal partnership, like I think about the Google Cloud partnership that you recently announced? Obviously, top of mind large name in the industry, but how do you decide to go with them versus maybe another guy or think about Raytheon and the FIS, what's the process you guys go through on your side internally there?

Well, the truth is that Baker Hughes selected us, okay? The truth is that Google Cloud selected us. So -- Thomas Kurian decided that he was going to compete in the hyperscaler market on the applications layer rather than compete based upon selling against Azure and Oracle or whoever it might be, AWS based upon CPU seconds and storage hours with an architecture that may well be arguably better. So rather than compete on that basis, he decided he wanted to compete through the applications layer. And so he approached us because we have 42 turnkey applications that address the needs of the entire value chains of oil and gas, of utilities, of financial services, telecommunications, manufacturing, what have you. So if you want to knock, I mean, if you want a family of 40 turnkey enterprise AI applications in multiple verticals that are tried, tested and proved in the market that you want to immediately be selling worldwide, there's exactly 1 door you could knock on in the world, and it's this is it, okay? We're the only company in that space. So they approached us. And Thomas said, "Hey, we want to sell your applications worldwide through my global sales organization. And as the CEO of Siebel Systems doing the best I can to represent the interest of our shareholders it is kind of an IQ test, right? As it relates to another example would be Baker Hughes. Baker Hughes, as you might recall, was formerly General Electric. General Electric was formally a thing called Predix. Okay. Predix was by about a -- as I recall, $6 billion catastrophe, okay, to try to build something equivalent to the C3 AI platform. Okay, after years of failing at Predix, Baker Hughes, the CEO of Baker Hughes showed up in my boardroom and said, Tom, I want to partner with you in oil and gas and then turn 12,000 people lose selling these products globally into Aramco, Rosneft, Gazprom, Shell, LyondellBasell, Flint Hills, Coke, what have you, and that was an offer that was pretty difficult to refuse. So I would say, in many cases, these guys are approaching us and they are approaching us. And I think we can agree that companies like Raytheon, Microsoft, Google, Baker Hughes, ENGIE these are very, very high-quality go-to-market partners.

Right And I know other companies have tried to do this, right, unsuccessfully. They'll try their hand and then they'll eventually turn to you or what is it -- do you think that the market still has players out there who are trying to go it alone and develop it themselves? What is the realization that, hey, instead of continuing to go down this route, why don't I just go with the C3?

Virtually, first of all, yes, okay? Just like every other market we have seen develop in information -- software information technology in the last 40 years, relational database, CRM, ERP, whatever it might be, I mean, companies tried to build all these things themselves. I mean how many companies succeeded to building their own relational database system. That would be none. Okay. How many succeeded. This is what IT used to do. IT wants to build -- wanted to build their own CRM systems, how many people succeeded at that, not very many. So virtually, every one of our customers, Baker Hughes, Shell, ENGIE and now Bank of America will have tried 1, 2, 3, 4x to try to build this platform themselves before they throw in the towel and decide that maybe that isn't their core competence is building extraordinarily complex enterprise application software and they should focus rely on a professional organization to provide it and focus on areas that are there areas of core competence.

Great. Really just to focus on your core competencies, outsource to this guys, if you're developing those 42 turnkey applications that leverage so well with your partners.

And make no mistake, our competitor today, okay, as companies that is basically build it yourself, that is the competitor, okay? And if I had to look at our revenue, rough numbers, I would estimate -- and this is just a guess, okay? So you can't hold me to this, okay? But I would say that 80% of our revenue last quarter accrued from companies that at one point in time turned us down because they were going to build -- they were building themselves and came back to us 1, 2, 3 or 4 years later and standardized on our technology. So this is how we do business is letting companies go build themselves for a while and fail in it and then we come in and help them out.

Great. And one of the things, too, I know -- I think it was the last earnings call, but you guys also spoke to the entry point for these contract deals. I think it was $16 million a couple of years ago. The most recent quarter, I want to say it was somewhere around maybe $4 million. And I know that there are pieces that go into that from your platform or whether or not we're talking about Ex Machina as an example. Can you talk about how that's making C3 may be more approachable, even in the mid-market or making it easier for enterprises to the bite the bullet rather than spending billions of dollars trying to build this themselves before they finally do turnaround company?

Our average transaction value has continued to decline over -- in recent years, and this is from memory, but I think it went from -- average from $16 million to $12 million to $7 million to $5 to about $3 million, okay, in terms of average transaction. This is creating greater diversity in our revenue and should be taking the lumpiness of our bookings, okay, last quarter is being an exception, okay? And the -- but today, I mean, we sell products, if you go on to our website, download Ex Machina which is a application for citizen data scientist, he can use it for free for 30 days, and then if he decides to keep it, I think it's maybe $400 a month. And so our transactions vary from very, very large to really quite small. And whereas if we look at 2014, 2015 -- I'm sorry, 2016, 2017, we were really focused on only the world's largest corporations. Today, we sell to large corporations, midsized corporations, small businesses, local governments, San Mateo County government, Stanford Hospital. And so they go from some of the world's largest organizations to small businesses.

And with something like an Ex Machina, right, if you have this 30-day free trial, people are aware of like you're almost building up your cheerleader base, right? People get familiar with the technology, they fall in love with it. Is that the approach then that you're seeing from this almost bottoms up while you guys are still attacking from a top-down? Is that a way to think about the go-to market there?

Top down, middle end and the bottom up. I mean we're going after the -- I mean, our goal is to establish, Mike, a global leadership position in enterprise AI. And so you can expect to see us to continue on major markets, enterprise accounts, middle market accounts, okay, and citizen data scientists.

Terrific. Looking at the vertical partnerships, I know we were hitting on that earlier. But with those different verticals and the different partners because obviously, everyone's going to kind of look at this differently, which verticals -- which partners are you most optimistic on today or do you have the highest expectations for as far as gaining traction out there in the market?

Microsoft is huge. Google is huge. Baker Hughes is huge -- Baker Hughes, I mean those numbers speak for themselves. Those are -- I think our relationship with Raytheon is really productive. So those are the ones that -- those are the partners who are -- I know we are hitting a long ball, and this is where I spend the bulk of my time on the creation and development and nurturing of those partnerships. And I'm very pleased with the way it's going.

Are there certain verticals like if I think about FIS feeding into the financial sector, when you say the financial sector is further ahead or behind other verticals out there that you're currently feeding into?

Well financial services is a relatively new vertical for us. I think it's going to be huge. I mean we started in utilities, then we went into oil and gas, then we went into manufacturing. Now we're in financial services. I mean in the long run, all enterprise applications will be predictive. Okay. All enterprise applications will require the kind of capabilities that we offer. All enterprise applications will require different forms of machine learning. So I believe the largest market for what we do in the long run will be unquestionably precision health, okay? But it's not an early adopter.

Terrific. And I think we have to leave it. I'm sorry, I thought we had more time here, but we are at the top of our time. So thank you very much, Tom, and to our listeners. I really do appreciate your participation in the conference.

Thank you, Mike. Enjoyed the conversation.

Absolutely. Be well.

Bye.