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

Welcome to Piper Sandler's keynote discussion with C3 AI. I am Arvind Ramnani, senior research analyst at Piper Sandler. I have the pleasure of introducing Tom Siebel, Chairman and CEO of C3 AI. Previously, Tom was the Chairman and CEO of Siebel Systems, which merged with Oracle in early 2006. Welcome, Tom.

Good morning.

Yes, I've known Tom since the late 90s, when I was starting on my career at Deloitte and Tom was running Siebel Systems. So Tom, the world has changed a lot over the past 30 years since we first met. We now have the Internet. We have mobile devices. And of course, the global pandemic that we all lived through. Amazingly, we have a fully functional life, fully functional work and all the comforts even through the depths of the pandemic. I guess my opening question is you've been around for a few years. What is your overall assessment of the tech landscape? And how do you see the world benefiting more from some of the newer technologies such as artificial intelligence?

Well, I know if I can be politically incorrect, as I always am, Arvind, if we're going to look at roughly, I guess, 100,000 years of homo sapiens on the planet, pandemics come and go. And this is not going to be the last one, and they don't tend to be -- we have the Bubonic plague which we had, the Spanish flu. And I think the relative order of magnitude of what's going on with COVID here is insignificant compared to those. And I think the idea that this is going to fundamentally change mankind is unlikely. Now that being said, if we look at technology, fire, the wheel, domestication of agriculture, the steam engine, Jacquard loom, the transistor, I mean these fundamentally changed things, forever, in the -- with the Jacquard loom and the steam engine that brought Industrial Revolution, the Gutenberg press that brought on the reformation. And now we have this phenomenon of the transistor that has changed everything in the last 40 years about the way we live, the way that we communicate, the way that we entertain ourselves, the way that we do business. Now the big change we've seen in the information technology business grow from order of a $200 billion business maybe in 1980 to what, say, $4 trillion today, you'll know better than I. And now we have this accelerant with the elastic cloud and I think the elastic cloud just changes everything. Because all of a sudden, we have infinite computing capacity and virtually infinite computing capacity available essentially for free, infinite storage available for free, and we're able to solve classes of problems that were previously unsolvable. So you combine this with the phenomenon of the Internet of Things and big data and we're going to see clearly a dramatic acceleration in this, what we call, digital transformation or some people call the -- Daniel Bell called the post industrial society. And this is big, Arvind. This is the biggest thing that we've seen since the invention of the wheel.

Perfect. Perfect. Just kind of drawing through your breadth of experience, right? You built Siebel from scratch, you grow it to a $30 billion company. You've written some books, I have one of your books right on my table, which is, by the way, is a really good book. You had a wrestling match or you've been trampled by an elephant and now you're running C3. What is the inspiration for starting C3? You could have done a number of things. Just really trying to understand, why did you build this company?

Well, after we merged Siebel Systems with Oracle in 2006, we just thought about what was happening next. And when we were thinking in 2006, 2007, 2008, when we planned C3, it occurred to us that we're about to see a new step function in information technology that was going to change everything. And those included -- step function included elastic cloud computing, included big data, the Internet of Things and predictive analytics. Now when you think back to 2006, '07, '08, the elastic cloud market was this big, right? I mean it was nascent. Well, it turned out to be a pretty good bet. It was true. And so we began in 2006, building an integrated set of software services that we call the C3 AI Suite because we thought that people would want to use this new set of enabling technologies to design, develop, provision, operate applications to take advantage of these technologies. And there's -- the beauty of this step function is that now we're able to solve a class of problems that were previously unsolvable in the area of predictive analytics. And so this -- this is -- predictive analytics is what is popularly called say, enterprise AI. And so we've been at this for about a decade. We spent hundreds of millions of dollars, perhaps approaching $1 billion on development of this technology stack where we've provided an integrated set of software services in 1 cohesive platform that provides all the services necessary and sufficient for companies to decide to develop, provision, operate enterprise AI applications. On top of that, we use that platform to build 40 turnkey applications for utilities, for oil and gas, for aerospace, for defense, for intelligence, for banking, for telecommunications that allow them to deploy these applications to do things like stochastic optimization of supply chain, demand forecasting, fraud identification, anti-money laundering, ballistics calculation whatever it might be, that are applications of just enormous social and economic benefit. So it's really fun. It's really hard. It's a -- it is the next-generation of computing. I believe this will be an entire replacement market for everything that we've done in enterprise application software in the last 3 decades. And if we look at general analyst reports, this is predicted to be -- this software market for AI applications is estimated to be like 1/3 of $1 trillion software market, come on, 1/3 of $1 trillion is a pretty significant addressable market opportunity and objective that we have with C3 AI is to see if we can establish and maintain market leadership position in that market. Will we succeed at that? I don't know. Will we be a significant player in that market? I think by the way if not us, who. We're surely going about it like we did at Oracle and like we did at Siebel pretty methodically and aggressively. I think if we come up short, we will still have built one of the world's great software companies.

Perfect. Perfect. And I think your go-to-market approach is also very unique, right? I mean, it -- kind of building AI company is difficult, but also the approach you've taken is, I think, quite different. You're looking to really leverage partners such as Baker Hughes, FIS in specific verticals. What is the rationale for partnering with these vertical leaders versus really going after market by yourself?

It gives us market leverage. I mean we are 700 people. I have 12,000 people selling for me at Baker Hughes into oil and gas. At FIS, I have thousands of people selling with me into financial services and banking market. I saw with Microsoft across a wide range of markets. Now most significantly is the announcement that we made, not too many weeks ago, with Google Cloud. So Thomas Kurian at Google Cloud decided that he wanted to 2 years ago. He made it -- when he took the job, he said, we're going to go after this market in a different way rather than selling CPU seconds, okay, and storage hours with an arguably superior technology at maybe a lower cost and rather than compete on that basis, he decided that he wanted to compete through the application layer that he was going to offer applications for -- he's taking a sales organization from 400 people -- order of 400 people to order of 4,000 people in the last 2 years, and he said, we're going to approach the market through the applications layer and offer supply chain optimization, supply network risk. All of the things that these SAP guys are really good at, manufacturing optimization, banking applications, what have you. And so after 2 years of looking at the market, you think about it. If he wanted to partner with a company that offered a turnkey set of applications that meet the needs of energy, oil and gas, utilities, telecommunications, aerospace, financial services, manufacturing, there's only 1 door in the world that he could knock on, okay? That's the door right there, and so that's only 1 door in the world that he could have knocked on. And so he knocked on that door and we formed a very strategic partnership with Google Cloud where they will be selling our 40 turnkey applications through all of their sales and distribution organizations globally. So this increases our distribution capacity by 1 to 2 orders of magnitude and will be a very important partnership. But we are looking for market leverage. The name of the game is put the pieces in place with Microsoft, with Google Cloud or in aerospace with Raytheon, in financial services with FIS, oil and gas with Baker Hughes, others to be announced where we have a very, very large sales and service echo system where we have literally hundreds and thousands of people delivering our solutions in Asia, in EMEA and in north and south America across all verticals. And if we do that, I think there's some probability that we will carve out a substantial chunk of the market.

Yes, make sense. I mean I totally see the value for C3. What's the value proposition for Google or for Microsoft or for FIS to basically sell C3 services?

Well let's take Baker Hughes. Baker Hughes competes against -- from their oil services company and they compete against a roughly $24 billion business. And they compete against Halliburton and Schlumberger. And now well, they have a turnkey set of AI enabled applications that address all the needs of oil and gas, upstream, downstream, midstream to be able to bring AI to optimize those applications and Halliburton, Schlumberger doesn't, that puts them in a unique position. Let's take Google Cloud. Google Cloud now has 40 turnkey applications in their bag that they can sell and their competitors are selling CPU seconds, storage hours and kind of various micro services. These guys are delivering turnkey applications. So it's a very unique position in the hyperscaler market. If we take Raytheon, which is our partner in defense and aerospace, here, again, they're able to offer turnkey solutions for AI-based predictive maintenance for aircraft, say, F-15, F-16, F-18, F-35 Joint Strike Fighter, logistics applications, insider threat applications. And the other Beltway bandits are basically offering large science projects with thousands of people where they take multiple years and billions of dollars to build things from scratch. It gives each of them a competitive advantage in their market to serve their customers better.

Make sense, make sense. How are you weighing -- what's your criteria when you chose a vertical like what's next in terms of additional verticals to enter with the offerings?

Well what's next is one that I'm particularly excited about is AI-enabled CRM. The CRM is a market that I'm not entirely unfamiliar with. And I've been thinking for it in the last 8 years about applying AI to CRM to really take this to the next level. You recall when we started Siebel Systems, there was no CRM market. We invented the CRM market. Today, it's a $60 billion business, $60 billion in software annually and $60 billion in services annually. So this year, the CRM market is between Accenture, PwC, Deloitte, et cetera, Salesforce, Oracle, Veeva, that's a $120 billion business this year. So the installed base out there has to be, what, $1 trillion? I mean it's $0.5 trillion? It's whatever it is, it's big, if it's $120 billion this year. Now so what we can do, what we -- what the C3 AI CRM product that we announced today, and you can go online and you can see it is basically this is really something where we can sit on top where we're not going in and ripping, replacing. We're not competing with Salesforce, we're not competing with Siebel, we're not competing with SAP, we're not competing with ServiceNow. We sit on top of that and we enable -- let's take aerospace market. Let's take Boeing. Maybe for Dave Calhoun, okay? And Boeing used to be about $50 billion commercial aircraft manufacturer. And imagine, in addition to the CRM data from salespeople opportunities, forecast or whatnot, we're able to incorporate all of the data about aerospace industry, all of the data about their customers, Lufthansa, Emirates, United Airlines, Southwest Airlines, NLP on their annual reports, equity prices, NLP on their analyst reports, NLP on social media, commodity prices, Jet A, GDP growth rates, unemployment rates, travel policies, COVID incidents, what have you. If we can aggregate all of those data into a unified, federated image, and these data sets are now 2, 3, 4 orders of magnitude larger than the data that are on the CRM system. With those data, we can deliver very, very precise AI revenue forecast, very precise, look, these revenue forecast that come out of the CRM systems, today, are absolute garbage, right? They're just the sum of what all the sales guys put in and the sales guys just put in whatever it takes to get their sales manager off their back. Very precise revenue forecast, next best product, next best offer, customer churn, what you'll see or what we've done with CRM, and I really encourage you to look on the website to see it. We have fundamentally reinvented. So not only did we take all the data out of Salesforce or Dynamics and integrate it with Outlook or Gmail and SAP and Ariba and Workday or whatever it might be, and all these exogenous market data. And then we have fundamentally changed the human computer interaction model as it replace the CRM, where we're visualizing all the relationships between people, between markets, between customers, distributors what have you, in terms of sales vision, market vision, customer vision. It is way cool. And this again. It isn't rip and replace. We can put it on top of this maybe $0.5 trillion to $1 trillion investment that's out there and make all of these CRM applications that are out there fully predictive. So this is an exciting space, keep your eye on it, and it's something I'm -- I've torn up [ ten ] product specifications over the last 7 years to get us to the right one, and we finally made it. And it's truly exciting.

Yes. I'm sure I'll work out quite well, just given your history with CRM. But just kind of focusing a little bit more on the product side of the equation. I think one of the key components of your tech strategy is your model-driven architecture. This enables your clients to leverage their capabilities in a structured manner and consume your offering in bite sizes. I also think like it really enables you to have a more enduring product offering of the -- so 2 questions on this. Can you talk about the differentiation with your model-driven architecture? And second, conceptually if you take us to kind of the blueprint, what was the rationale for creating this model-driven architecture?

Well, the alternative to using C3, the model-driven architecture. So what we have -- what we already have built into C3 is hundreds of micro services in the C3 AI suite that are designed to interoperate together, persistence, queuing, ETL, data science services, what have you, okay? And when we build an application on C3, that application will run without modification, whether it's anti-money laundering, whether it's stochastic optimization of the supply chain, whether it's demand forecasting, customer churn. That application will run without modification on AWS, on Azure, on the Google Cloud, on the IBM Cloud or on Bare Metal. Now the fact that we've designed all these systems to work together avoids the problem of these companies have to engage with these massive science projects that always fail, where they take, let's say, AWS primitives and combine it with Databricks and DataRobot and the Oracle Database and this and that the other thing and try to use structured programming to connect all these things together. Well, IBM spent scores of billions of dollars on that with Watson. How did that work out? I mean, [ for a second now ] Watson is gone, right? I mean, GE spent, I think, $7 billion, and 10 years trying to do this. And how did that work out? So I think GE Predix as far as I can tell is gone. Virtually every one of our customers, Shell, Enel, ENGIE, Coke, United States Air Force, United States Army, Standard Chartered Bank spent years trying to build these systems themselves before they shelved it and put in the C3 AI suite because it works, it avoids them having to spend 5 years building an enterprise data lake and integrating all these services and allows them to get on the business with designing and deploying applications. So it allows them to get the job done, say, 1 to 2 orders of magnitude faster, 2 orders of magnitude less code. And that is what we've done that's unique with the model-driven architecture. Now it will also be entirely open architecture. So if somebody wants to use Databricks for virtualization instead of we use open-source version of Databricks, that's fine. Or if somebody wants to use H20 or DataRobot for AutoML instead of what we use, which is the open source version of H2O, that's fine. They can put it in there. If they want to use Snowflake for relational database store, they put Snowflake in as a platform independent database store. So it's an entirely open architecture where every company has the standards that they've chosen, and they can put those standards into -- seamlessly into our architecture. It makes it future-proof, and it seems to work out quite well.

Perfect. Perfect. And I assume essentially this playground that you've created for pretty much allowing any of the other tech vendors to come and play with, that's well received. Like, I mean you're -- both from the partners, but also customers, they like this kind of platform that allows basically plug-and-play of multiple, multiple technologies.

Well, it enables them to deliver solutions to their customers quickly and at lower cost and customers tend to like that.

Yes. Of course, of course. So the product road map for a company like C3 is certainly complicated. Oftentimes, I think you're building soft in a vacuum, right? Like when it comes to like AI technology, there's really no customer requirements compared to just sort of building something that -- there's actually no clear specification. Can you talk about what informs your product road map? How are you prioritizing what you're building versus what you're partnering with?

Well, we thought, first of all, there were some fundamental principles. We spent 4 decades now in enterprise application software. So I think we do have some understanding of that, okay? In terms of the DevOps requirements, in terms of the multi-platform requirements, it was clear to us that people would want to be able to -- they would want to have independence from the cloud providers and be able to run AWS or Azure or Google Cloud or IBM and didn't want to be tied to any one. The idea of using metadata to describe the nature of the relationship and the nature of the application, rather than hard coding these things is a concept that we embedded at Siebel Systems with Siebel version 7, so we're pretty familiar with that. So I think the fundamental requirements, multiple language, multicurrency, multiple language, I mean we kind of have experience with that, Arvind, access control, end factor authentication, logging, what have you. So those are requirements that we really did understand and we built them into this platform. And I think somebody doing this for the first time just doesn't understand that. As it relates to the industry-specific requirements, those were developed as customers walked in the door. When Enel and Con Ed and Duke walked in the door, we had to build industry-specific requirements for utilities. When Shell walked in the door at Baker Hughes, we had to build the industry-specific applications for oil and gas. When the Secretary of the Air Force walked in the door and the Chairman of the Joint Chief of Staff, we had to understand and build -- apply this to aerospace, defense and logistics. So we do it as when Bank of America walked in the door, then we focus on AML and cash management. So it's relatively trivial for us to apply this to different use cases. I say what's really unique about this technology is whether we're doing cash management of Bank of America, AI-based predictive maintenance for the F-35 Joint Strike Fighter or production optimization for polyethylene or say paper -- yes, polyethylene at foothills, which is Coke Industries. 97% of the code is exactly the same across all of these applications. Now that is unique, okay? And that is one of the advantages of this model-driven architecture.

Yes. Yes, that's interesting. I mean, I know we've talked about it before, but I think -- good to hear from you again. How is cash management and predictive analysis? Like how is that code unique? Because different industries and applications.

It's I said unique. I misspoke. It's common, okay? I misspoke. It's common to all these. But the things associated with ETL, the things associated with cache, the issues associated with event processing, machine learning services, data persistence, these are common across all of these applications. What varies in each application are the data sources, so the data sources that we aggregate are unique to each application. And sometimes, they're quite large. For example, Enel which is a large utility based in Europe. It's the largest utility in the free world. In that instance, we've aggregated 100 trillion rows of data from 50 enterprise information systems, 47 million sensors, 42 million smart meters. I think 18 instances of SAP, Salesforce, Siebel, Dynamo, Atlas, 2 different SCADA systems with regard to the extranet to get weather, terrain, social media, what we call open source data, that gets [ up to dated ] 62 billion times a day. This is a 100 trillion rows of data aggregated into a unified, federated image and some of these data are being upgraded at 90 Hz cycles. Now that capability to do that is really the same, whether you're doing it for a utility, whether you're doing it for oil and gas company or whether you're it for bank, they're just different data sources. All of the issues, the event processing, analytics processing, machine learning services, queuing ETL, access control, those are common, they're common to every application persistence. The user interface expression varies in each industry needs for application, but most argue that that's trivial, okay? The data aggregation issue, this is nontrivial, but we have made it trivial, okay? So we have some secret sauce there as a result of the model-driven architecture. And then what really is non trivial is the development of these machine learning models. So we allow data scientists to focus on building machine learning models rather than spending 95% of their time on wrangling data, understanding how data are interconnected and understanding what processes act against those data. So we abstract all that away, so data scientists can do what they're best at, and build machine learning models and put them, used at scale. So we have some applications. And we have one application at Shell, for example, it's a predictive and descriptive maintenance application for -- they have 150,000 valves at Shell. They need to [ check ] the valves very carefully because when they stick open or stick closed, things go real bad real fast. Again, in that application alone, we have 2 million machine learning models that we're managing in the pipeline. But -- so we allow data scientists to do what data scientists are best at rather than have them wrestle with how our data persisted, how do I persist the data, how do I wrangle the data. Most of these efforts, that's where they all begin and end is trying to figure out how to deal with the data image.

Perfect. Perfect. And I think like the simplification, I think, is a very powerful concept. We have talked about Ex Machina, which is really kind of No-Code AI and analytics platform. And I think certainly, simplification probably drives broader user adoption. Can you talk about the market opportunity and where you're seeing traction?

I think there's a pretty -- actually it's a market that's been around for a while, but now it's growing very rapidly. This opportunity for the citizen data scientist. This is somebody who is an analyst who's used to doing business with Excel pivot tables or maybe Tableau where you want to be able to analyze data. And Alteryx has achieved some degree of success there. I think they sell I don't know $400 million, $500 million, $600 million worth of stuff every year, which is basically a canvas where you can drag and drop data from a variety of different sources and then apply AutoML techniques to do structured learning, nonstructured learning or deep learning processes on these data sets. And so Ex Machina is a product that anybody who's on this call can go download it for free on the web, I encourage you to do it. And don't forget to leave your e-mail there because we'll be calling you in 30 days and asking for your credit card number. But okay, but you can use it and it allows -- it's a very, very powerful product that allows citizen data scientists to aggregate data from a multiplicity of data sources in the unified, federated image. And then as somebody who's knowledgeable about data, but not knowledgeable about data science, they can now provide -- apply very sophisticated techniques to analyze data for market segmentation and analysis, for whatever it might be. And that product is available on the web. It's available for free for trial. And the way that we're selling it is a little bit different. We're really focused on the CIO who wants to bring 100 citizen data scientists live or we had a call this morning where they with one of the departments of -- in the U.S. military, where they have a need to bring thousands of citizen data scientist live. And so rather than going after it 1 at a time, we're going after in a manner that we're -- kind of feel comfortable with, which is large enterprise selling. So I think that will be -- as we get into next year, I'd expect that to be a pretty substantial market opportunity. We have that fueling growth. We have CRM that will fuel growth. We have the Google Cloud relationship that will fuel growth. We're doing a lot of work. For those of you that want to go look at go click on the CRM site today, and you'll see something called Data Vision. This is a fundamental new paradigm, okay, in the user -- human computer interaction model for enterprise applications. You can see this paradigm. So rather than being restricted by tables and by forms, we're now whether we're representing data in a digital twin, in a supply chain, in logistics, in demand forecasting, relationships with people. We're doing this in a visual graphical metaphor that changes everything about the human computer interaction model and enterprise computing. And Arvind, as I'm telling you now in whatever in -- this is going to be an enormously, enormously exciting and, I think, game-changing technology.

Perfect. Perfect. I mean we've talked at a high level and you've sort of given us some examples, but can we dive a little bit deeper, specifically into some case studies or customers that have seen success? And can you kind of just walk us through everything from what was the problem? Where did they apply C3? And what were some of the sort of business benefits of working with C3?

Well, now it's a smart grid analytics problem. We're providing -- I think this is the largest AI application -- enterprise AI application in production in nonclassified space in the world, okay? And it now runs about 60 million meters in most of the European and South American grid. And we're using it for predictive maintenance, for devices in the distribution and transmission value chain. These are things like transformer substations, reclosures and we can predict device failure before it happens, so they can replace the transformer before it explodes, okay? And then when it explodes, 2 things go bad, okay? The lights go out in San Francisco and you start fires that like kill thousands of people, see PG&E for details, okay? And so that's -- the social and economic benefits of that are huge. There's an application in there called volt-VAR by bringing voltage and imaginary voltage into phase, it results in -- reduces the amount of fuel that you need to burn to power the grid by like 8%, distributed energy resource management is a very complex problem that can only be solved through AI, where in any given millisecond or the amount of power that you've generated and the load need to be equal. Well, this was relatively easy when in the old days, when you were generating power and everybody was a consumer. Well, now everybody is a pro server because out there on the consumer side, you have solar, you have wind, you have all these forms of energy that are coming both ways. In order for this grid to work, you need to balance this distributed energy resource management problem, energy efficiency, integration of renewables. So if we were to look at the annual report for Enel, Francesco Starace, their CEO is looking for, I think, EUR 5.1 billion in economic benefits from these digital initiatives with C3, EUR 5.1 billion a year, I mean come on, this is significantly non-0. Shell, Shell is the I think fifth largest company in the world, I believe the second largest hydrocarbon producer in the world, and they are using C3 AI to completely reinvent Shell upstream, downstream, midstream, who are applying what they call, they call the Shell.ai, which is C3 AI combined with the Azure Stack for everything from AI-based predictive maintenance for offshore oil rigs, AI-based predictive maintenance for ESPs, submersible pumps, production optimization and wells, placement analytics and wells, integration of renewables as they reinvent themselves to a hydrocarbon -- 0 net hybrid footprint company, production optimization and refining. These are the types of fuel station analytics. There are 5,000 fuel stations. They sell more coffee than Starbucks does, it's true. And so these are hundreds of applications across Shell. These are 2 kind of very large use cases that are probably the 2 largest commercial use cases, I think, on earth of enterprise AI.

Perfect. On the flip side, has there been any clients or kind of projects where you didn't reach the desired outcome, so you had to really kind of go back and relook at your core or your offering to see some success?

Yes, there have been cases. For example, there was a manufacturing company in the Midwest that did a trial with C3 AI, it was hugely successful. And then they decided -- then the CIO decides that he's going to build this from scratch. So Krishna's got 20,000 people in Bangalore and he has his PhD from Florida State and how hard can this be, right? And so they spent 4 years trying to build it out of DataRobot and Databricks and AWS primitives. And so then 4 years later, when nothing is working and the CEO gets kind of cranky and then they come back and they license it from us. So there are people every now and then who will have a very successful application and some CIO want to come in and try to build it themselves, him or herself. And by the way, this is what happened when we introduced relational database technology. I mean believe it or not, all these companies were going to build their own relational database systems in the 80s. When we introduced CRM and ERP in the 90s, everybody was going to build their own CRM system, their own ERP system. I mean how many companies succeeded in building their own ERP system? So that would be 0. And so this is a natural phase that companies go through. So sometimes we get started and they're quite successful, and they'll hire a new CIO from some place who says, oh, I'm going to build this from scratch. There's a bank that will remain unnamed in New York that is spending $1 billion. They just spent $1 billion in [indiscernible], they're going to spend $1 billion trying to build the stack. And when we talk to the person in charge about why do you want to build it when it already exists and he mentioned the CEO's name that is a very well known and respected name in banking, and I won't mention his name here, but his response was, this person's first name that we all know very well, okay, gave me $1 billion to solve this problem, and I'm going to spend it. This is a true statement. So yes, you do run into that. But is it the way -- this is a natural part of the adoption cycle of new technologies.

Yes. Yes, makes sense. Last couple of questions for me. So far, you've largely been focused on larger enterprises. But earlier this year, you have shifted the focus more to small- and medium-sized enterprises. Can you talk a little bit about kind of the rationale for the shift and how you're progressing on the -- with kind of servicing the small- and medium-sized companies?

Well, I think it's a really important question because historically, our bookings and billings has been characterized by lumpiness because we're dealing on very large transactions. In fiscal year '19, our average transaction size was $16 million, average transaction size, which is -- this would be in order of magnitude larger than any software company you and I have ever seen. So we talked about this, and we wanted to reengineer our organization to put in not only elephant hunters, okay, they're outdoing these large mega deals, elephant hunters, something near and dear my heart, but deer hunters and rabbit hunters and hummingbird hunters, okay? And partners to do that with us. So this involves now people who are motivated to selling the departments, telesales, selling on the marketplace, the idea was to bring the average transaction value down, also lower-priced products like CRM, okay? So selling to smaller organizations. We had some success with like we're selling what would be a Spearstone, Novela Neurotech, Skillion, CerebralEdge. Well, most people will not recognize these companies, okay? And it used to be all of our companies were like Enel, Shell, United States Air Force, they're a multibillion-dollar organization. So they resell to smaller organizations, and we sell a lot of lower-priced products. So this has been pretty successful. Our average contract value came down the last 3 years from $16.2 million in fiscal year '19 to $12 million in fiscal year '20 to $7.2 million last fiscal year to I think $4.5 million, if I'm not mistaken last quarter. So we're doing a very good job of bringing the average transaction value down. And we need to continue to do that to get the lumpiness out of our booking stream so that we have more predictability in bookings and billings, which I think the market is going to -- investors are going to demand from us. And we have -- it's still a work in progress. We haven't completed it yet. It's lumpier than I would like it, and it's going to probably take 4 to 6 quarters to get all the lumpiness out of it. But we're on the job, and I think we have a good plan to do it.

Perfect. Perfect. Yes, I got asked a question on the elephant. I know I asked this when I was in your office earlier this year. You have a picture of elephant right by your desk. And I think most people know kind of -- yes.

Okay, right there?

Yes, that's the one. That's a big elephant. You had quite a few stories to tell. But can you -- like how do you have this picture of this elephant right next to your office after sort of the wrestling match that you had with this elephant?

Arvind, I was attacked and mauled by an elephant August 1, 2009. And after that, I had 19 reconstructive surgeries. At first, there was a lot of questions if I was going to live. After we got through that, there were like 3 surgeons that wanted to cut off my leg. And we got through that and we just fired those guys. And you've seen me and you know the leg that's on here is my leg, and I'm in a reasonably good shape. It changes your sense of risk and it changes your chance of perspective. Every now and then I have elephant on the wall and it looks -- for those of you who ever get charged and attacked by an elephant, that's about what it looks like right before it's about to take you out, okay? And I don't recommend it. Now the -- I have it there just to put things into perspective. Every now and then somebody will get on the phone, doesn't happen much anymore, and explain how they're going to dismantle us and get -- act pretty tough. My reaction to this is if you don't weigh 5 tons and you can't run 30 miles an hour, you got nothing. And so it kind of just puts everything into perspective when the pressure is on.

Perfect. Perfect. Just last 1 for me. You're certainly seeing some sequential improvements in technology and then some step function improvements in technology over the last like 30 years, 40 years. When you look out over the next 1 to 3 years, what are you most excited about from overall tech perspective and specifically from C3's perspective?

I think we have trivialized. So everybody's been wrestling with these issues of how do you solve the digital transformation IT problem, okay? And I think, honestly, we have trivialized that, through model-driven architecture. We have solved that problem. I think the next -- I think what we're going to see next -- what are important issues. I think we're going to -- we need to fundamentally change the human computer interaction model. I think that's going to be through visualization and voice. I think we have some -- this AI is going to bring on some very serious ethical issues that need to be addressed so these systems are not misused to enormous social detriment, see Facebook for details. The -- so I think we need to be concerned about that. I think as it relates to the impact of cloud computing, I think we're in the first half of the first inning. I mean this is going to be huge. I think computing and the edge is going to be very important, particularly with a lot of the work that these GPUs and the work that people at NVIDIA are doing, but we're seeing dramatic, dramatic acceleration. I think we're also seeing acceleration in things that are very dangerous that we need to be worried about as it relates to social media. And I think some of the consequences of what are going on in social media are really troubling. And we need to start giving that some thought or else we're going to have to live there.

Perfect, Tom. We're out of time, but thank you so much for taking the time to speak to us. And thank you, everyone, for listening in.

Great to talk to you, Arvind.

Of course. Thanks, Tom.