#271 Steve Lucas: Why AI Agents Will Automate 75% of Business Operations by 2026

0:00You'll have an intelligent agent that knows who you are, your job, what you do, your company, and it'll say, just FYI, three expense reports came in, I already approved, let me know if you need anything, we will build that agent, no code, just describe your job, what you know, you know, you know, you know, you know, you know, you know, you know, you know, you know, you know, you know, you know, you know, you know, you know, In business, they say you can have better, cheaper, or faster, but you only get to pick two. What if you could have all three at the same time? That's exactly what Cohare, Thomson Reuters, and Specialized Bikes have, since they upgraded to the next generation of the cloud, Oracle Cloud Infrastructure.

0:52OCI is the blazing fast platform for your infrastructure, database, application development, and AI needs, where you can run any workload in a high availability, consistently high performance environment, and spend less than you would with other clouds. How is it faster? OCI's block storage gives you more operations per second. Cheaper? OCI costs up to 50 % less for compute, 70 % less for storage, and 80 % less for networking. Better? In test after test, OCI customers report lower latency and higher bandwidth versus other clouds. This is a cloud built for AI and all your biggest workloads. Right now, with zero commitment, try OCI for free.

1:53Head to oracle.com slash IonAI. IonAI, all run together, E-Y-E-O-N-A-I. That's oracle.com slash IonAI. So I'm Steve Lucas. I've been in enterprise software for 30 years. It's terrifying to say, but 30 years. I got my start at Microsoft at a very young age. High school, it looks like. Yeah, a solid 52 years old and counting. But all I've done is software. And I love it. I love everything about enterprise software. I love the transformative nature of it. So that I love. I've led three companies as a CEO. So, Bumi, iSIMS, and Marketo. So, three of those organizations. And in doing that, prior to that, my gosh, I was president of platform group at SAP.

2:56Prior to that, I was the SVP of Force.com, that's Salesforce. So, I had a lot of what I call platform-specific roles in my career. but if you look at how I got to Boogie which we'll get there in a moment a lot of it has to do with every business I've run whether it's at a large company or as CEOs it's been a platform business of some sort and the SaaS oh yeah 100 % like I mean I would say I got involved in cloud like super early on at business objects like way back when they were a business intelligence and analytics company in 2000 and i i think it was three i don't i don't want to fear three years 2003 or four i launched crystalreports.com which was the world's first cloud native reporting solution that's over 20 years ago and at the time it felt super cutting edge and i was using in my my pitch to the board i was using cutting edge metaphor like we take pictures of data so with a digital camera take pictures and post them with Flickr if you remember analogy totally worked at the time so and Boomi was started in 2006 I want to say yeah Boomi was founded roughly 20 years ago and the founding of the company was based on a simple observation which is none of these major software vendors, not Oracle or SAP or Microsoft or even Salesforce at the time, want to talk to each other.

4:42No, their systems talk to each other. The data is not compatible. Well, that's not super helpful because every customer on the planet needs them to talk to each other. So Bumi solved that very painful problem of fragmentation and complexity by bringing an integration platform. So that being at the time, that was kind of mind-blowing. holy cow, I don't have to build all this myself. It was kind of like 40 years ago, or 40 years ago, companies actually wrote their own email servers or their own database as well, or 40 years ago. No one does that anymore. And similarly, there's no sense in building all your own integrations by hand, painstakingly.

5:20We have an integration problem. So that was the innovative thing we delivered 20 years ago. Yeah. And I'm going to ask you the tip, the thing, yeah. Yeah, that's better. Yeah. And what were they integrating 20 years ago? So, well, 20 years ago, I think conceptually a lot of the same things we see today, applications, databases, those were the things that needed to be connected in order to build business processes and improve business outcomes. And the whole notion was like you take the analogy of I get an order from a website and I need to process that order through an SAP system. And then I need to fulfill it through any logistics type of application.

6:05All of those things need to have the data move from system to system, from order management to fulfillment, all the way to shipping, logistics, et cetera. Well, that data or sets of data packages, most systems, like an SAP natively, they'll help you as long as it stays with an SAP. Oracle will help you as long as it stays with an Oracle, but good luck once it's kind of outside the boardwalk, so to speak. And that's where Boomi shines. we can literally connect everything to everything yeah and and the and i'm gonna uh do uh at the end we'll do a screen share of the platform and then i'll fold it and while we're talking for the video yeah um in those days there was you weren't handling uh ai applications or systems so how did that change first of all the supervised learning craze when everyone was labeling data i've worked with a couple of platforms one called label box that would take the data from your data store and then put it in front of the labelers with all these tools and then pass it along uh did you guys handle labeling or or did you have supervised learning workflows that you were working on yeah Oh, yeah, we sure did.

7:33I think this is worth 60 seconds of a walk down memory. Sure. So here we go. So, you know, in the beginning, whether it was on-premise or in the cloud, we had applications and databases. And various standards have come and gone for how do I connect these things? Everything from, you know, the old school days, already BC, you know, open database kind of compatibility protocol or communication protocol. So those things have come and gone. And then you have enterprise or system bus architectures and all listed other things that came along. Boomi has existed through all of those, especially the past 20 years kind of phases.

8:16Well, as we had applications proliferate and databases proliferate, then we had the API economy. Everybody kind of, in a good way, put APIs in front of everything. Well, in application, our API is pro-integrated. So now you have tons of apps, tons of databases, tons of APIs. And Boomi has evolved with all of that, where we've said, hey, our platform allows you to get your arms around all of your apps, your databases, and your APIs, and integrate those in a single layer. Now, why that's special and powerful is I'm not writing bespoke code. I'm not using bespoke connectors every time. I need to move data payloads between all these things.

8:59So that's the beautiful part. Now, to your point, well, then what about AI? First of all, AI is fueled by data. So that's irrespective of the airbag. So that's always been good for us. But even probably 10 years ago, we were building machine learning technologies to do things like automatically bring data, as you can imagine, when you do a table joy, something fairly straightforward, simple. You actually have to know which elements in each table match if you're going to create links to run a SQL query. So we had AI that would actually automatically link tables, even if they were from different databases, different applications, whatever that may be.

9:44So for us, we've been building machine learning type technologies for quite some time. We've also been collecting metadata about customer processes. So think about I'm a large beer manufacturer, right? So I make beer and I ship. Pretty simple. But the logistics of manufacturing process, not so simple. So I'm building thousand step processes at a time that require connecting many systems to many systems. And you're building these complex processes. So our technology, we were actually capturing and genericizing, it's anonymized, the metadata about those processes thinking someday it might be good to train AI on this data, right?

10:33That was 10 years ago. Today we have nearly 400 million unique processes that we've categorized, characterized that we're now training on. It makes sense now, right? So we did all that through the process. We've evolved. The platform has been rewritten, modernized multiple times to embrace new technology. And here we are today in this era of generative AI, pre-treened transformers, and the like. This was like a perfect world for UMI because AI and agents, which I know we'll talk about, it needs data from your apps and your databases and your APIs. But it needs context like models. We've even, and just one more thing on your question, we've even, over the time we built a data catalog product, we were part of the whole data catalog and how do you label your data process.

11:26I think all that stuff, and we're going to be part of that as well as it's going to be replaced by AI. Why would I want a human to do things like build an oncology for me manually? It's just AI for that. So yes, we've been there, done that. I should have said that. That would have been a much shorter answer to your question. Yes. Yeah. And even with AI, once you started integrating APIs into the platform, make the platform conversant with APIs, most of the big model providers are all going towards APIs. So how different is it? I'm not talking about integration of AI into the product, but in handling workflows that include AI.

12:15If you're hitting models by API, it doesn't really change your process or platform, does it? Well, no. I mean, I think that the lingua bronca for all AI, it's APIs, right? That's how they talk to each other. They talk to other systems. The beautiful part about what we offer when I say API management, what it means is you can take any system, hyper-modern, super-antequated, you can wrap a series of APIs around it and allow me to talk to you. Simple, right? This is a profound transformation, especially for legacy systems. There's a lot of banks that I talk to, but they have systems that are still sitting.

12:58They're literally, they were written in COBOL. Yeah. The guy that wrote it is on a beach somewhere in Florida, and there's no one around to support it. So how do you get a modern interface on a relatively ancient system, right? So things like agents can interact with that, you know, in a modern fashion. Well, that's what Boomi specialized in. We can connect to those legacy systems. We can then apply modern APIs to those systems and allow it to interact with the modern world. So kind of like what's old is new. So we can do that amongst many things. That is super critical in the world that we're talking about, right?

13:36If you want agents, modern AI, to interact with banking systems, you better be prepared to have an integration platform that can kind of do the translation. I'm sure someday there will be some capability where AI will be able to say, hey, just talk to it for me and it'll figure it out. But we're nowhere near that. Yeah. Yeah. Well, that's an interesting point. Are you working on that? Yes. Yeah, I would guess. I mean, the transformation, look, it's going to go from, you know, I have enterprise applications, databases, and APIs, and I need an integration layer. Absolutely. There will be an agent fabric, an agent orchestration layer where instead of me saying, well, I have all these systems of record and the like, and then I have an integration tier.

14:28And that will always exist for whatever character is like classic enterprise software. But there will be a day where I will just say to a swarm of AI agents that have expertise, like one agent will speak COBOL. Right. Another will speak ABAP from SAP, right? These will speak these things. We will be able to direct those agent swarms to go do things like connect to this system and produce this outcome for me. So I will describe my intent. And then our platform will orchestrate this swarm of agents to go do that for you. So less like, you know, connect this thing and it's very rigid and much more flexible.

15:09but I think it's just fabric. Yeah. And this question might be naive, but there is a class of companies called systems integrators that work to integrate different is, does this replace that or are they the customers for this sort of thing? Well, I think system integrators, I hope they will be the innovators for this kind of thing. Here's what's going to happen. I think with all system integrators. The system integrator today is defined as large consultancy with a specific set of expertise. It could be an industry or many or a technology at many, and then a ridiculous number of human consultants.

15:55There is not an SI out there that isn't already replacing consultants with agents. And by the way, they'll still bill by the app. Yeah. So that's still going to happen, right? It's just going to be an agent by the hour. So their profit margin, if you think about traditional consulting, why have software companies stayed away from having big consulting? Well, because the margins are fine, right? They are. So the margin high, human margin low. Well, what if the humans are consultants are replaced over time by agents? Well, now you've got a ridiculously high margin for an agent high. for cheese that's an irresistible proposition for si's they will most not will they are doing that yeah um but they could use they're not building their own platforms necessarily no well some some are building kind of these like agent factories that will produce lots of agents look i think every company like abumi we're building agents that are kind of in our wheelhouse We're an integration company.

16:58We build integration agents. We have an integration power. We have an agentic platform that allows our customers to build integration agents. So we build them. Our customers will build them. So every company is going to kind of be in their wheelhouse. However, what AI does is it gives you the ability to also look around and step outside your wheelhouse a little bit and do more. I think the biggest impact in the short to medium term across all of software, it's not going to be with the SIs. What it's going to be with is we're going to see the user interface layer that enterprise software companies like Salesforce, Oracle, SAP have dominated.

17:39They count on, they are used to you as a human logging in. You do things and then you log out. You won't do that anymore. You'll literally have an intelligent agent that knows who you are, your job, what you do, your company, and it'll say, hey, just FYI, three expense reports came in. I already approved them. Let me know if you need anything. That's the experience I think when I want. I don't want to, no offense, but I want to log into an expense report system and try and remember how to do that. Like, who cares, right? So that future we're heading for, the user interface layer is going to be wrested from all major enterprise software companies and replaced with AI.

18:24Yeah. You were talking about the consultation end of system integrators. Part of that is understanding the landscape of applications and configurations and data products and then advising a client, okay, you need this, this, and this. you guys are sort of agnostic right about about what's getting plugged in how do people know what they should plug in is is that uh up to the uh the cso or cto at a company or do you think that ai will take over that where you can ask maybe an agent but certainly a general model or maybe a fine-tuned model you know this is my business process what tools would be best to automate this workflow and then you know i can then use boomi i'll grab these tools or boomi on the platform the tools are listed right you're not You're not pulling in, but you can, I imagine, pull them in from outside.

19:44Well, I think that so I think companies that deliver capabilities like process discovery are in a uniquely good position. Right. So process discovery technology generally will do things like run against tons of system logs and figure out what is your actual business process. And then it will give you kind of a mind map, if you will, or process flow saying, well, here's your business process. It could be order to fulfillment and it'll discover that. That process discovery, those tools and technologies kind of become the prompts for AI in the enterprise. So you use process discovery to create prompts.

20:25Then from those prompts, you can ask AI, now make a better one. What should my process be? That's where, you know, us collecting hundreds of millions of indexed processes over the past decade across 23 ,000 companies. It gives us kind of a unique moat that we actually know we've already built an AI agent where I can actually say, create a fulfillment process for me. It looks across near half a billion processes and go, well, here's the best one. So it's going to be a unique thing where I think process discovery vendors will be kind of like the enterprise prompt vendors. They'll find it, share it.

21:07And then integration platforms, specifically Boomi, will be the optimized creators. Yeah. Yeah. You mentioned how many companies did you mention? 240 ,000 or what? 23 ,000. 23 ,000. I'm sorry. Yeah. This is part of the broader digital transformation of the economy. And a lot of people assume that's gone a lot faster than it has. How do you see that? I mean, how big is the market for Boomi? I know you guys are growing very fast. Well, yeah, we are. Well, so first of all, I think, I mean, digital transformation, as you and I know, is a loosely defined term, right? Originally, digital transformation was, you have paper, let's be digital.

22:07Now, digital transformation means something more than that. It means let's be modern. I think we're going to see that shift from digital transformation to agentic transformation. It will be the dominant term in the enterprise. Two years ago when I was talking about AI agents, I was met with a lot of eye bubble roles. Like, we'll see that in five years. But I think Sam Altman and his band of Mary, Women and Men have been proven right is that agents will transform the enterprise. It's beyond the sumptive now. So as we move from digital transformation, paper to digital, to agentic transformation, digital to agent driven, to me, what that really means is we're going to see the rise of the self-driving enterprise.

22:54And I don't mean fully, but you think of whatever, pick your electric car. You know, the reality is all electric cars started with kind of like super cruise control on the highway and then eventually stop on their own, light turns and go on their own. So we're seeing that evolve. This is exactly how enterprises will evolve with agents. We'll see them self-drive on the highway. But eventually you'll see an enterprise or a series of agents just build the income statement and balance sheet for you. why would I want humans doing that anyway? The error is a long list of things. You'll see things like the approval processes for soul-crushing work, like expense report approval, you can tell I don't like expense report, or, you know, hires in an HR system.

23:39That will all become self-driving aspects of AI within the enterprise. So that's exactly how this is going to evolve. So, you know, I just, I look at what we do is, all of these existing apps, databases, and APIs at this layer. And then above that are the agents working. But an individual agent, much like a human, it doesn't know how to connect to 100 different systems. That's where we come in. It doesn't know what this API is. That's where we come in. It doesn't have enterprise context. If I said to an agent, what's my forecast? Well, what do you mean? You mean sales forecast, retention, the weather?

24:24Well, what system is that in? Is it Clary? Is it Salesforce? Is it NetSuite? Is it Game? What is it, right? And so that context matters. That's where we come in. So we can give agents connectivity, so authentication, authorization, connectivity, context. That is the fuel for a successful enterprise right now. Yeah. And what level of enterprise are you targeting? What level are you targeting? Is this Fortune 100, Fortune 500, or anything making more than 100 million in annual revenue? Well, we'd certainly like those. But we have 23 ,000 customers all over the world speaking hundreds of languages with, you know, all different kinds of currencies.

25:21Our aim. Look, I would love to have Walmart as a customer. I would. But I view Boomi's job as something slightly different than that. Look, the average company, most companies, they're not going to be Walmart spending billions of dollars on AI research. Right. And look, the technology age already allowed Walmart to, you know, with hyperscale technologies to gut all their competitors. The road is littered, I think, unfortunately, with tens of thousands of mom and pop retail shops that couldn't compete with the Walmarts of the world. That was just the technology era. Now we're in AI. Walmart is spending, as I said, billions, forget billions, tens of billions of dollars on transforming their whole business, optimizing with AI.

26:10Who in retail is going to survive that? So I view our role as to be the great equalizer, to give companies that, and I'm not trying to vilify Walmart. No, of course. We give companies that are not Walmart the ability to run their business with AI hyper-efficiently. It's why we built our, not just the ability to connect all these systems, literally we are making available, generally available on March 21st, the ability for any of our customers with no coding expertise whatsoever to build their own AI agents. I can describe my intent, which I just talked about, and we create agents that do that.

26:49So if I describe intent as I want to run a promotion for any new customer that comes in through my website, we will literally build an agent that does that. And you go, well, hang on a second. I thought Boomy was an integration platform. Yes. And AI can do anything for it. Right? So I can describe an agent that I can say, hey, anytime I have a shipper that cancels an order, they can't deliver for me. automatically find another one that's a lower cost that meets my customer requirements we will build that agent for you no code you just describe your intent and we build the agent that's the great equalizer for everyone that's not on it yeah and so what's that product going to be called agent studio agent studio and we've got a long time ago yeah uh the the uh what's happening under the hood are you are you uh are you using other companies agent building platforms and who are you using and i assume you're using model context protocol and there are all these different protocols coming up i was just yeah yeah we're super buzzword compliant on that front so we have all kinds of interesting things.

28:11So first of all, we're actually using third party models, for example, like LAMA, but not limited to. And what we've done is we took models and we taught those models how to build agents. So it's models that build agents that use models, if you kind of get the circular reference, right? The second thing we trained them on are hundreds of millions of business processes that we, you know, over time. So they understand how these processes work. We trained them on things like how to interact with API, applications, the database, all using the existing Boomi platform. So when I describe intent, that prompt in English is passed to one of our models that we fine tune.

28:53And then those models produce what it believes to be the best agent for your business. Aware of the fact that you're already using the Boomi platform. So all that just happens automatically. When you do that, we then create the agent to put this in any business process you want. That's where I think the game-changing magic comes. A customer that built processes to connect a bunch of apps, now they go, holy cow, they're going to use that same work to create these agents that do anything. Which begs the question. I used to run, I was the CEO of Marketo, marketing automation. A very specific product that automates marketing, sends emails, runs a campaign, right?

29:38What if I could use Boomi to build an agent that does that? Do I need Marketo anymore? Those are the kind of questions that you ask. And by the way, yes, a ton of protocols coming out like model context protocol, which literally is like, I can give a model context for my application. That's going to be a huge player in the world of AI and agents. But I do have a little bit of a beef with MCP, and here it is. So Anthropic comes out with it in like November of 2024, right? Hey, any application vendor out there can give our model context about your application. Oh, well, thank you, Anthropic. That's truly altruistic, right?

30:21Think about it. What I just mentioned, which AI becomes the new UI. Well, this is an easy way for Anthropic to say, hey, just hang out in Claude. Use it all day and we'll talk to SAP for you. We'll handle the expense report. And then what makes me even more squint AI is then OpenAI. And Anthropic and OpenAI do not like each other. Then OpenAI says, we're going to support SAP. And then Google says we're going to support MCP, right? So all the major model players out there are supporting MCP. They're doing it because it aids the extraction or removal of the classic user interface and puts them atop it.

31:00If you think it's anything noble, it's not. It's selfish. And look, the toothpaste is out of the tube. This is going to happen. So we, of course, support MCP and will continue to do it. And you have other companies like Google comes out with agent-to-agent protocol. Look, the world, and Boomi's going to release a number of capabilities here as well. This is something we're just redefining right now in real time. We're literally redefining how processes, workflow, and more will work in the enterprise. So you're going to see hundreds of protocols come out. Yeah. And back on penetration, how is Boomi priced and how small a company can take advantage of this?

31:49We have companies with, you know, 25, 50 employees using Boomi. And by the way, our message to every single one of our customers, if you want to scale more efficiently in the future using AI, it doesn't mean fire humans. at all. It might mean higher or less going forward because of the efficiency of our platform. So our model is really simple. We have two very simple models. One is it's just an all-you-can-e platform model. So you literally get every product we make. And then just depending on how much data you move in and out of the platform, it's literally like a power reader. Use everything, go transform your business.

32:32The more of it you use, we make slightly more. Then we have a second model that's kind of more of a, what's called a connector-based model, which is if you have a limited number of systems that you want to connect to, that works as well. But it's super easy to do it out. In fact, you can actually go get a 90-day trial of the entire Ruby platform for nothing. Yeah. And both of these categories will have access to the Agent Studio. Of course. Yeah, it's included. So we're, you know, on May 24th, we will make, so our agent builder, which is part of the agent studio, we also have something called agent control tower.

33:13I'll get to that. Agent builder is create unlimited agents. All of our customers will be able to build unlimited AI agents. Now, deploying them costs money. You can deploy a certain number for free, and then if you want to travel beyond that. But you can build as many as you want, which is pretty cool because the experimentation is there. the reality and i think the question that begs to be asked is what happens when there are millions billions tens of billions of agents so we humans who have been the gods in this scenario we see much of that ground to ai yeah who's watching the ai when do humans get pulled that human in the loop kind of concept.