The Next Generation of Enterprise Integration: AI-First Platforms vs. Drag-and-Drop Legacy

Dan Magid 00:04

Okay, great. Well, welcome everybody. We are very excited. This is kind of a special webinar. It’s about what we think is not just the next generation of enterprise integration, but really the next revolution in computing. If you think about the kind of history of computing, we’ve had several revolutions. The original batch processing, and then we moved into interactive real time process processing. And then we did client server computing and personal computing, and then we did the Internet and we did the web and then we did mobile computing and then we did enterprise integration. And now we’ve got the AI revolution. And AI is really going to revolutionize how things are done. I think we won’t even realize the impact until later on and we see the results. But if you listen to all of the sort of industry leaders, they’re all talking about how this is going to really change how things work. And we want to make sure that our customers are able to take advantage of that. So we’re going to talk a little bit about why this is such an important thing to be paying attention to and we’re going to focus a little bit about in the integration space between kind of the most recent generation of how things have been done versus how things are done in the AI world. If you kind of look at the previous generations of integration, originally it was paper going back and forth and people doing data entry. So you’d have data entry on both sides of the transaction. So there’d be a lot of manual effort. And that was extremely error prone. And then we started to do batch file transfer. So as we started to move into things like EDI where you could then send a sort of a standard file out to people and then you could process it on each side. We moved into that way of doing things. The problem with that of course is it wasn’t real time and EDI became the sort of non standard standard where everybody customized it. And so it was very difficult to create an integration because they were all different. Every single partner did things differently. So we started to then go to hand coded integrations where we would actually write computer code. We’d actually write programs to take EDI documents and translate them into something we could use into our system, so into database updates, into program calls that we would write the code to do that. But that also was time consuming, took resources and you’d often get backlogged on trying to onboard new business partners because the integration process was so difficult. So we said, okay, so how can we make that easier? So a lot of tools came out that allowed you to do drag and drop integrations where you could say, I want to drag this set of fields and drop them into this other set of fields and I can drag the fields into, map the fields by dragging and dropping them into the, the appropriate places. And then under the COVID the system would, based on that, create the integration code or create the mapping code to allow you to move through that process. The problem with that is it was based on. The underlying infrastructure is pretty rigid. The vendor has created a particular way of interpreting what you’re doing when you’re dragging and dropping. And as things change, you have to wait for the vendor to update things in order to be able to take advantage of the latest technology, the newest thing that might have come out, or any kind of changes that may occur. So we’re going to focus today on the difference between that and what’s happening now with AI first integration. We talk about it as AI first and Aaron’s going to talk more about this. But it means starting from the idea that, okay, so now that I have AI, how am I going to do things? And this is, it’s so new. It’s hard to believe that it’s only been two years or so since the, the, the ChatGPT announcement, which really revolutionized the use of AI and really because it gave you the ability to use large language models that you could actually talk to the AI, you could, you could use regular language to talk to the, the AI. So the question becomes, okay, so now that I have this capability, how would I do things differently? So not just how would I kind of wrapper the old way of doing things in some AI, but starting from scratch, how would I do things? And that’s when we, when we say AI first, that’s what we mean. So we’re going to be talking about, and this is kind of be the focus of our webinar today, is what does it mean to really start from the point of view of, well, now that I have AI, how would I do this, what would I do differently? So we’re going to talk a little bit about the, the, the traditional integration tools of how things have been done up until, like I said, up until a couple years ago, how things have been done and what were some of the challenges of doing things that way. And again, those, those ways of doing things represented very large moves forward in how productive we could be and how easy it was to do things. But we’re going to talk a little bit about some of why there were challenges associated with them. And then what does it actually mean to do AI first integration. So from an architecture standpoint, what does that mean? And then how would you build things then, given that you’re working in an AI world and what does the flow in the real world look like of building and deploying AI integrations? And then Aaron is going to spend actually most of the time giving you a live demonstration showing you how things work in an AI first environment. And then we’ll talk a little bit about next step as we go through this. Please feel free to post questions in the Q and A section. So just go ahead and put in whatever question if, if, if we don’t have, and we’ll answer them via email. So with that, I’m going to go ahead and turn things over to Aaron and let Aaron start taking you through this. All right, great.

Aaron Magid 05:56

Well, thank you for that introduction, Dan. This is very exciting. All right, I’m going to share my screen and we’ll get going here.

Aaron Magid 06:16

All right, cool. I’m sorry, hold on one second. I’m a little too far. There we go. All right, sorry, one second. Okay. Apparently my machine doesn’t want to work for me. All right, that’s fine. Okay, so let’s jump in and let’s talk about, about what AI first means. So the first thing that I want to make sure to cover here is in order to understand what we mean when we’re talking about AI first integration platforms. The very first thing that we need to understand is where were we in the past and where are we going now? Right? So Dan talked a little bit about the history. You know, he talked about batch processing, he talked about, you know, paper processes, he talked about, you know, the, the ages of EDI and of drag and drop tools and a couple other phases, you know, that have come through over time. But fundamentally, the difference with this technology is we want to recognize that the new wave of technology, this new AI wave, is different than what has happened before and that it’s here to stay. And once we recognize that, we have to ask a very important question. And that question fundamentally is, if I was going to build something for the age of AI, given that I have AI now, now that I have these tools, how would I do it? Right? And one of the critical points that I want to cover in here is, and this is, this is broader than just the integration space, we want to make sure that with this new technology, we’re reimagining what we can do because there are a lot of pre existing ideas, there are a lot of traditional methods of doing things that have served us well. But are no longer appropriate in this era of this technology, meaning there are now better ways of doing things. And so we’re going to talk about that here. And that’s really the major distinction when we talk about traditional tools versus the AI first model. The difference is am I sticking with an existing model and basically just adding a little sparkle onto it for the AI, or am I actually looking at this top to bottom to figure out how can I do this, how can I solve this problem the best possible way in this era that we are in, right? Without all the predetermined notions and all the predetermined or all the, the previously known problems that, that we’ve run into in the past now the limitations that may or may not still be accurate. So with that, let’s talk about traditional tools before this technology existed, which frankly is not that long ago, right? As Dan mentioned, the original, your chat GPT kind of, you know, hit the scene for most of us about two and a half years ago. And it’s kind of ridiculous to me to think about how much the world has changed in every job in the last two and a half years. I don’t know that the world has ever experienced a disruption of this magnitude that moved this quickly. It’s, it’s insane. So when we look at any platform, any of the sort of household name platforms you talk about like you, we talk about like the, the, the boomies and the mule softs and the Zapiers and the, you know, those kinds of platforms, these are tools that were built and designed and had their ideologies defined before this technology existed, right? And that means that the platforms, the tools, the people, the ideologies are not starting, they don’t have this technology at its, at their core, right? So I’ll give you an example of that when we talk about drag and drop tools, right? One of the biggest limitations of those tools they have always had for their entire existence before AI and now is if you build a system where the way you build an integration or the way you configure something is by dragging and dropping components together, there’s a very simple question which is what happens if there’s no block for the thing I want to do, right? And in the age of modern technology, even before AI, just in the age of real time integrations, that was a huge problem for these platforms, right? What happens when a new technology comes out or what happens when a niche technology comes out that their developers don’t build a block for if that happens? There are two possibilities for me either I’m in a no code system and I’m just done. I mean, there’s, there’s nothing I could do about it. I’m locked into this very rigid platform and I can’t get out. And I basically just have to find another, another way of satisfying my business requirements or tell my executives that I just can’t do it because the platform’s just not capable of it. Or if I’m in a low code platform, they basically resolve that problem by saying, well, you can write some code, our proprietary format. And then what you end up with is an army of consultants who. What I’ve seen working with companies is you end up spending more on the consultants trying to customize and bludgeon the drag and drop tool into doing what you want than it would have costed you to just build the thing yourself. Right? And that problem, that problem has been there for the entire lifetime of these tools in every generation that they’ve existed in. Before it was drag and drop, it was, it was, you know, fourth generation languages, they had the same limitations. Before that it was case tools, they had the same problems, they’ve always had the same problems. And so when we talk to companies about this technology, what we usually see with these existing tools is as long as you stay within the very narrow bubble of what is expected, as long as you don’t create anything new, as long as you only do the things that the developers of your platform already thought of, right, it’s going to work great. You build systems very quickly, you get them out to production, it’s super easy. But the moment you want to step outside of that bubble, that’s when you start running into problems. And what we’re seeing is that in the age of real time integrations and AI technology is moving faster and faster and faster. And what’s happening is these platforms are having a harder and harder time keeping up. And so what happens is more and more frequently we end up hearing stories like my integration took weeks or months, it took way too long to get this thing out because we had to go into a whole custom development phase to get this thing done, or I had to spend a ridiculous amount of money with an army of consultants to try to get this thing done, to make this platform do what I needed it to do because I had a complex business requirement or one that we commonly see in the companies we talk to is I just had to tell my executive, sorry, can’t do it. My team, we just can’t do it. We’re just not capable of doing that thing that you want us to do because our platform just doesn’t allow it. So from there, some of these platforms have tried to address that to an extent by adding what, what I refer to as AI sparkles, right? Meaning basically, right, We’ve all seen that little icon, right? You know, a little, the little two diamonds icon that you know, every single application has now, right? And what that does is it’s companies coming in and saying, okay, well you know, what if I added an AI assistant to my drag and drop platform so that you can explain what you need in my drag and drop platform and then it will do the dragging and dropping for you, right? That’s what, that’s what platforms are doing. The problem is that technology is layered on top of the existing platform, right? It’s an AI that is dragging and dropping the blocks. But the fundamental limitation of the platforms isn’t that we’re too, you know, that we are unable to drag and drop things together. That’s not what we need help with. What we need help with is when the fundamental platform is not able to do the thing we need it to do, that’s where all the problems come in. And this kind of tool does not help with that because it’s built on top of the existing pipelines in the existing application. It’s only able to do the things that its tool set is, is able to do, right? So it’s really just a convenience tool. And that’s what we’re seeing with these kinds of applications that are, that are getting updated. So with that, let’s talk about what, what does this AI first model look like? What does this mean when I say AI first? Fundamentally, when I say AI first, in a nutshell, what I mean is that the AI is the primary portion of your integration system. It’s not that I have a drag and drop tool and then I tacked on some AI here and there and I, so that I could put the sparkle on there, you know, to make my invest happy, right? It’s not that, it’s the, it’s. We’re starting from the ground up with AI at the core and saying, in this era, what would I do if I was starting from scratch? Practically, that takes a couple of forms. Number one is the form of the subject matter expert, right? Number one is one of the critical differences between large language models. And what. Right. What everyone is nowadays, right. When everyone says AI, what they actually mean generally is large language model. But with these tools, one of the critical differences is that they can actually understand and reason about Information, Right? Meaning that I don’t necessarily have to be as reliant on human subject matter experts. I don’t necessarily need a person who understands architecture for software and who understands my application and my business rules. The system itself can actually understand that. So that’s the first piece is in an AI first system, the AI actually learns your environment. It actually learns what it needs to know in order to build integrations. Right. It actually becomes the subject matter expert. Right. In our system we go through that with a couple of phases, but this is the first phase that we’ve been going through with Aerodyne Connect’s AI first model, which is that we start by documenting the ecosystem and that’s an automated process, an AI system that’s going out and saying what is in this ecosystem, what applications are here, what interfaces do they have? How can I talk to these systems? Once it understands that, it’s then able to reason about business requirements, it’s then able to actually say, okay, you want an integration that will pull some customer data, for example, but you don’t know where the data is in your database. No problem. I know where it is, right. I have the documentation, I have that I have pre processed your entire system. I know where all this stuff is. I know how you do business. Right. I’m the subject matter expert speaking for the AI, right? And because of that, it’s able to generate the, the actual integration code. Like the drag and drop platform. It’s able to generate transformation logic to fit your standards. It’s able to generate monitoring code that will allow you to watch what’s going on and actually will allow it to watch what’s going on too. It can help you visualize your data, it can generate documentation, it can test your code, it can deploy things, it can get the whole system set up because it actually understands what needs to be done. Right? It’s not just dragging and dropping in a, in a limited system, it’s actually building from the ground up based on its understanding. And what we’re going to do to interface with it is going to be based on business requirements, natural language expressions of business requirements. This is a critical point because the system understands all the technical details because it knows how to reason about our applications. If I can express a business requirement, not a technical requirement, if I can express a business requirement, it can generate an integration to do what I need to satisfy the business requirement. And then what I need to do, what my role becomes is expressing the business requirement and then validating that it did what I want it to Do. One of the ways that I phrase this when I give presentations on AI technologies is these days everybody is an architect, right? Or everybody’s an executive, right? Meaning my job working with these tools is not to generate the code, it’s not even necessarily to check the code. My job in working with these tools is to understand the business requirement, express that business requirement in a way that the AI that is clear to the AI and, and then validate that the business requirement has been met. That’s my job, and then move on to the next one. And I can say from personal experience, this is one of the craziest things to me. I’ve been writing code since I was 10 years old and I basically don’t write code anymore because I use these tools. And what I do now is I express business requirements, I express architectural requirements, and my tools are building it all for me, right? So basically what’s happened is I’ve moved up, right? You move up a layer in the organization and that previous layer that you were at is now automated, right? That’s, that’s really where this, this technology is going. And that’s what the AI first model does. It allows us to operate at a higher level than we were operating at before. Instead of having our hands in the tedium of, you know, writing test cases and making sure that everything is, you know, that, that we got, you know, exactly each line of code working correctly and the tool can do that for me. So I can just focus on the high level things that I actually need to be doing. I’m providing business value.

Aaron Magid 21:44

So hopefully at this point I’ve made it clear what this AI first thing is and why it’s so powerful. I want to talk about some of the technical requirements to make this happen, right? Some of the things that we want to do, some of the do’s and don’ts for this technology from what we’ve seen, and this is coming from, by the way, working with many, many companies on integrations. We’ve been watching these AI tools evolve for a long time and we’ve gone through a lot of iterations of how to use them and we will go through more, right? I mean, this technology is changing almost daily. You know, I guarantee you there will be more to come in this process. The critical things that we need, right, the most important pieces are we need our system to understand our ecosystem, right? If it doesn’t understand the ecosystem, it’s not actually that it can’t generate integrations, it’s actually worse if it doesn’t understand something, it Will hallucinate.

Dan Magid 22:53

Right?

Aaron Magid 22:53

That’s what LLMs do. If they don’t have the information that they need to answer a question, they generally won’t say, sorry, I don’t know the answer to that. They’ll make one up.

Dan Magid 23:02

Right.

Aaron Magid 23:03

So first thing is, we need to make sure that it understands our systems and we need to make sure it understands them clearly, because if it doesn’t, it’s going to start making stuff up. We want it to generate clean industry standard code. Why do we want that? We want that because that gives us a couple of things. It gives us automated tools that can help us check it for accuracy and it removes the limitations of clunky drag and drop tools. It removes that limitation of, oh, you’ve just got this big block that you can put in there to do something. Well, no, actually I can do that with a slight tweak because I have access to the underlying code. I can change whatever I want in this application. Another key point is we want to make sure that the AI is driving the process. The AI is not just answering a question here and there. It is not just, you know, one sparkle that’s in a little box somewhere. It is actually owning and driving it. We are saying to it, I need an integration that does this and it is then planning and executing and testing on that. Right? That’s a critical piece because any part that it doesn’t have control over becomes a part that it can’t account for in its stages, right? AI tools are able now to loop back and review what they did based on a future stage, right? So you can have an AI that’s able to generate an integration. This is what we do, actually generate an integration and then go back and test it, and then take that testing feedback and go back and generate more if it needs to, right? If the testing isn’t owned by the AI, you can’t do that process. So that’s a critical piece. We also want to make sure that we are generating a production ready, portable application, right? We’re not throwing this up into a SaaS service that we don’t control. We’re generating a package that we can then automatically deploy onto our infrastructure. So we control it. We have all the controls on it, right? We, we know everything that’s going on with it and we own it, right? We own the product of what we generated, right? So if that SAS platform, you know, that we might have used in the past, that drag and drop tool, decides to raise their rates or if we decide to cancel our license with them, we don’t instantly lose everything we’ve ever done, right? We don’t get locked out, right. We’ve actually generated something that’s valuable and what that’s going to do for us is it takes away that lock in. It takes away the limitations, it takes away those missing capabilities that we run into in these applications. And it also takes away the need, because of those missing capabilities, takes away the need to bring in external resources and pay a whole bunch of money to build custom things in really frankly, what is the most expensive way to do it to, to do that? So from here, right, there’s, there’s a couple things actually that, that, that I wanted to mention. You know, I mean we work with a lot of companies. We work with companies that are at all phases of, of the history of, of integrations, right. Work with companies that are, that are, you know, on batch processes. We work with companies that are on hand coded things that are on drag and drop tools and you know, and, and again we’re, we’re working with people to try to pull them into this era, right, to take advantage of these, of these technologies. And it’s important to note that there is a shift here, there, there is a difference in, in how these tools are going to work, but hopefully we’ll see when I go in and use the tools in a minute that it’s extremely powerful. I just wanted to make a comment here about the, my point about the production ready system of building an application that we can deploy. Right. And I want to, I want to really explore that with an example. Right. So I was working with a group recently that needed to build a complex integration that went through multiple processing steps. It’s basically a process of reaching out to their internal systems, grabbing some data, then reaching out to a separate internal system and, and talking to their internal application to enrich that data and then going through some business logic steps and then sending a notification out to their users. It was actually a text messaging based integration, right. And they had tried doing this with one of those drag and drop platforms. The problems that they ran into, there’s two of them actually three actually. It’s the first one that I was going to skip over is the architecture, right. In order to build it, they actually had to have somebody on their side who understood the here is the right way to do this. This is the way to build a clean, solid integration, right. I have to actually know how to do that. I have to actually know with the drag and drop tool what blocks do I need, how do I build a robust Integration that isn’t just a very simple brittle thing. Right. So that’s the first piece. The second piece with the other platforms is they had to use pre built widgets. And the problem they ran into was twofold. Number one, there was no pre built widget for their internal company systems. Those homegrown applications, drag and drop platforms, aren’t going to have anything for that. So they had to build a whole custom connection system to allow the drag and drop system to reach into their internal application and actually get what it needed out of it. Right. Which ended up being larger than the actual integration that we wanted to build. And the other piece is it’s all on a proprietary runtime, which means that if they were to ever leave that platform, they lose everything. Everything that they’ve built in that platform is actually, they don’t own it, it’s actually owned by the platform. And if they ever want to leave it, because, for example, it’s incredibly limiting, they can’t, or if they do, they have to start over from scratch. Right. So we came in and we talked with them and what we did is we provided the information about their applications, we gave our AI access to their applications so that it could reason about their systems. And it then proposed a clean architecture for their integration that did not require them to understand how to do it. It then generated the application that they needed to do exactly what they need without any drag and drop limitations, without any of those pieces. And, and it built it out into an application image that they could just drop onto one of their servers with an automated deployment system that runs it on infrastructure that they control so they don’t have to worry, you know, one day if that drag and drop system goes away, are they going to lose everything? They actually own it. They actually have the thing that they generated. Right. What we’re seeing with these strategies is that these integrations are becoming more effective, they’re rolling out much faster, and they’re frankly more cost effective because they don’t have to work within the confines of the previous models. And I do just want to make a point that it’s not that previous tools, that the drag and drop tools are poorly architected. They’re not. They were architected very well. They are, they are frankly in many cases, you know, the sort of very elegant applications that are extremely powerful. They’re just from a previous era. And so the new technology that we have sets the floor above the level, the base level of the tools that we have now are above the level that those tools can reach. And so we need a new architecture. And that’s, that’s really what we’re, what we’re out here doing, what we’re, what we’re proposing here. I do want to take a point here before I jump in and show this stuff to talk about documentation and why this is so important, right? And this is, this is especially important for companies that are in danger of losing knowledge about their applications. This is something that I see all the time, right. I work with a lot of companies, especially you know, with IBM I companies where either I hear the story all the time, either I come in to talk to a company and the person who knows their system is, is gone. Right? Nobody knows how the system works, right. I had one of those, I think actually earlier this week, right. I hear this all the time. So, so I go into a company, they say, we lost the person who understands the application. How are we going to build these integrations? How are we possibly going to tie these things together, right? The other piece that I see is, well, the person is still here, but they desperately want to leave. You got a person who has been trying to retire for years, but they can’t because the company doesn’t know how to function without them because they’re the only person who understands the application, right? And one of the powerful things here is that if our AI model becomes the subject matter expert, if we are able to move the knowledge into the tools themselves so that the application can tell you what the application does, then we don’t have those limitations anymore. We don’t have to worry about that as much. And this is a critical thing, especially if you’re concerned about losing your application. Knowledge is a critical thing to do this preferably before you lose the knowledge. We do work with companies to help them build integrations after they’ve lost the knowledge. But it’s, it is definitely easier if you still have that knowledge in house and you can basically train your new AI subject matter expert on your applications while the people, while the human knowledge is still there, right? And one other critical point, and then I’ll get in and show you some stuff, is that the documentation that we’re going to generate, true to that AI first model, the documentation that we’re going to generate is optimized for the AI, meaning it’s human readable, but it’s not actually meant for people. It’s actually meant for the AI. It’s meant to be traversed by the AI. It’s meant to provide the AI with a handbook for everything that it needs to talk to be able to reason about your applications. So we’ll jump in and we’ll, and we’ll, we’ll show that. So I’m going to jump in, I’m going to show this. What we’re going to do is we’re going to generate an integration with a tool that understands the ecosystem, that can generate from natural language, from business requirements, from high level prompts, and can also set up metrics and, and visualizations on those. Right. And get the entire thing basically functioning as the integration expert. Right. And that’s, that’s the key thing.

Dan Magid 34:49

Right.

Aaron Magid 34:50

I’m just going to provide business requirements. The AI is actually going to do all the hard work for me. It is my integration specialist. Okay, let’s talk about our scenario. So I have an, I have a problem here. I have a database in my application that has all my customer data in it. And when a customer calls support, I need my agent to be able to pull up their information. I need them to be able to pull up, let’s say, their contact information. And right now that data is scattered throughout my backend database. But I’m not sure where. I know that it’s in there. I mean, it’s got to be right somewhere. I’ve got an application that’s reading it, but I don’t actually know where it is. And the person who knew is gone. Right. And this is, you know, a real scenario. I just want to point out there’s a real scenario that I have actually seen based on real things that I’ve seen companies go through. So let’s go through this. So what I’m going to do here is I’m first going to pull up my workbench and what I’m going to do is I’m going to generate a base API. Right? So the first thing that I’m going to do here is I’m going to create a base integration.

Aaron Magid 36:14

Okay. I’m going to give it a name, say API, demo, contacts, details or something like that. We’ll just skip through this, give it a name, get contact details and say generate. Okay? So this is going to go through and it’s going to generate a base integration for me. Right? Now what that’s going to do is it’s going to get me started, it’s going to build out that platform for me, to give me what I need, that foundation for my application. Right. And you can see here, right, Remember those points that I mentioned? We’re generating industry standard clean code now. I don’t have to worry about that. I can see it if I want it, but I don’t have to worry about it because the AI is actually going to handle the entire process for me. Now from here what I’m going to do is I’m going to ask it to make a change. I’m going to give it a high level business requirement for what I want it to do. So I’m going to go over to my assistant here and I’m going to say, right here, I’m going to say I generated an integration at API Demo Contacts, details. Can you set it to get the relevant information from the database for a support engineer to contact a customer? Right. Notice this is a business requirement. Right? This is not a technical requirement. I did not say where the data is, I did not say how I want it returned, I did not say how I want to monitor it. I didn’t say any of that. I just said, hey, I’ve got a problem. I need to be able to get the data for customer contact information for my support people. That’s all I told it. Now as it’s thinking, you can see it going by, what it’s doing is over here, you can see it’s actually pulling up documentation. It’s got this, this customers table docs. This is part of the generated documentation for my application. And if I open that up, we can see that there is actually a manual in here that has generated documentation for my entire system, right? And it says, okay, so here’s your customer master table, right? This is where, this is where it is, this is what it does. Here’s how you can find it, here’s what its purpose is, here’s how its columns are structured and what’s in it. And then if I scroll down a little bit farther, it’s going to have the related tables and programs in there. Because before I went in and did this, I ran our documentation system over it, over my application and it walked my entire application, my whole database, all of my programs, all of my interfaces, everything that my system does. And it figured out how my system works, right? That process, that, that process, actually that’s a multi pass process. That process is going through first module by module and generating everything that I need per module. And then it’s going back and it’s then relating everything to each other and making sure that all of these components are documented correctly in their context. Right? And then it’s going through and it’s actually producing completed documentation of the system. Right? And again, remember, it’s not structured for people. This documentation is structured for the AI. It’s structured so that my assistant is able to access what it needs. It’s able to actually grab things in real time as it encounters objects in my system so that it can grab just the information it needs. Right. And this is a, this is also an important point. Right. Most real applications are too large to put into context all at once into an AI. Right. You know, even if you’re using the larger context windows, you’re not going to be able to, you know, bring the entire documentation of your system together, you know, all at once. Right. So we’re going to need to have this set up in a way where, you know, where it can actually look things up in real time. And that’s what the system is doing. So you can watch it as it’s going through and it’s generating the controller logic, it’s generating validation code, it’s deciding.

Dan Magid 40:49

Yeah, before you jump off the documentation thing there, there actually is a question in the, in the, the Q and A section that says can the documentation output include BRDs, flowcharts, data connectors and flows, application help forms and user training documentation? If so, how does that work?

Aaron Magid 41:07

That’s a really good question. You’re preempting me a little bit here actually. It’s going to do that. So yes, it absolutely can. Yeah. That is one of the things that we do. If you want it to generate human readable documentation, you can do that. It actually will generate a flowchart when it gets to the end of this process. It takes a couple of minutes to generate everything. Because I asked to do, I asked it to do a lot. But, but yes, to answer your question as to how it works, it’s going to generate, we have basically a multi pass process, an intelligent process on the documentation. So we have a documentation process that goes through. It’s going to generate, it’s going to look at things individually and in context to try to understand how everything works. And basically by setting that zoom level differently. Right. Is how it can generate flowcharts and data flows. Depends on what scope are you looking at. Right. Am I looking at one segment of code? Am I looking at one integration? Am I looking at one application or one database? Right. What’s, what am I actually looking at there? So that’s, that is a, that is a really good point. It’s actually, if I can just stretch this out by, by 10 seconds. It’s, it’s actually generating the documentation right now for my integration and I think it’s working on the flowchart right now. So this is really Good timing.

Dan Magid 42:35

So there’s another question out here. So if you have a minute. With AI technology changing rapidly, how do you protect your platform and the customer’s investment from quickly becoming an irrelevant or a legacy solution?

Aaron Magid 42:50

That is a very, very good question. So the, so the first thing, what I like to say is nowadays every application is always legacy because by the time you get it updated, the technology has moved so far that it’s just, you need to update it again. So how do we, how do we protect it? One is by, by staying flexible. This is one of the main reasons why we do the industry standard code generation, not the drag and drop. Right. What’s slowing down? If you, if you, if you have any drag and drop platforms, you have probably noticed if you’re actively building things, that they are not keeping up with this technology. And the reason they’re not keeping up with this technology, not truly, is because they have so much work to do to actually catch up. It’s a completely different paradigm. They have to rewrite their entire platforms in order to take advantage of this. Right. And so what’s happening is they’re becoming irrelevant or legacy. What our platform does is it generates the application as clean industry standard code and it also maintains the high level business rules separately. Right. So there’s two pieces there. One is that the code is generally going to be able to keep up with most of what’s going to happen. I believe there will be future shifts. I don’t know what they are. And anyone who tells you they know what they are is making stuff up because nobody knows what’s going to happen. So the other piece that we’re doing here is we’re maintaining documentation of the business rules of the, of the high level functions, what we’re doing in the application and in an AI that understands them. So that when the next revolution of, you know, having a quantum computer plugged into your brain where you, you know, just think thoughts to GPT and it does stuff for you when that comes out, whatever that is. Right. We can take that abstract understanding of what you’re doing and map it onto the new paradigm. Right. We’re, we’re actually not suggesting that this way of doing things will be the end all, be all. What we’re, what we’re trying to do is say what we’re doing is generating it in as future proof of a way as possible and then making sure that this platform is ready to move to the next one. Right. We’re not trying to lock people in, we’re not trying to stick your applications in here to where you can never get out. We’re actually trying to facilitate your moves to the next technology because we’re gonna, we’re gonna aim to provide that next technology whenever it comes out and, and help keep you up to date. Right. Rather than trying to sort of stick you into our existing model and kind of, you know, chain you there forever, which is what I’ve seen a lot of companies run into with, with, with traditional platforms. So I, I see that I’m, I’m overtime here, which I know is characteristic of me. So I just really quickly before I, before I sign off here, I just want to run that API just really quickly so that I didn’t do that whole thing and just kind of, you know, have to go on, on, on faith that it actually works. So I’m going to actually call my API and, and I just want you guys to see that. I hit my endpoint and it came back and it actually gave me the information about this customer. It actually worked and it pulled it together from multiple different places. This is pulling from a database that I happen to know has, has I think 10 different customer data tables in it that are all interacting in different ways. And it’s got a whole documentation system where it knows, okay, here’s how, here’s how account statuses work, here’s how the accounts table works, here’s how the high level database is mapped out. Right. This is all generated documentation that the assistant generated for itself to allow it to reason about the system and generate integrations. So hopefully that’s useful, I guess. One other side note that all of this stuff is monitorable with the same standards that we’ve seen in all of our other webinars. And I hope that’s useful. I hope that it gives you some, some ideas for how you can keep your technology moving forward because frankly, I think this era is all about speed. You know, we, we have to keep moving forward because it’s, it, it’s constantly, the technology is not slowing down. So I hope that was useful. Thank you all for listening. I want to turn it back to Dan, to, to close it out.

Dan Magid 47:24

Aaron, can you just, just in a couple sentences, so, so just summarize what did you just do?

Aaron Magid 47:30

Right, so, right. Thank you. So just to summarize here what we did. I have an AI assistant where my tooling was able to understand my ecosystem, my entire application and all of the surrounding systems.

Dan Magid 47:45

And so was that an AI process?

Aaron Magid 47:47

That is, yes, that is an automated process. So I have a pre existing application that process takes more than I could show in the webinar. So I ran that process last night. It went over my whole application and documented out the entire thing. All the programs, functions, databases, resources, servers, everything. Got the whole thing documented out. And now my AI understands my ecosystem. And so if I give it a high level prompt like give me the customer data, it’s able to actually go in and do that because it knows how my whole system works. And it has a platform, has a foundation to build on where it can actually deploy integrations onto my infrastructure and get them all out really fully autonomously. That’s. That’s really what we’re.

Dan Magid 48:34

What.

Aaron Magid 48:34

What we saw. That’s what I did there.

Dan Magid 48:38

Right. All right, well, thanks, Eric. Appreciate it. And thank you everybody for. For hanging out with us. Sorry we went a little bit over. If you have other questions, do, feel free to. To shoot us an email or, you know, or give us a call, whatever. And we would be happy to get on and talk about this a little more and go into a deeper dive on some of the things that we can do.

Aaron Magid 48:59

Yep. Thanks for coming, everyone. Look out for an email from me with that recording. Hopefully we will see you next time.

Aaron Magid 49:08

All right, thank you all. Thanks. Have a good one.