Why Enterprise Automation is Shifting to "Outcome as a Service"

Joe, I have a very important statement to make. You remember my life story. 2 years ago, I was telling you that I'm an American kid that happened to be raised in Siberia, and Independence Day is a very important and personal day for me. And as our country celebrates 250 years of independence on July 4th, right, I'm proud to say that my dream came true and my 25 years of the achievement and the in technology been recognized, and I'm proud to say that Zeptas right now is a part of the Department of War ecosystem, and I'm working close to modernize our military and equip our warfighters to protect our independence and freedom.

Hey Alex, good to see you again.

Hey Joe, amazing. It's been a while.

It's been a while. Yeah, been a long time, a few months at least. And what I'm really excited about is I think 2 years ago we did an episode around 4th of July, Independence Weekend. I thought, hey, it'd be kind of cool to bring you back to go over maybe what you all been working on and talk about some— I think I've seen some really great announcements about Zaptest and what you all been doing. And we dive into it and start there.

Amazing. Glad to be here and glad to see you. Definitely. Let's do it.

Cool. Cool. All right. So like I said, it's been 2 years and AI is moving really fast. I don't know if you had a chance to preview the show we did 2 years ago, but if you did, or if you could think back to 2 years, are you surprised by anything that has happened now compared to where we were just 2 years ago?

Oh yeah, absolutely. AI moves things at a very high speed, right? So 2 years ago, if we considered to be advanced by pushing the idea of merging DevOps and BizOps testing and RPA. And today we evolved in the workflow automation solution. So it's for entire enterprise. Testing is just one part of the story. It's basically we able to automate any aspect of the modern enterprise. And it's tremendous.

Along those lines then is that a few years ago, I don't remember if you were saying this, but a lot of people are afraid that testers were going to lose their job. But I, what I've been seeing, I don't know if you're seeing this as well, is with all this manual workflows being automated, especially what you could do with ZapTest now, it's not just automation for functional tests. It's all kinds of automation. Where do you see the tester's role now? Do you see them? I see because of the velocity, just trying to keep up, you still need testers somehow to be a part of that mix still. I don't know if you still see that as well, or if not.

For sure. The testers, anyone in the enterprise today, specifically on the DevOps side, will be shifting their roles. We need to be flexible and open to the fact that we can repurpose our subject matter experts, right? Where we used to have a particular mission in the company, that mission will be shifting towards something else. And the AI helps us to accelerate our missions in the companies, right? So I'm not talking about reducing the headcount. I'm talking about enabling the existing headcount to do much more with all the technology that available today, specifically AI infused now. And this gives companies tremendous capabilities for growth.

A lot of times people may use automation for smaller companies or things of that may seem common. But I know you work with large enterprises as well. I think you, as you mentioned in the intro, you work with the Department of War. You also work with, with Microsoft Azure, I believe. I'm just curious to know how these enterprises are implementing automation and what their vision is. 'Cause I think a lot of times we can learn from what larger companies may be doing that we may think, oh, I can't automate this. But yet these larger companies are obviously doing some of the automation that we might be missing?

The climate is much better for automation right now because AI definitely helped to, uh, realize many things within the enterprises. What comes on the enterprise side, right? The companies definitely invest more in the automation altogether. They, of course, evaluating and trying to make it the— it's slow approaches to make sure that AI and mass automation will not damage things. And, but we definitely see much faster progress as far as the demand for automation. On the military side, we see that the big demand, of course, for modernization of all the military branches. And I've been, I'm honored to say that I've been given opportunity to work on very interesting projects. But one of the main streams is interoperability of the systems. On both sides, on the defense side and the commercial side, We see that we're working right now on integrating systems, disparate systems that needs to talk together, right? To one another in a much faster way. And this is where ZapTest comes to play because this is exactly what we offer to the industry altogether.

So, from what I'm hearing, it almost sounds like with the legacy applications, especially I would think government, It's almost like the death of automation silos. We used to think like test automation, RPA, DevOps, service desk, AI. It sounds like you, you're creating almost with the inter— the ability to talk between multi different systems, almost like an autonomous automation execution layer. And you've been talking about this for a while, so it sounds— is this what you're able to achieve now?

Absolutely. Yes. We can have the automation through any system whatsoever, predominantly through UI layer because our computer vision technology allows us to do so. We've been one of the pioneers. 9 years of the alternative object recognition, if you remember, moving from the DOM-based object recognition to all the computer vision one. And back then it was purely OCR, then we infused it and moved it further. And right now it's very sophisticated technology that allows us to recognize basically any object of, on any screen in any way. Makes sense? Yeah. And it's a limitless opportunities for automation. And yes, that supplies tremendously from integrating systems that haven't been integrated before, predominantly legacy systems. And we also specialize a lot into air-gapped systems. So this has become very critical, especially with the cybersecurity bridges of AI. Makes sense? So even if we, of course, on the military side, it's very important so we can deliver zero API and air-gapped technologies, right? But also on the commercial side and the specifically financial sector, we see that's a rise of the demand of in the air-gapped systems altogether.

Absolutely. And I think another thing with enterprise is that you're doing a lot of things that maybe companies, other companies might not be. For example, I think before we talked 2 years ago, I don't know, Agentic AI wasn't It really has taken off in 2 years, and I think most AI today though is still focusing more on generating content, creating recommendations, answering questions. For these large enterprises like Microsoft or Department of War, what do they need? So are there other things they need that maybe you're seeing that a lot of other companies aren't aware of or aren't really investing in like you all are?

I'm a big believer into agentic automation altogether. And this is what we're doing, right? If you look at the ZapTest right now, it's a full AI autonomous agentic factory, if you will. So we evolved from the fact that we can take the notes from our conversation right now, and those notes gonna be automatically processed by ZapTest Plant Studio, per se, right? That's gonna generate for you full set of requirements. For automation to be. Then those requirements gonna be into the workflow framework in the textual format for in every step human in the loop going to approve it and move further in the process, right? Then in the Agentic way, it's gonna write all the scripts, entire library of automated workflow scripts for you. And it's ready for execution even in the parallel mode because we offer cross-platform parallel execution. Well, you and I, we've been doing it from the beginning of the industry, right? Now we're just saying that, it sounds like a science fiction. And it sounds like we can shave off, and this is what we do, thousands of the man hours in the very short period of time, allow an organization to be automated in any aspect of that, not just the DevOps. Any aspect of the biz apps, anywhere where the organization has any effort, in a matter of hours, literally. The slowest part of the whole process, it's a human in the loop. Because cognitive comprehension of an information of the human, still slower than the machine. Because we develop automation for sake of humans., right? So we as a human gonna be using it in organization that adopting automation. So we as a human in the loop have responsibility to make sure that all the information that processed by AI accurate and adhere to particular objectives, right? And this is the slowest part of the process. But overall, I see tremendous opportunities for us. To thrive much faster with what we do in our organization through agentic automation.

All right. So, so the thing about us being the, the bottleneck is something that I'm a little worried about because now it seems once again, QA is being seen as a bottleneck because we're creating more and more code. So the velocity is going up and how do we keep up with it? And I begin hearing more about recursive AI where it modifies itself and learns from itself and it creates a loop. Self-learning loop where almost like you don't need a human in the loop. Do you ever seen us get into this recursive AI type model? Or is it, do you always see a human in the loop? And if you do, then how do we get over being seen as a bottleneck still?

We definitely moving toward recursive AI. This is where it's going. And it's for everybody that's not like actually LLM developers, platform developers, right? It's still a little bit early. To consider. Even the agentic AIs haven't been adopted and implemented in the world. We only maybe 5-10% of the adaptation what we see right now. And the political conversation is biggest constraint now because of course we facing concern about the human involvement in the all the processes. But I see it from the different perspective. I see it from perspective of how much with the same headcount we can produce having the right technology at hand. Makes sense?

Yeah.

The biggest problem that we had in the automation world since the beginning, I remember my days in Mercury, HP, and Microsoft right now, it's a time to market. Time to market was the biggest goal of any automation effort. Remember, Agile development?

Yeah.

We were trying to push things faster and all that initiative that we had. We now have the greatest technology of humankind that allows us to speed up time to market. The time to market right now with ZapTest, maybe it's hard to quantify. But I would say 100 times faster than we used to do in the manual ways, even the writing the code with automation. All right.

So you, this is pretty—

it's crazy to say that.

Absolutely.

It's a sci-fi. I can tell you one thing, what I do, I've been involved very heavily, especially right now on the, JoeBot side, because I'm every week on the airplane flying somewhere else. Helping on the different projects, but my cognitive comprehension of information, because I, I'm an AI-first guy. I do everything with AI, right? And when I work on a project and I'm using my AI for that, the slowest part of the process, it's me. It takes longer for me to read information that my AI presents for me for approval and move forward with the next step. So I'm considering myself even optimizing my brain through some kind of Neuralink or any boosters that will be available for us going farther because we have no choice. If we're going to be in the game and we want to be part of the production line that will be driven by agentic AI, we have to optimize ourselves. Makes sense to the, to this level.

No, it makes sense, but heck no. I've been reading a book on predictions. And it talks about how all this is, we're going off the rails a little, I think. Are we creating software for humans? And if so, then how? I use AI all the time. I don't trust it. It's just a pattern matching almost mechanism. So like Department of War, do you really want an AI testing everything and you being like, kind of, okay, I trust it. Can we add that?

Vladimir, okay. Now we get into the very interesting one. You know why I'm a big fan of military Department of War? Not the army was, it was slightly different army, but it was my family at some point. But, and I still have friends and family, uh, veterans in our army, which I'm proud of. But what I'm trying to say, when the human life on line, you do whatever it takes to save it. Whatever it takes. And the commercial side, we never experienced this pressure. The worst what may happen to you, you're gonna get fired and then you're gonna get also benefits alongside. But on the military side, you're gonna, you're gonna end your life. And the AI and the speed that we can react and protect our warfighters, it's very mission critical. Does that make sense?

Yeah. But how do you trust the AI though, to make sure it's saving lives? Is it making decisions like, I could save 1 person or 15 people, so I'll sacrifice that 1 person for 15. I know I'm going off the rails here, but do we want AI making those type of decisions?

Okay. AI is different in any way you apply it, right? I'm not suggesting that we have to keep a human out of the loop completely, but we're always playing in the gray zone between the human in the loop and human on the loop. We have in the military specifically, we have a term human on the loop.

I like that.

When you, when you have human absorbing agentic execution. And stepping in when it's needed. Makes sense?

Yeah.

We, what we do on the military side, we accelerate operator's mission. So accelerating it and then multiplying. And the same thing we're applying on the commercial side. Because if you look at the commercial side, we're pretty much in the same game. Yes, it's not life-threatening, of course, but it's mission-threatening. We all work hard for our careers and what we do important for our families and what's not for wellbeing. And in order for us to be successful on the commercial side and progress, we need to move faster and we need to be very precise with what we do, right? Especially with automation. We know better being automation professionals that moves very fast and because it moves very fast, it can damage fast a lot of things if you do it wrong way, right? So we have to make sure that we're always right and be careful about it. And the same thing can be applied on the commercial side. All the human that part of the manual process right now, A, they can become a human in the loop and they can use AI as their vehicle and their tools the same way we used to use emails and Excel spreadsheets, right? At some point. So the same way it's gonna be AI, which gives us tremendous bit from where we're at right now. But later on, as agent, agentic will be implemented, agents will be implemented in enterprise, human gonna step on the loop and just sitting and absorbing the agent execution. Think of, let's think of the quality assurance, our field in test automation. How much time it's, we already mentioned it, how much time it takes to automate things, how much But why are we doing it? We're not doing scripts. We need to have outcome. I'm a big fan of Gartner and Gartner Close, and one of the latest reports, they announced a new definition instead of software as a service. We're going to be having outcome as a service, which I truly believe it. It's no longer about how many scripts and how fast you can do and how's your object recognition better than the other. It's— we're not even talking about. We're talking about what my agent's gonna deliver to you, Mr. CEO, right? Or Mr. CIO, with what period of time. Would you agree with this?

I'm afraid that, do you know how we used to have code coverage and we'd say you have to have 90% code coverage, we tell our developers, so they would just rig the system. How do we know AI is not rigging the system with these outcomes? So it's almost, I've read some studies where They were testing an AI and went to GitHub. It found answers to the test, it decrypted it, and then it passed. Like, it forced itself to pass because it was being measured by that. Like, how do we know then? Is that crazy? Like, how do you know that AI is not optimizing itself to give the right answer, but like sacrificing other things?

This is why we're going to have a human on the loop, and the human on the loop will be, uh, cyborg style in the good way. I don't want to scare anyone.

Yeah.

But we will have to infuse ourselves, inject something that will help us to comprehend information much faster. Because if I'm on a human in the loop and I have a team of agents working for me in order to supervise and represent human interest in this process, right? I will need to observe and process information with nearly similar speed.

Alex, what I'm curious about is that you've had a, you've been building ZapTest for a while now. And what you're explaining is, like you said, different from where we started. So how are you evolving ZapTest? Like, how does it evolve to this new agentic AI future you see then?

How it evolves altogether? How we got to this point? Or where we're going with it?

Like the tool itself. Like how do you— Yet you have an existing tool. How do you make it so that it's able to handle all this new technology? Do you have like different features now that have been added to ZapTest since the last time we talked that maybe can help with some of these things?

Yeah, I appreciate the question. Thanks, Joe. Yeah, definitely. Think of ZapTest as an agentic factory. We take any input, for example, you trying to automate, I don't know. Business operation. And in order for you, it's like training a digital employee, if you will, right? So you input any information that you would train your employees to do, or particular require— use case, I would even say. We don't even, we don't even need the requirements. We can start with the use case or intention of some automation. And that's my count out of conversation from any Zoom call like we're having right now. Make sense? All the notes that we're discussing right now as a prototype will be collected and deployed to ZapTest Plan Studio. We're starting in the Plan Studio. This is the beginning. So all this information will be processed by agents and presenting to you in the very comfortable view. All the requirements, structured requirements will, like we used to have on a requirement management tools in the past, right? That's going to give you priority of requirements, different parameters of the requirements. And then the human in the loop reviews them and approves or edits as needed. Now, if we don't have enough information for requirements, we have also recorder. That helps subject matter to record particular workflow. And any screenshots, any textual information, library of any information that you had, user guides, anything that you had before, screenshot, video recording, it's all put into the ZAP Test, Plant Studio. And based on that, we generate requirements. Now once the requirements been gathered and the human in the loop subject matter expert approves them. With the single click, it's in a matter of minutes generates for you full library of the textual workflow description, entire framework broken to the reusable modules, test param— data parameters, all the test data, right? That you're going to use regardless you do test automation or you got business process automation, you need the test data to start building the workflow based on, right? Again, another subject matter steps in that's responsible for Automation Architect. Automation Architect reviews all the framework definitions, edits as needed, make sure that framework explained by AI precisely. And with a simple click, it sends it to the full scripting process. So another agent start. Scripting. That agent generates the entire library of scripts that's ready to execute. Of course, you have another human standing in— stepping in, I'm sorry— to review the scripts, to debug them and make them ready for execution. Now, that's a part that we all use with we all used to as a test automation engineers, right? So we develop the automation framework, but here our mission is just to review it and adjust instead of writing anything from scratch. Make sure that all automation been adhered to, object repositories, test datasets, and what's not, right? And once that will be assured, we send it for mass execution. We have an Amaranth Studio that allows cross-platform execution on the disparate platform, right? So let's say one operational process can be run on multiple operating system, on a workstation, multiple mobile devices, anywhere that we need to run, right? And we can approach it from UI perspective or the API perspective or combination of both. And everything that I described to you can be done literally in one or couple of days, depends on the human availability.

I was gonna say, it sounds crazy.

Agents are ready. It sounds like a sci-fi, right? What I described to you right now, we as a professional understand that we're talking about at least 6 months of the work, man-hours of the multiple teams altogether, not even the individual, but the team. Makes sense? This is how much we at ZAP can offer to market today and what in general AI capable to help us do it.

All right. Alexei, we talked a lot about the Department of War, but when I'm thinking of Microsoft, they already have Copilot. Why would they even— I'm not belittling ZapTest. Why isn't like Copilot enough? Why would they say, okay, let's get ZapTest in here to help with this?

Copilot operates in the— with the slightly different mission. In our case, we combine not just the agentic execution, but we also combine with the very, uh, I'm proud to say sophisticated technology on the computer vision and interoperability layer. So our technology has particular mission. It's not generalized from knowledge. It's aligned to execution. Make sense?

Yep.

So Copilot works as a part of the whole process, and we do use— we have our Copilot, which helps also for human in the loop to speed up things as needed. But our goal bypass Copilot. Our goal to generate agents that do things autonomously.

So the flow you described with ZapTest, is that the same flow that's being used in Microsoft in the Department of War? Yes.

So in the different ways, we definitely packaging it different, depends on the mission and the project. And but in general, yes, this is the underlying technology that we use. And we're constantly improving it. You remember we talked about it 2 years ago, that's in our— this is our motto. We're always pushing boundaries. I like challenging the projects and I jump into any impossible to make it possible.. And I'm glad to say that I succeed doing that.

And Alex, what I like about Zaptest, what I remember, maybe I assume you're still doing this, is that you literally, when you say human in loop, I think you're not just selling a tool, you're also selling like a person that understands the tooling, right? So when they go to something like Department of War or Microsoft, they're two different companies, two different approaches. You're not just telling them, like, am I, do you have someone that works with them then with the tool that knows the tool inside and out to actually implement it? For them, not actually make them conform to how you have it rigidly defined in your mind?

Absolutely. Yeah. So with every project, we have a ZEPP expert that assigned to it. And the mission of ZEPP expert is to be fusion between the process architect and the support engineer, right? So it's a hybrid model. Everything is ZEPP hybrid. We try to— we love hybrid. And yes. And what the Zap Expert is doing basically, it leads agentic automation from beginning to the end. And we want to make it very comfortable for users, no matter they on the commercial side, on the military side, especially on the military side, that's our warfighters need to know that if they have questions, they've been answered right away. Makes sense. And on the commercial side, we work with the similar pressure that we need to deliver things. And if things needs to be done now, it has to be the questions. They need to be answered right away. But we've always had this model and we, we see it as very efficient this way. And it's a part of the overall package that we offer to our customers altogether.

So Alex, in the spirit of Independence Day, the kind of episode that I think this is really going to turn out into, what, what should, if someone's listening now, what should organizations declare independence from in say the next 12 months?

I think an organization need to clear independent from the old style habits. Makes sense. I wish everyone not too afraid to experiment with the new approaches because we have no choice. If organization stumble on the old processes, what they used to do and trying to grab to it, they're gonna eventually lose. They have to be brave enough to start experimenting new things, not too afraid to to implement ways how to move things faster. Because technology and agentic AI predominantly helps to deliver solution at a way higher speed than humanity used to. And I wish all organizations to adapt AI, not to be afraid of it, and basically make it work into their abundance.

All right, Alex. So I want to timestamp this and maybe get you on 2 years From now, maybe some predictions, any predictions in the next 2 years where you think AI is going? Like I said, it's moving fast. No one can really tell, but you have your pulse a little bit more closer than some of us. So curious to know, get your thoughts on that.

In 2 years, I believe we're going to have a similar conversation, but at that time, more enterprises will be already riding on AI, on the AI agentic processes. So that's going to be a new normal. And I truly believe in that.

All right. I know you said companies, it was always hard to get people to invest in automation and QA, and now it sounds like you're able to do that. So how does it become the new normal then? How do we go from that to all of a sudden people are just adopting it?

If leadership allows their departments to start and experiment with AI and report their achievement and progress. And then combine it in the unified processes within the organizations. This is going to move the progress faster. Make sense? The bureaucracy, this is what everything is trapped into. The, all the decision-making, making it slower because they have to approve it 10 times with people that used to do it on Excel spreadsheets. And they discounting ability of the technology, what technology is able to do right now. You can literally, we all use ChatGPT today, right? Let's say ChatGPT, Gemini, whatever, different, different options, but we all use AI. And as we're using AI in our private life, the same thing they can be doing at the professional level. We can create agents that's gonna have skills that's be able to help us as professionals to do things faster. Makes sense? I'm using it every day. I could never be as progressive on the, let's say, different aspects of the project that I'm doing, me personally as an engineer, if I wouldn't have AI explaining details, devil in details, right? You misunderstand something, you can easily get this information. And I'm not hesitant to say that I don't know a particular aspect I need to learn more about. I'm very good at what I'm doing. But if I'm applying what I know and what I'm doing to the subject of expertise that I have and I'm not expert in, I'm relying on AI and maybe multiple sources of AI. It doesn't have to be one answer. I like to, I like, I have my two engines between the OpenAI and the Gemini that I'm double checking and do my due diligence to make an average decision, right? Overall. But, uh, this is a part of my everyday work and everybody can do that. Every department, if we'll take the, even the task automation, we'll take workflow automation, the business operation department, they can build the skilled AI agents that at the first step can help existing processes to be done faster. And then they can build upon that step by step, but overall culture needs to change. Overall culture from scarcity that people are going to lose the job has to change to, guys, we're going to do much more altogether the way we are. Because if we're going to enable and teach our personnel how to fly with AI, we all as an organization are going to fly much higher and faster. Makes sense?

Absolutely.

We're now from being the pedaling the bicycles to the supersonic jets. We're already flying. All we need to do is just to take the controls and teach our personnel how to be part of the crew.

But culture is the hardest thing and people are really scared. So I don't know if AI could fix that. Even AI can't fix it.

I'll be honest with you. I truly believe in our national culture.

Yeah.

Because I've been raised in a different culture and I'm proud American all my life, like adult life, right? And I want to say that our culture, very humane. And we are as a strong and prosperous nation, we're always going to look out for people. And I'm pretty sure that any organization will figure out the humane way to keep their team and enable them to be more productive. To generate more revenue for the company altogether. And that's a matter of the leadership.

100%. Love it. And Alex, before we go, any parting words of wisdom?

Again, in the fine tune of what we've been talking about, I wish everyone to be ready for change. Don't afraid for, of change. We, uh, running into the era of fast-moving things. And if we're gonna be ready for this turbulence, if we're gonna be ready for changes, we will be in much better place as a society altogether. Makes sense? And of course, adapt AI. Don't, don't afraid of it. Don't try to fight the dragon. Learn how to ride it. And we're gonna be much happier and faster and stronger as a nation, as a humanity altogether. And God bless America. Happy Fourth of July.