Automation That Works Every Time and Quickly Becomes 2nd Nature with Greg Rog

Welcome back everyone to another episode of Mastering Tech Growth. I'm your host Mike, and today we're diving into a topic that is reshaping the tech industry: AI-driven automation. So from automating business processes to improving efficiency, AI-driven automation is a key driver for success for the modern business. Joining us today is Greg Rogg. He's a serial entrepreneur and I would say an expert in automation with ample of experience developing smarter tech-driven solutions. So Greg's journey includes co-founding EasyKart, a platform that enables educators to create and sell courses with ease, Automation House, a company that that leverages no-code tools and robotics to grow online businesses, and Zenkal, which is a game-changing calendar app that automates scheduling across multiple platforms. Recently, Greg and his team released Alice, which is an AI assistant for every task that works with your apps. And full disclosure, I am a user of Alice. The moment I installed it, I was like, Greg, come on the show, let's talk about this. So in this episode, we'll cover everything that implements automation, exploring cutting-edge solutions towards that, which is— Alice is one of that. So before we begin, a quick reminder to our listeners: if you enjoyed the show, please subscribe, please leave us a review on Apple Podcast, on Spotify, on CastBox, or wherever you're tuning in from. Your feedback helps us grow and bring more insightful content to the tech community. Okay, let's dig in. Greg, welcome to the show.

Yeah, it's good to be there. Thank you for the invitation.

Of course, that's going to be an interesting one. I love that. So before we begin, Greg, um, when I say mastering tech growth, what comes to your mind?

Yeah, for, for me, it was always, uh, doing leverages to, uh, spin off my multiple online businesses and ventures. And I also like to build on leverages, and leverages are especially automation for me for the last maybe 5 or 6 years. It's kind of a new approach to automation that I managed to utilize, uh, for different projects that I run. And, uh, we have like really small teams that are kind of handling big projects thanks to automation itself. We have millions of operations and thousands of robots that are working for us daily, and they are inside of all the processes, uh, of our online, um, or product companies. So basically, we employ automations to handle customer service, marketing, billing, you know, sales. And in all of those processes, we kind of inject those small automations to help us accomplish better results and also just run businesses kind of on autopilot. So this is the biggest leverage that I have for the past years. And once ChatGPT came out, we started to explore how this can be handled with AI or with the help of AI, or how we can utilize AI to just kind of like complement the automations that we already have. And this is actually the idea that, that brought Alice to life. We, we did Alice, which is like a simple— maybe not so simple anymore, but an assistant, a native app that, that helps you connect to automations. And this was the original idea to bring this software to life, was basically to use our existing stack, our existing automations that we had, mostly on Make.com, some of them on Zapier, some of them custom built. But we wanted to have an interface that can talk to our automations. So let me explain this way. In the old times, we had to create a really like robust data structures that would enable our automations to run. I mean, we had to have this structured data, you know, every field has to be correct. If there was an error for triggering some automations, it will ultimately stop, and it was pretty difficult to debug. So you had to create a structure of data, and then the structure of data should be really clean in order for automation to run. And, uh, one of the really bright ideas of, uh, of Adam, who I work with on Alice, was, okay, what about we ask AI to create those triggers? What, what about we ask AI to create those inputs to automation so that we can just chat to an interface and we have these strict data formats that are produced as the output, therefore running our automation. So that's the idea for Alice, so that we can easily chat with the interface and ask for, for example, to issue an invoice, or maybe to create a subscription for our users, or maybe to just, I don't know, add someone to a mailing list. And within this chat interface, whether you chat with it, whether you just write messages Alice behind the scenes will format it as data for our automation that fuels our automation, will trigger our automation, and then make sure that the process is finished within our API or some other API and, and that the job is done. So that was basically the idea and one of the biggest leverage we managed to create.

Okay, so we set the scene, we have an Alice which is a Windows/Mac app, effectively an interface which is, you know, think of ChatGPT, you're able to chat with it, it's, it's literally the same, right? But if you issue a command, these commands actually now go to the backend, and the backend is the Make.com, Zapier, or potentially your own hosted workflow automation tooling, right? So these commands will just go in and do— well, I mean, imagination is your limit, right? It's like, correct, as many tools you can connect up and all the workflows you can come with, as long as you can actually draw the workflow you could automate it, right?

Yeah, as long as you can create a scenario in a no-code tool like Make.com or Zapier, you're good to go. But like you said, you can also connect your own API. So, you can connect to your own API to your own app and you can talk with AI. You can talk and send commands to your own app because Alice will format a perfect JSON structure for you and just send it out so that any of the tasks can be performed. So, apart from all the tools that are you know, popular and have their own APIs, you can use your own. And you can even use— we have a custom AI feature which is, uh, just, um, an option to connect your own backend to Alice as well. So we are exploring that now. We are leveraging the knowledge bases of companies, uh, even behind firewalls with offline models that can be plugged into Alice, and you can chat with, with those. So this is also an option to, to connect different sources. So in a way, like you said, it's very similar to ChatGPT as a frontend, as an but under the hood you have a lot more possibilities because it's a lot more like geared towards different models using, you know, not only OpenAI's, also Groq, which is super fast and insane, but also, you know, Perplexity and Claude and many other models that you can use in parallel. And also it can, it allows you to connect even your own custom AI with, let's say, offline models like Ollama.

I mean, when I installed it, Alice, and I, you know, I looked at the setup where you can have a different assistant. So you can have a, you know, an assistant for show notes generation and assistant to, you know, look at the keywords for the particular things and, you know, marketing bits and stuff. And for me, you know, a lot of around the podcast production and, you know, social media and things like that. I was like, this was what I was missing because I had a huge amount of just bookmarks with, you know, to the prompts with ChatGPT where I would have my, what I call the playbooks in, you know, in a Google Doc where I would click on this and go to that prompt, fill in and get the results and sort of manually drive that. And then I was like, oh my God, like that's, that was a missing bit, you know? So, so I, my mind got blown like instant after, you know, a couple of hours of usage, just, you know, invite all the listeners to give it a go. So if we put Alice now on the front and just go back to the, in the back and go back to the topic in general for automation, I really love how you put it, that this is a huge leverage for the businesses to actually get their processes to another level when it comes, you know, to just efficiency and, you know, and people. So what, like, because you're doing this, what was like the main challenge? Like, let's say business, you know, listening right now, they go like, 100%, you know, I have Stacy in HR who does the same thing, copy paste or whatever every day. If I could use that, you know, she could just take that and get that efficiency. Like, From your perspective and your business, this is like, what was the main challenge to actually onboard these automations at scale?

Yeah, that's a great question. And actually, this is, this is a really tough one because it boils down to many of the processes that company has in place. And we've been doing that for many different companies, big and small enterprises as well. And we try to look at what they do. And we try to propose the ways that that could, you know, use, utilize automations to help them, uh, make tasks faster, the, the ultimate problem was always the same. They, they didn't have the processes in place. Even if they had the process, it was not like, uh, you know, it was in someone's head basically, uh, not, not put in a guidebook or even, you know, Notion in a simple form. So basically the first thing that the homework that, that you have to do in your companies to basically decide what kind of processes drive your sales, that drive your customer service, and name them and put them as a simple instruction, even with bullet points, you know, because it's like, it's the starting point where also most of the optimization comes from. We realized that when people started doing this for, for their businesses, they found a lot of, you know, waste in place that they already have in place and they can fix right away. So they started to map out the process and they realized while mapping the process, oh yeah, this is, this is clearly wrong. We have to, we have to do it differently. And this is the first step that will, uh, that is not connected to automation at all, you might say, but it will improve, uh, the quality of what you do. It will, it will improve the speed and save some time of, of the people who work with you. But then once you have those processes in place And then you, then you have to decide which of those is most likely give you the most advantages when you try to automate it. There are processes that are very robust, manual, require experience, require people, you know, uh, the connection between some people, and, um, and they, they simply can't be automated like, uh, in, in a way that we, we try to do it. So there are, and it's great because We just, we just don't want people to be, to, you know, do the monkey jobs, do the things that robots can do for them. We don't want to erase relationships and what business is all about. We just want to give people more space with automations to do what really drives the business, what's really the relationship part. And this really helps. But if you find and qualify the process that you already have as a process that is can be easily automated. For example, um, invoicing process and also maybe contracting process. You have different contractors, they all fill in some forms, and this can be all automated. Um, in my companies we have processes for that. It's super easy. They fill in the form, then it's automatically transferred into an agreement, and the agreement is automatically processed so that they just click and process the agreement. And we even schedule a transfer for the job that is done after some, you know, marking it like, as done. So, so basically there is, uh, there is a lot of things that you can automate in most of the processes that, that require, um, people to move data between, uh, let's say, you know, uh, one software to another. This is, this is easily, um, this can easily be automated. So you can think of those tasks that, uh you can just create a map of processes and then decide, okay, this process is a great example of something that we can automate, and let's focus on this one process and do it end to end. Try to, um, try to do, uh, as many things as we can in this process automatic. And for the process, if you can imagine like, um, like a map of the process with bullet points, if you have like 100 bullet points that leads you from beginning of the, of the cooperation with someone to basically maybe paying him out as a contractor. So you have hundreds of hundred things, and maybe 80 things can be automated, maybe 20 cannot, but basically most of it can be automated. And what you have to take into consideration is also I want to make those processes proactive. So basically we don't, we don't need people to check, constantly check in for something. We, we are, we live in the, like, there is a lot of waste in companies where someone is waiting for something. He's waiting for an email to come, and then when the email comes, he processes it, but he also checks the mail every few hours and, you know, wastes some minutes on that because there is no, there is no email coming. So basically, I try to design those automations so that nobody has to wait. So that, for example, we have very simple tasks, to-do lists. You imagine like a the simplest, uh, to-do app like Todoist, or maybe, uh, I don't know, Google To-dos. I think it's called like this, but very simple. I mean, it's super straightforward, and I want my automations to fuel these apps with tasks for people. So we use Todoist, for example, and in different processes that we automate, people just get tasks from the machines. So they are not waiting for anything. They just open the to-do app and they see what they have to do in this very moment. They don't have to check anything. So this is like reactive versus proactive, and I don't want people to be proactively looking for, for searching a solution for a problem. I don't— I want them to be acting reactively to what they have to do because automation just said that, okay, now we have to do this and that. And this is far easier, and this is where a lot of like time is wasted in companies.

I actually love that very much. I, I It took me years to develop proper habits in business to actually work with the to-do apps. And I still see people just starting their days with just random email checking, responding to some Teams or Slack messages without really any structure within your day. Where it took me years to sort of develop the habits where I actually start my day by opening my to-do app and going like, okay, what did I add yesterday? Yesterday that I need to do today, or what did I add a week ago with the reminder that popped up today, and just get that filled. So effectively what you're saying is that with a proper automation you can fill your to-do lists automatically, so you don't even need to like chase and add, and it just, just comes in. So you open that up and you just start knocking out items. That's powerful. That's powerful. I love the monkey jobs. Monkey jobs, that's the first one.

Yeah, because they are, you know, it's like We try to do the robots so that people don't have to do the robot job, you know, it's like, you see it everywhere. You come to any company and you see people just, you know, pasting, copying and pasting things from one place to another, or maybe trying to log into interface, you know, we try to eliminate with automations, we try to eliminate the interface, the use of the interface, you know, so that people don't have to log in, forget passwords, you know, reset them. We spend like hours on that if you can't So basically, we try to eliminate that and we try to communicate with all the software on the API level so that we don't have to, you know, go to the UIs and do those operations in the UIs. Of course, people would say, yeah, why do you have to talk to Alice to, I don't know, even add someone to a newsletter, you know, or add a client to a CRM? Well, you can log into CRM and you can add this client, you know, probably a few clicks away. Well, maybe you are a few clicks away, but you have to open the browser, it has to load up, you have to log in, forget the password, sometimes reset it, then remind yourself where, where do I add customers in HubSpot? Then, you know, you have this huge pile of fields that are unnecessary in the most cases. Instead, I just open Alice, I press Command+Shift+D to summon Alice, and I just say, hey Alice, can you please add Mike to my CRM. His email is this and that, and that's it. And she will, she will do it for me.

Or you can actually open a mic and say that instead of typing, which is another— yeah, it's very cool if you have like a proper headset, you know, like people used to make calls, like sales guy. I mean, I play with that and, you know, I was like, oh my God, you feel like you are in like this, you know, battleship control station. Basically, you're just clicking buttons and going like, do this, do that, and things just happen. It's such a powerful concept. And seems like movie-like, but it's just, just go ahead, do it. And, you know, it'll take some time to set it up, but once you, once you do, it's just so freeing by nature.

Just like, we'll transition to that, you know, we'll transition to that. No UI is natural for a human. It's like, uh, you know, no, no UI, uh, that we have with thousands of options, uh, that you have to click. Uh, Google UI is simple, right? You have one input, and this is ultimately what we're gonna accomplish with AI as well. So we'll just have an input which will be very simple and straightforward. This can be typing, but like you said, speaking is more, much more natural, and we'll get used to it. For now, it's not ready. You know, people struggle with Alexa and Siri, and this, this is just not— this doesn't feel ready. But AI will enable this to be ready in, in, in a short period of time. So in Alice, it's also not yet ready. Like, you don't feel it's like super natural. It's not a natural conversation. But you can still do things by just talking to Alice. But give it a year or maybe a few months and you'll see that you'll be able to just like fluently chat to, to AI and she will just do the task that you ask her for.

I love that we had the like AI and data, big data person in the show about like maybe 2 months ago and we literally talked about what Alice is. We were just like talking about AI future and, you know, the projection was that we're just going to talk with the modules. And we'll get the work done. And, you know, I've seen Alice and I was like, it's moving so quickly. Like the market is just coming up with these awesome things.

Yeah.

Can I take you a little bit back, Greg? So when you were talking about bullet points and, you know, doing the flows, do you know the concept which is called value stream mapping?

No.

No. So if the listeners put it in Google as well, it's effectively, you know, you have a range, right? And the range is very often is, you know, some type of value stream, right? You go from, hey, client asked for this, you know, and it ends at client has that, right? And everybody in business, stakeholders involved in that, they come in a room, they sit down and they go like, okay, what is the very first step we do at the high level? You know, maybe we'll get requirements, right? Okay, what happens in that biggest step? Like, how do we actually get the requirements? You know, okay, you know, Stacy sends an email, this and that. And the value stream map effectively, you know, forces the stakeholders to spend days in a workshop to just map out what is called the current state. Right. And then you look at this current state and you go like, well, okay, you know, we have some issues over here. Lead time is big. Process time is huge. You know, maybe another term is called complete and accurate. When you move through the stages, it's like how complete the previous stage was and accurate to move to the next one. So it's a very good framework to really do what I believe is a very first step to go like, okay, well, all this area, once we do value stream mapping, is actually quite poor. So if we can go in that and just automate segments of it, of course, yeah, you know, value stream mapping wasn't created for automation. It was created to create initiatives of how business can streamline their processes. But ultimately what we are right now with the technology is that you can look at this and go like, Should that be a person or should that be a proper automation? So I feel like these two methods, value stream mapping and automation, especially AI-driven automation, come together very, very nicely. So just a great idea.

Yeah, that's a great idea. And, you know, I have to add that most of us, especially in, in the bigger companies, we kind of know the, the what RPA is, this robotic process automation. Which is like totally different to what we do right now. I'm a, I'm, I'm, I'm a proponent of like super modern approach to automation, not the old one. The old one, uh, was basically someone offered you automation. And this is also why so many people and also businesses, they, they kind of turn, turn down the, the automation at some point because that was the process where someone would make you an offer to automate your processes then he would come to your company and maybe ghost your employees for, for a few months. And then it would take him like few additional months or a year to create like a custom version of software that you would use to automate the processes. And then you, you'd have to, you know, pay this guy indefinitely. And, uh, this is old ways and wrong ways of doing that. Uh, this can be, this can be, um that the modern approach to automation is much simpler. Uh, once you have those value mapping in place, it can be done really easily with the modern tools, with simple no-code automation tools. You don't have to write your own software, you don't have to spend like a huge budget on it. It's, it's really inexpensive in this, in this regard. And also you can think of employing AI on top of that to basically help you with this, because if you have a lot of automation and you need to trigger them with this proper data, like I mentioned, the data that you have to correctly format and stuff like this, a lot of people who are managing this will make mistakes if they don't do it daily, you know. AI is like, uh, I know that it's silly because we know that AI makes mistakes, right? We know that models tend to hallucinate, but What we figured out in our way to approach it is basically it's mistake-free. It's really rarely that we— that any of the mistakes occur. If they do, there is no panic mode, you know. There is just— Alice gives you a friendly message that, sorry, something went wrong, and this is the details of this error so that you can easily handle it. But I think, you know, there is this promise of agency and that we'll be able to tell AI what to do and it will do it for us. I think it's very far in future, to be honest. Like, the general approach to agency is close to impossible right now. We created so many different interfaces. We created so many different things that let us, I don't know, maybe book a holiday, you know, it's not an easy task to tell AI to book a holiday for us. It involves so many things that I don't think it will come pretty soon, you know. And business-wise, it's also not like this. It's not going to work, you know. I'm going to say, yeah, can you please optimize my sales? And no, it can give you some ideas, but it won't be able to, you know, connect to any software that you use. Uh, someone use Teams, someone use Google, someone use HubSpot, Pipedrive, and stuff like this. It's really not possible in a way, in a general way, that people tend to, like, believe it will be working. So our approach is, is a bit different. Our approach is still using automations so that we can strictly control how the data flows, what kind of software is used in the process. We can see how it works, we can debug it, we can see every step of the process, but we are initiating those automations, triggering those automations with AI. Which helps us with basically skipping having to remember everything, having to structure the data properly. And, you know, yeah, this, this is the process that was a pain in the ass in, in the way, the way we automated things. And now we can eliminate it with the help of AI. But still, it's the agency that, that we invented, kind of like, you know, with automations. I think it's the best and the most reasonable approach to agency today, not the promise of the future, you know, agents doing things for you, which might eventually happen, but it's going to be quite difficult.

Okay, so we have a couple of things here. First question, you said it's almost error-free. So is that that you trained your own module? Is it just like a very good prompt, which makes sure that the JSON data, you know, they get from API calls is able to be sort of understood and and, you know, Alice sort of gets it, or was it something in the middle, which is your stuff, like your LLM?

Yeah, we are operating in the areas where things are not that difficult, really. Because if, for example, you want to insert some data about the customer to the CRM, it's like, it's simple data. If you want to extract, for example, results for this client from a database of, I don't know, real estate, um, it's pretty simple and strict data. It's not talking to the model where it hallucinates about, you know, what do you want to do in the future. No, basically it's kind of like, uh, strict things that we want to do. And the job of AI is basically to take our inputs and to structure it in a way that automation can then take this data and distribute to APIs, to different software. Um, so this is fairly easy, and I think that's one of the easiest tasks that you can give, uh, AI. And for example, if we are talking about, you know, adding a customer again to— or adding an event to calendar, it's pretty easy set of data to work with because you say, I want to meet with Mike tomorrow at 9 PM on Zoom. You know, you have all the details, and there is almost, almost 100% success rate that AI will be able to, you know, understand this data and format the structure of the call of the trigger for the automation correctly, because it's really, really simple. So if you format a prompt and you also give some examples that someone, you know, can, um, can give to, to, to AI, can, can talk to Alice with these examples, it's really— the success rate is really high. So yeah, that works.

So when you mentioned agencies, uh, Greg, and you said, you know, we're not close and it's probably going to take some time, What do you mean by agencies? You mean a set of assistants, of AI assistants coming and building, you know, almost like a business?

Yeah, correct. And to simplify this, and this is really big and something that everyone like kind of like in the tech world obsesses about, and this is part of the future that we will be able to ask an agent to, for example, create a website like Amazon for us or something like this. And then what will happen is we won't have, to simplify again, we won't have this really quick relationship in the question we ask and the answer we get, like in ChatGPT. But we will have different agents that are responsible for different tasks in this process. And we will just ask to do it, then, I don't know, go for a coffee or maybe go to sleep and wake up tomorrow. And then in the meantime, this different agents who specialize in different tasks and understand different areas of, let's say, web design, you know, development. Again, I'm making it a bit simpler, but they will communicate and iterate through ideas. And this is like when you work with Cursor or some other software for web development. So now it's kind of like— and also with ChatGPT, the idea is pretty clear. So sometimes you ask ChatGPT something and it will give you about you know, incorrect answer, and you say, no, no, I don't think it's like— and he will like correct itself. And if you ask him to autocorrect, if you ask him to apply critical thinking to what, uh, to what he does, to what— to the answers, you get better answers. So this is ultimately the same idea, uh, about applying this critical thinking, but in a chain of prompts, in a chain of questions and answers that you don't necessarily see, but they kind of give you at the end the result that you want. So if you, uh, if, if you ask, um, an agent to build the software, then it will try to build the software, but not as a one response with a ready-made code, but it will try to, to build small pieces, small parts of code, then iterate, test them. If they don't work, improve them and stuff like— and, you know, it will take time. It will be a process that, that leads to a solution that you want. So this is what I mean by agency. And, um, and this is something really important that will kind of like, you know, hoping to, uh, to accomplish in the future.

I'm sure it will come together. I mean, maybe it'll take a couple of years, but it's a, it's really manageable, uh, future. Like, you know, you, you see the, the bits of that future and you go like, yeah, I can see how that will come together, rather than, you know, it's being some type like disruptive heads in the cloud type of situation, I'm sure will come together. So, um, when you were talking through the range of, of automation examples and, and, and just sort of me thinking about business processes, I sort of went like, okay, well, we have what I would call like a simple automation and then something which is just complex automation, which effectively can be defined by just how many steps you need, you need, really need to do in that automation and how many maybe different systems needs to sort of talk and get connected up. So from your experience, maybe within your business or, you know, serving others, like, what is the, like, most complex automation you made it work? Like, how many steps, maybe how many systems, just, just to give a flavor to people what's actually possible?

Yeah, the funny thing is that I, I don't aim for complexity at all. So, um, this is funny, but, uh, once you start learning automation, and it's, it's a really cool experience, I really recommend you, uh, start, I don't know, maybe learning make.com, which is my favorite tool. They have academy. I'm not affiliated in any way, I just use them like day one. They, they used to be Integromat, now, now they call themselves Make. And, and, uh, you can start building those simple automations, just connect the dots, and this is fully no-code, uh, You kind of need to become a bit technical, even if you're not, while trying to learn it. But once you're in the process, it's, it's so much, you know, dopamine that you get from it. You feel like, like, you know, a kid in Disneyland when you create an automation that actually works and can, you know, pass the data from here to there and then, you know, notify you on Slack but also do something else. And once you start doing that, you start complicating things. You start adding new modules. You, you kind of like, oh, what if I do this and that in this automation? So, so I've been in this place. Uh, I had automations with like hundreds of modules, and then I realized that when I, when I come back after, you know, a month or two to debug it because it stopped working, because, um, I don't know, maybe API changed or something like this, I can no longer understand it and I have to still rework what I did. So the beauty is actually in simplicity with these automations. So instead of like creating a huge chain of commands that, that, you know, um, that do like super complicated tasks, I really prefer to create like scenarios which are super straightforward and super simple. Sometimes they can contain a lot of steps, but you can clearly understand the value that you get from the entire process, and you can name it as a name of the scenario. So you can say that this is an OCR for an invoice, for example, and you have this process. If it's complicated, okay, let it be complicated. I don't know, multiple steps, uh, it might be 20 or 30 steps, but you know what it does. But don't do a process like, I don't know, managing invoices, you know, and then you have OCRs, you have like, uh, 1,000 different modules working, and you don't even know what, what's going on, what's, what's going where. So basically, um, I really, I really, uh, prefer those simple automations even if they work in a kind of chain. You can, you can link them together. So for example, if I have this OCR for an invoice, then at the end, at the very end of the scenario, automation scenario, I can say, okay, now that you've extracted all this data, pass it to the other scenario that will parse this data and add it to my QuickBooks or something like this, right? So So then, then the, the other agent kicks in, you know what I mean? And then I have one agent that does the OCR and the other agent that does the filling of my, you know, um, Income or, or, um, or QuickBooks or whatever, so that I can easily debug it. I can easily see what went wrong and where, in which scenario, what, what part of the— what module, like you mentioned, I need to focus on. And that's far simpler. So It's really about simplicity at some point and making it available for AI to work with. It's also about simplicity because if you imagine, if you think of it, we as humans, we also don't process super difficult tasks all at once. You don't want to say, you know, that I want to deal with this contractor and you want to, you know, make a transfer, make an agreement, make a contract, sign it. You don't do it all together. You do it step by step. So This is very important to have those small steps available for you, and then Alice can do it for me easily. And, uh, this is how I approach it. It's, it's like chain of simple commands that I know that, you know, once won't, won't be, um, uh, won't be adaptive to, to like an errors that much.

Okay, so there's a definitely then the concept of complexity leads to, uh much higher cost in maintenance. And cost is, say, time, right? So, so you sit down, you go like, okay, I'm going to create this automation. If I'm going to make it complex, I'll need to maintain it. You know, maintenance becomes much more prolonged, it's much harder to debug, it's much harder to look inside. So I'm wasting too much time over there. I'd rather take it in isolation steps and connect these steps, which are much smaller, much easier to, to manage. Is that —like a right assumption, right? Okay. So there's some concept of, like, you know, coding as well, right? Because when you were talking, I was thinking, okay, so we just want to decouple basically that the whole chain in the smaller bits, which, you know, just goes back to, like, almost like a serverless or microservices approach where you just don't want to—you want to have one service as small as possible serving one problem, and then you just connect them up.

Yeah, one function at a time.

Yeah. Okay, okay, that makes sense. So, okay, so you are where you are right now with Alice and, you know, the concept of automation. And you have, you know, a lot of experience in that, you know, with the business and as well as product. What is waiting for us? And you mentioned the agencies, Greg, but like if we put like some creativity, like what we going to be seeing in like 10 years in the space of automation?

Yeah, I don't have a crystal ball anymore. It's like it's moving. It's moving at such pace that you cannot like— this is some indication of something really serious if you compare it to some other, you know, hyped things that we had in tech in the past years. I think AI is— the indication of AI being really serious change is how fast it kind of goes. So at this point, it's really hard to predict what we're going to have in future. To be honest, when I'm watching some, some, some bright minds behind models and how they approach it, They openly say that they don't know as well. So basically they say that, for example, you know, they trained the model, they changed some parameters, and it started to work, but they don't really know how exactly this happened or why it happened. And, you know, they have the equation, the math is there, but at some point with, with, with, you know, billions of parameters that you have, it becomes so complex that you can no longer perceive it as, as a human being. I, I really think that at this point we can, uh, you know, hit a wall, uh, maybe, and, and, and just, uh, it will stop developing. You know, there is a scenario in which you can say, yeah, it's, it's over, we don't know what to do now, we cannot go, you know, improve it anymore. But I don't think this will happen, to be honest. I think that We proved that with the, with the computing power, we can improve it at least to certain, to a certain point right now. So it will, during the, the next year, it will improve thanks to only the computing power that we have. So this is an easy guess that it will get ultimately better. The agency is coming. I think that maybe it will require some other software layer, maybe a special API for AI, you know, some, some kind of like what Apple does, for example, an extra layer in apps that will allow communication with, with AI and with models. That's probably going to happen, but that's a technical layer. So, um, this, this is gonna like change, ultimately change the, the whole ecosystem of, of, uh, of business. And if I, if I should predict something, I pretty much believe when I see people, um especially in some other tools that we use, because we, uh, we use— we, we do tools for creators as well. Uh, so Easy Tools, uh, Easy.Tools is a platform where they can sell and create like digital— where they can sell digital products. And I see a lot of innovation in that. So for example, for them, a lot of markets now become accessible because of AI. They can easily translate the courses that they create, you know, ebooks, knowledge, uh, things like that. And they can like really expand with that, with what they tried to do, what they were doing locally. So this is quite big, and I think that, that we will, um, we will see a great like, uh, change in how people, uh, you know, consume knowledge, perceive knowledge, create knowledge, and sell it on the internet. And this is going to be big. And also I think that company-wise, we will see a slight change that kind of like will eliminate the, the particular roles in the company and make it like, uh, you know, now you have a marketing department, you have sales department, you have, uh, you have customer service. I, I really think that, uh, what's— what AI and automation would allow us do is just like, uh, you know, um, smooth the border between those, maybe eliminate them whatsoever in the company, and we'll just have operators who can plug into different areas of the business and AI that helps manage those, uh, those areas. So I don't think we'll have people who are like super specialized in, in one thing, because having Alice, for example, will enable them to, you know, tap into marketing, tap into sales, tap into the other things, and this low-level things, uh, AI and automation will handle for them. So they don't need this expertise to, you know, create a website themselves, or, you know, be a developer to fix some small thing, small bug in the software. Currently with EasyTools, uh, we, we can even enable, you know, customer service representatives to basically, I don't know, fix the label in the software. And this is thanks to some, some, you know, automation and AI. Running in the background. So I pretty much think that this will all come together in a way that we, we won't have the jobs that we have right now. We'll just have kind of like different, more general jobs, and we won't dive so much into the details and UIs of things. Okay.

It's a great future to some, maybe not so great to others. But it's a, it's a common future where I think majority who works with the technology of AI and automation sees that sort of coming in place and slowly starting to become real. So if I'm a business owner, so let's say, you know, I'm listening to it right now to the show and I'm a tech venture founder, right? So I'm instantly going, okay, well, I can squeeze out so much efficiency and as well just a cost management perspective is just is just great. Like, uh, like, where do I start? Like, do I, you know, do I go to the company which does automation? Do I start playing, uh, you know, with Alice and I give the tool within the businesses? Like, in general, like, what's your advice if, if company wants to do automation but is really don't know how to go from -1 to 0?

Yeah, uh, yeah, that's, um, that's especially For, like you mentioned, a founder or someone who is kind of like organizing this business, and it's not like, you know, a small manager in a huge structure, it's like a really, really great chance to basically explore it and have a lot of fun on the way because it's real fun if you kind of like exercise with it and start playing with AI. And also connect it to automation. So what I do is obviously one of the most important skills is the ability to talk to an LLM. So something that people call prompt engineering is one of the things that, one of the areas that, that one could look for to improve in this area. Obviously, this is great because it lets us kind of like understand how LLM works and try to force it to give you better answers, better answers, better results. So this is extremely important. So, uh, allowing ourselves to, to, to, to learn a bit more about that, uh, is important. Learning basics of automation is also like a pretty cool thing to do. It's, it's so much fun. Um, like I mentioned, Zapier or Make.com, that I prefer, are great platforms to start, and this is fairly easy. You just have to, you know, spend a few hours. I, I, but I think it's pretty, pretty fast as this is a no-code platform. After an hour or two, you already have a result that gives you like, gives you this dopamine and you can continue doing that. And also Alice. Yes, Alice is, uh, probably a really nice, uh, thing that you can try to implement into your daily, um, routine and productivity routine, uh, which kind of blends it all together. So this is really nice because you have those different models that you can experiment with, but from time to time you can also create some simple automation and, and accomplish something more, you know. So this is how I approach it, and I really recommend doing this, this way. So you don't like, you know, uh, just focus, like take a week off and try to do everything that you can, try to automate everything that you can. This, this kind of like I don't think it's a great approach because you will surely make progress, but at the end of this week, um, later on, you will probably forget most of the things that you've done and you won't use it that often. So what I recommend instead is just being curious and experimenting all the time, but also kind of like an introspective, uh, thing that will allow us to, to catch ourselves, like to stop for a moment every time we kind of see that we are doing something that's repetitive, that's boring, that's, uh, kind of can be automated. So every time I have a thing like this and I, and I think of it like, okay, how much time does it take? How often do I have to do it? You know, and I kind of like make the equation whether it's worth to automate it right now. And if I spot such a thing in my, in my workflow, then I try to automate it with with simple automation or maybe with Alice now, but it's only that I recognize it. I try to focus on that and spend some time to make it better for myself. I don't try to reinvent everything that I do. It's like, you know, a lot of people try to do it like this. Okay, I'm going to be more productive, so I'm going to buy 10 books about productivity, then about automation, then I'm going to sit for a month, automate everything, create a huge dashboard and in Notion, and you know, it's going to work. It's not going to work. It's like you said, you— it took you years to kind of like make those habits a habit. And this is also the thing with automation and AI. So the, the mindset is very important, that you recognize small things in your workflow that can be automated and you pursue this, uh, automation. This is the way.

I love that, and I resonate with that so easily, Greg, because I have we are like 4 people doing the show and I got like 17 things to do, like any day I have like 2 hours, right? And I'm like, okay, do I send emails? You know, do I look at the show? Do I do guest sourcing? You know, do I look at, you know, YouTube? Do I chase some other initiatives and things like that? And it's like at any given point I'm like, okay, well, there's so much going on. And I always think about what's a repetitive task and then just connect with the workflows and just take the low-hanging fruit. But it's a constant thing which happens when you think about efficiency of processes. It's not like you take a pause and then you automate everything and then you go suddenly to some other levels. A very, very compounded exercise which actually drives that. Yeah, yeah, yeah. I love that. So you mentioned prompt engineering, which 100%, and I'm sure our listeners know what that is. I wanted to just maybe add on top of that is that for business owners who potentially don't understand prompt engineering well and they just use like a sentence, you know, hey, please rewrite this email to sound more professional, which is probably like the simplest prompt you can do. You can find online platforms where they actually sell prompts, right? So if you don't want to spend days understanding how to actually do a good prompt, you can find professionals who sell prompts, you know, for various tasks. I'm not saying that all of them are good, but if you're really thinking about complex prompts and stuff like that, it's probably worth exploring Endeavor because it might save a little bit of time. Yeah, definitely. Yeah, definitely.

More, I think, on the like kind of inspiration part. What can I do? How can I, you know, structure the prompt? What can I use? What are the tricks, tips and tricks? How can I like structure it? It's, yeah, it's great for inspiration. The tricky part about this, like using ready-made prompts, it's kind of like a little bit like those Notion templates that I mentioned for organizing your life. So you basically buy and download them, but you never use them because someone else created them. And this is like the important part is the process of creation. So when you put like the prompt that you see, the ready-made prompt, is just the outcome of some process that is super important to understand in order to use this prompt correctly. So when you just get the outcome, you don't get the knowledge behind it, and you don't get like the iterations. If you run through iterations yourself, you're in a much better position to kind of like, you know, understand it better and make and get better results. So I tend I treat it like a great inspiration, especially for beginners. But ultimately, there is kind of no shortcuts. The more you talk with the model, the more you like spot this like aha moments where you get like, ah, this is where I got it right. This is where I get it. So basically, it's just like iterating through this. And this is how you learn how to chat with LLMs. This is like the ultimate way. But of course, you can make some shortcuts as well and see how others do it.

Yeah. I started like that. I remember, you know, ChatGPT came out, you know, everybody were just sort of like simple sentences, simple sentences. And then I stumbled upon a website and just, you know, I think it was Happy Face or Hugging Face. No, it was Hugging Face, I think. Yeah. And it was like, you know, act as this and be this. And I was like, oh my God, there's a structure to that. Like there's a, there's something which I have no idea that just became a rabbit hole where I went in and I was like, okay, well, there is an actual science of how you actually prompt work. Definitely. Yeah, definitely. Yes, definitely. Okay. So another thing which I just wanted to sprinkle for our listeners is that people who buy businesses professionally and there's just, you know, a lot of people and not necessarily tech ventures you buy, you know, you can buy a simple businesses you know, cleaners, cutting grass, HRs, or whatever. But people who are businesses professionally, the very first thing they do when they buy a business and when they look at how can we, you know, get the cash flow better is automation. They literally walk in an office and they go like, okay, well, a lot of what has been being done over here is just manual, you know, signing papers, printing, sending, or whatever. And the first thing they do when they walk in, they go like, what can we automate? And, you know, which roles we actually don't need? So it's just, just putting that, uh, emphasis on that leverage which Greg mentioned in the beginning, that automation, it is a huge leverage for literally any business. So whatever you're doing right now with your business, automation might bring a lot, a lot of value. So this is, this is where I want to wrap this up. So thanks, Greg, for coming to the show. I mean, AI-enabled automation and the tools you guys creating and Alice in particular, it's insanely well. So if people want to know more about you, businesses you run or Alice in particular, how do they find you?

Yeah, so you can find me on X. I'm mostly nose down in projects, but you can DM me there. I'm Greg underscore Rog, R-O-G on X. And you can find me on LinkedIn as well if you type my name, or you can just email me. It's, it's, it's different emails. But let me, let me give you an email that you can always use because it's my private. It's webstandards@gmail.com. But yeah, it's best for you to reach out probably on X where I check DMs pretty often.

Okay. We will have everything in the show notes. So if you're watching this on YouTube or listening that just just scroll down to the show notes, you know, click the more button and, and somewhere you're gonna find all the links to, to be able to reach out to Greg. And, uh, just before I close, you know, to our listeners, please subscribe, leave us a review. Your feedback is something which makes our day, especially if it's positive one. If not, let us know how, how can we do better. Until the next time, say stay safe and see you online.

Yeah, thank you. Bye.