So erledigt das neue GPT-6 Astra meine Arbeit! (ganzer Workflow)

0:00You may already be using ChatGBT regularly, but in the end, most of the work still falls on you. Writing documents, answering emails, planning tasks, and so on. In this video, I will show you my personal AI workflow in four steps, which allows the ChatGB team to take several hours of work off my plate every single day. By the end of this video, you will possess not just a few prompts, but a complete system that allows you to know exactly how to accomplish more work in less

0:22time using artificial intelligence. Let's get started, and I will show you in just four steps how you can set up this AI system directly for your own personal use. I'll show you the whole thing here in ChatGBT. Theoretically, this also functions with any other chatbot, and I realize that this video here is going to be a bit longer, but it could very well be the most significant AI video that you are going to watch throughout the duration of this year. That is why I recommend that you take the time right now to work through this video thoroughly, and I am certain that you will gain a great deal

0:44of value from it. In the first step, I am going to show you exactly what my current ChatGBT setup looks like right now. Secondly, I'll show you the AI tools that currently have the biggest impact on my day-to-day work. In the third step, I will show you what you are already able to automate within your professional workflow by using AI in this day and age. And in the fourth step, I will show you highly advanced AI features that, in all likelihood, fewer than 1%of all ChatGPT users are actually familiar with. This video has no prerequisites, but I imagine that you have at least had some experience

1:08using ChatGBT or ChatGBT Work at some point in the past. This is not an absolute beginner's tutorial; therefore , it is important that you are able to correctly classify the steps which I am going to show you right now. I would also like to ask you not to expect that you will be able to understand everything that is in this video right

1:20away. This is my personal AI workflow. Knowledge that I've built up over many years, and even if you only take away two or three ideas from this video, this video will have been worth it for you. Just a quick note beforehand: this video is not sponsored, and I am not being paid by anyone to show you everything that I am presenting to you here right now. It is simply important to me that you are well prepared for

1:38the work that is ahead of you tomorrow. That's why I made this video. If you'd like to support me and my work, please give this video a thumbs up. Thank you very much. So, in order to show you exactly what I mean here by this specific difference, I am now going to open up the browser version of ChatGBT to demonstrate it. I'm now logged in with a different account and now I'm writing to him here: "Hey, ChatGBT, as you know, I'm moving to Munich at the end of the month. Can you please track the current Six booking from September 27th and see if we can book it a bit

2:05earlier on that day? If so, please rebook the car for me directly on the website." So, now I'm sending this here and of course the result is nothing. Of course ChatBT can't do everything for us now. He doesn't even know that we're moving. He has absolutely no idea which Sixt booking we are even talking about here, and even if he did know, he would not have the necessary tools currently at his disposal to rebook this specific vehicle for us on the website at this particular moment. But let us see, if I now copy this message and then go into my personal AI system and send that

2:32same message once again, then it immediately gets to work and here it naturally knows that we are currently in the middle of our move. He knows exactly which Sixt car it was that he booked for me during this past week. He'll now check the confirmation in Gmail, then log in to the Sixt website, which he's doing right now, and see if the car he booked can still be rebooked . So now he is coming back here and stating that he has looked at the specific booking directly on the website, but he explains that a pre-booking is unfortunately not possible in this particular instance after all. That's why he's given me the

3:00exact date the Sixt car was booked. This is admittedly still a simpler example, but I believe it gives you a solid impression of how drastic the difference is between ChatGPT in the browser and this specific AI system here. So, let's start with the setup now, because if you set up ChatGBT correctly, you'll probably get two to three times better results than if you don't. It all begins with the simple fact that you should not be using ChatGBT in the browser as you do here, but instead should always use the ChatGBT desktop application on your laptop or your PC. You might be thinking now, okay Jonas, but what

3:33difference does it actually make if I use ChatGPT in the browser or via the application on my PC instead? Trust me, it makes a huge difference. ChbT is indeed significantly more powerful in the Daher version than it is in the browser. The app offers you significantly more functions, and almost all of the features I am going to show you now, which will help to accelerate your daily work, are only available to you within the desktop version of ChatGBT. So my main recommendation here is to begin by downloading the ChatBT desktop application for your computer, in the event that you have not already done so yet. The link to the official download

4:00page can, of course, be found right there in the video description. So get used to using the desktop app as quickly as possible, because later on, I would really be thankful to you if you had done so. Now let's move on to point 2. Just as important as the desktop app is the fact that you should never work within the standard ChatGPT chat, but instead use the dedicated ChatGPT work mode. This is just a small switch here. You can switch from Chat to the Work tab up here, but in doing so, you are actually utilizing two completely different sets of software tools. ChatGBT Work has a different

4:25so-called Hanis. Simply put, this means that ChatGPT, when using this work mode , is significantly better at writing and executing code, operating programs for you, navigating through files, and working on tasks independently for a long period of time as it performs its various functions. Therefore, my recommendation is that you use ChatBT not only within the desktop app, but also utilize it consistently in work mode. A quick note at this point: To use the work mode, you currently require the Plus version of ChatGPT for your tasks. You can currently get it for around € 23 per month. However, I can only recommend that you get the whole thing. I am not getting paid for

4:55it or anything. It's just my recommendation. I would suggest that if you are still feeling unsure about it, you should get the Plus version for just one month to start with. If you happen to be dissatisfied, you can always cancel your subscription afterwards; it is likely one of the best investments in your own education that you are able to make at this very moment. So, congratulations, because you have come quite a long way so far, because there is definitely a world of difference between the Chat version in the browser and the work version on your computer. You should now use the browser version of ChatGBT at most if

5:20you have a few quick questions, but in order to get work done, it is best that you always use the work version on your desktop. This brings us now to the third setup point, which is likely one of the most significant: the correct model selection. Because JetGBT, as you may perhaps know, has various different AI models at its disposal, and at times it is not entirely clear which of these specific options one should be making use of for the task at hand. You can find the models by simply clicking on the model name located here within the

5:43chat bar in the ChatBT Work interface. If you click up here in the middle, you'll see the different models that are available to you. This can be quite complicated in itself, but on top of that, there is the additional fact that there are also various different effort settings that apply to each individual model here. It starts at low, then progresses through medium and high to very high max, and after that, there is also an additional ultra mode available to use. Choosing the wrong model is truly fatal here, because with an incorrect model, you will either receive significantly worse results than you could have otherwise achieved, or you will end up burning through far

6:12too much of your available quota. In other words, the answer will cost you significantly more than it should. Choosing the right model and applying the correct amount of effort is one of the simplest, yet most effective levers you can use to get better results from your artificial intelligence. And don't worry, I'll tell you right away which models I personally like best at the moment, but you want to know what my personal AI workflow is like and what I

6:32usually do is look at the benchmarks. Simply put, there are a whole bunch of companies that are dedicated to testing and comparing the latest AI models every day. And that's exactly what you should know how to use for yourself. And to make a long story short, there are two sites that you should definitely be familiar with. The first one is the Artificial Analysis website, and the other is the Deep SWE Benchmark , which you should use if you are

6:52working with AI at an advanced level. Both of these websites are comparing the performance as well as the costs of AI models. Artificial Analysis, that is more or less considered to be the gold standard around here. Right here at the top, there is a view where the intelligence levels of various different AI models are being compared with one another. So you are always able to see right here which of the AI models are currently in the lead. You can of course also find the links to

7:12the sites in the video description. What is even more interesting, however, is this specific Intelligence vs. Cost Index. Here you can see how good the models are compared to their price. Another thing I really like to look at is the Deep SWE benchmark. This measures how effective the new models are when it comes to the development of software programs. The trick here, within both of these benchmarks— meaning both Artificial Analysis and Deep S—is that you are provided with the option to specifically select only those models that you actually wish to compare with one another. And then you will see here not only the performance of a model like GBT6 Astra, but you are

7:42also able to see the performance and the costs associated with each of these specific effort settings. And here you can see the problem right away, namely that with Astra, the performance hardly differs between Medium and Max, but the price has tripled in the meantime. With Artificial Analysis, we can see that the results here are also very similar between Max and Medium. With a model like 5.6 Luna, on the other hand, you can see that the effort setting makes a huge difference here. So, if you decide to use version 5.6 of Luna on the Max setting, you will be getting almost the same performance for your coding tasks

8:13as GPT6 Astra on the Max setting, while paying only a small fraction of the total cost. And that is precisely why the two models that I am currently utilizing the most are GBT6 Astra on the Medium setting, or, whenever I am looking to save some tokens, Luna 5.6 on the Max setting. And as I mentioned, we configure them here in the chat bar by clicking on this area, then selecting the appropriate model, and finally adjusting the respective level of effort required by using the effort setting provided right here. I would almost prefer it if you did not take the specific models with you here, but instead understood that you can now use

8:42these two pages for yourself at any time, in order to classify new model releases directly for your own purposes whenever they should appear. Now let us proceed to the fourth setup configuration, which corresponds to the correct approval mode. And so, if we now click into the approval setting located within the chat bar, you will see that you have exactly three different modes available for your selection here. There is the Request Permission option, where it will frequently ask for clarification before it even edits a single file or attempts to access an internet page. And to be honest, that actually hinders you more than it helps you. With Full Access, as

9:10it is stated here, it possesses unrestricted access to the internet as well as to all of the files located on your personal computer. I would definitely not recommend that. For me, the sweet spot in this case is this specific middle setting. It will only ask you for potentially unsafe actions. So do not choose that upper setting,

9:25because it is much too slow for this. The lower one is much too dangerous. The middle setting is the best sweet spot here. The next setup option relates to the fact that you are now able to have multiple AI agents working for you at the very same time, which is quite a convenient new feature. I know it sounds fancy at first and many people will tell you, "Oh, I have 10 AI agents working for me all the time," just to make it sound impressive. That sounds great, but it is substantially more glamorous than it ends up being in reality. Then take a look: If I start a

9:48new chat here in Chatt and I write the following: "Check my personal website and find ways to improve it", and then I send that off, and I subsequently start a second chat here and this time I write: "Please check the code of my personal website for any vulnerabilities", and I send that off as well, then you will see, if we now open up this sidebar, that both of those chats are currently displaying that specific loading bar that you can see right here. That means I can now truthfully tell someone that there are currently two AI agents out there who

10:13are busy optimizing my website for me. So that's all it is in the end. Having multiple AI agents running is nothing out of the ordinary. It simply means that you are assigning a new task in the chat before the previous one has been fully completed. So the next time someone brags to you about running 3 AI agents simultaneously, I would strongly advise you to exercise a certain degree of caution in that situation. But if someone tells you that they are running 50 AI agents at the same time and can explain to you how they do it, then you can definitely learn a lot from that person. So what I just showed you is

10:41having multiple AI agents work for you in parallel. This will become the most normal thing in the world as you become better at AI, because the better you are at guiding AI, the more likely it is that AI will work longer and more independently, and you will then simply have five, six or maybe ten chats running in parallel. This will transform the way you work in a variety of ways, and it is highly likely that no knowledge worker will be able to

11:00remain completely unaffected. Personally, I believe that we will all become managers of AI agents to some extent. That is why I would like to give you three tips that are particularly important for managers of AI agents. First, it is important to know how to parallelize tasks. In other words, you need to know what the goal is that I want to achieve. Which tasks can I now assign to different AI agents in such a way that they do not end up blocking one another? What tasks does a single AI agent have to do? To achieve this, you require a clear overview of your goal, and you are then ultimately

11:26the person who is responsible for the division of all the work. Second, I would recommend that you always find a way to quantify the results produced by your various AI agents. Unfortunately, AI agents are very bad at providing you with clear statements about how well or poorly a task has been completed. That is why I generally have everything ranked on a scale from 1 to 10. So, as a rule, I do not simply ask: "Hey, how far along is this project?" Instead, I would say on a scale from 1 to 10, just how close are we to actually finishing this project? Or I do not say, explain

11:51the three approaches to me, but instead I say, evaluate each of these three approaches on a scale from 1 to 10 and provide a justification for the whole thing. This saves you a significant amount of time at work, as you no longer have to spend your energy trying to interpret how serious the AI agent is being in any given moment. Third, and this is now the final tip from this setup, you should never spend an excessive amount of time attempting to fully understand the intricate work process of your specific AI agent. What I usually do is read only the first and last paragraph of the AI agent's output

12:15, because that is where, as a general rule, the most significant amount of information is contained within the text. And if it isn't clear to me within 10 seconds what exactly the AI agent wants from me right now, then I will go ahead and ask a follow-up question. To go into the specific techniques of an AI manager would, at this point, likely go well beyond the scope of this particular video. For now , it's important to me that you take away the basic idea of seeing yourself in the future as the manager of your various ChatGBT instances, assigning tasks and coordinating everything. With the ChatGBT setup, we've done that so

12:41far. Now we come to the second step, and these are the AI functions that are currently taking the most work off my hands and helping me to work in a significantly more productive way. In my view, these four functions possess the greatest potential to fundamentally change the way that you work with AI starting from today. Now, the first and most important tool, upon which more or less everything else is fundamentally based, is what we call the Second Brain . I believe the biggest and most significant difference, regarding why ChatGPT in the browser could not assist me with the move, while my own personal AI system knew exactly what to do, was

13:09and still is the second brain that I am currently using. You have to imagine that AI agents like ChatGBT or Cloud build up a collection of memories about you over the course of your various conversations with them. As of now, you can imagine it to be something like a mining system in its current state. So, it is a series of various texts written about you and your work. However, this knowledge is extremely rudimentary and it is not at all helpful when it comes to supporting you in the tasks you perform during your daily work. A second brain, by contrast, is a comprehensive knowledge repository that is owned by you. With it, they then

13:35know precisely what you are currently working on, what your current projects and goals are, which decisions you have made, or how you perform certain tasks. In this way, AI agents gain a significantly deeper insight into your activities and develop a much better understanding of the work that you perform every single day. I have mentioned second brains again and again in the last few videos, because it has become my number one AI tool. The longer I continue to work with this, the more effective it becomes, and I am convinced that everyone should build a second brain now to work more productively with AI. Based on my own

14:01personal experience, I can tell you that the more comprehensive and high-quality your second brain is, the more helpful the resulting AI agent will be for you. I have been working with my second brain for about 4 months now and I have simply noticed that the more comprehensive the second brain becomes, the more work I can get done, because I am making it easier for AI

14:17agents to understand my work in detail. At this point, I would like to take the time to show you how you can build yourself a second brain, in case you have not done that for yourself before. First, you need to understand that at the end of the day, a second brain in its simplest form is really nothing more than just a folder on your computer. In my case, that is the second brain folder. Theoretically, I still have a private second brain, but this is my professional one, and it documents almost everything I do all day long. This folder is also being automatically maintained by both Chatbt

14:42and Cloud as well. And here, everything is now stored: what I am currently working on, what my latest goals are, and which tools and topics I occupy myself with on a daily basis. And it keeps on growing the more I work with it; this specific tool that I use to visualize my Second Brain, that is exactly what Obsidian is for me. That is not at all necessary, but people enjoy using it because it allows you to visualize the Second Brain in a very beautiful and effective manner. This Second Brain stores my entire business knowledge, and you really should also build your own Second Brain, so that

15:05you can complete significantly more work using AI from this point forward. By using the link provided in the video description, you gain access to a specific setup prompt that allows you to easily configure your own first personal Second Brain. That means you take this PDF, copy the prompt and then paste it here in ChatGBT Work. I'll keep updating it. It is also best to ensure that you create a new Second Brain learning folder for yourself here , and once you submit the whole thing, the system will guide you through the process step by step to help you set up your own Second Brain. Once you have

15:32successfully set up Second Brain, your ChatGBT workflow will change in a fundamental way, because from this point forward you will always be working within the Second Brain folder whenever you are using ChatGBT Work. That means that whenever we start a new chat here, we go to Select Project, then New Project, and then we navigate to our Second Brain folder and click open; we might have to create a new project here first, and then you will see that this Second Brain folder is automatically selected for us. This means that ChatGPT is now able to see all of the files which are located here inside this Second Brain folder. That

16:00is exactly the point of it, that it can now tap into this knowledge accordingly . And no matter which task you are currently working on, it will be able to support you much more effectively because it has access to more context and because it understands far better why it should be completing this specific task and what you actually need the result of that work for. And the great thing is that with the setup prompt, Second Brain is set up in such a way that it updates itself automatically every time. So you don't have to do anything except start working in ChbT Work from now on and

16:24add this Second Brain. Second Brain will update itself automatically. Please take 5 minutes at this point. Set up your Second Rain. It is likely one of the most important measures that you can possibly take if you intend to work with artificial intelligence on a professional level. If there's only one thing I really want you to take away from this video, it's setting up your Second Rain. And if you happen to need any help with that, you can find my entire video on Second Rains right here above in the information box. The second feature is currently saving me more time than any other, and it is a

16:51tool that, in my view, is also being completely underestimated by everyone. It's the Google Chrome MCP plugin. To put it simply, this means that ChatGPT and other AI agents can use this plugin to log into your existing Chrome browser and operate the various programs located there on your behalf. That might sound a bit strange at first , but it just means that it takes screenshots of your browser and then decides exactly where it needs to click . As of now, this feature is only available in the desktop apps. So, if you are using ChatGPT within your browser, it is not able to control the browser for you in that way. But why is

17:19this so great? Chatt is already capable of connecting with other applications by using the plugin function, which you can locate up here in the sidebar of the interface. So, if we take a look at the Plugins section, you will see all the various tools that ChatGPT is now theoretically able to connect with at this moment. In Cloud, these are all called connectors, but they are basically the same principle. The problem is, we are noticing that only very specific apps are available here, primarily American applications that have been officially approved for use by the company Open AI. So, Gmail, Google Drive, SharePoint and the like, these are all very useful plugins, but

17:46what about all those other various systems that we are working with on a daily basis every single day? In my case, for example, I have to handle customer support within my sales platform, I must input receipts into my Lexware tax software, and I also maintain a dedicated CHM system where I manage all the clients for our implementation agency. These are all applications for which there are of course no such plugins available, and it is precisely for these kinds of cases that the Chrome plugin is ideally suited. To ensure that ChatGBT is able to access your browser, it is necessary that you download the ChatGBT browser extension onto your system one single

18:14time. I will naturally include a link to it in the video description as well. Furthermore, you might also need to enable computer usage settings within ChatGBT Work as well. You can find it by opening the sidebar, then clicking on your profile at the bottom to enter the settings. From there, navigate under Integrations to computer usage, and once you have reached that section, you need to make sure that the Google

18:31Chrome function is currently activated. So, let us say, for example, that I would like for ChatGPT to make use of my link profile for this task. For this specific case, there is currently no direct plugin for ChatGPT, but I can instruct it to open LinkedIn in the browser and evaluate my last five collaboration requests for me. And when I send the whole thing off, it then informs me that it is opening up LinkedIn right here for me to see. And that is exactly what happens here. Then he navigates over to the messages tab here and proceeds to click through all the different requests one by one. We

18:56can also see directly here, while he is working, that he has the browser tab open right here in front of him on the screen. And it is here in Chrome also opened up as an independent Chat GPT tab for you to see right now. You can always tell that JGBT is currently working in this tab by looking at the pointer icon right here. And now he is finished with his task here, having rated the last five link requests on a scale from 1 to 10 and then provided a corresponding recommendation. This means that our Chgbt can now effectively use our LinkedIn, even though we have never added any kind of

19:21plugin directly to it ourselves. For example, what I can also do is book the Sixt car for Schzug, but I additionally require one for October 2nd in Munich so that I am able to pick up IKEA items and various other pieces of furniture. Could you please compare the current prices and then click your way through the Sixt interface, so that all I have to do is confirm the entire booking at

19:40the end. Okay, send the whole thing. Okay, and now look, this time he opens the browser here directly in Chat GBT. So, this here is a window that I am able to drag to be larger or smaller, and it is now clicking itself through the Sixt mask entirely on its own accord automatically. We could also make the whole thing smaller, and then it will do everything automatically in the background for us. I want to show you how it does it. As you can see, it automatically sets everything up for us . It is currently in the process of selecting the most suitable van for this task. It clicks through here and

20:04automatically enters the various information. It has this directly from the Second Brain, of course. After a minute, it is finished. He has now directly entered all of the information for us in here. So all we have to do now is enter the credit card information, double-check whether this data is correct for all of us, and then confirm everything. In the past, I also used to communicate with my tax advisor

20:24exclusively through the use of email. In the meantime, they have established their own system, and now ChatGPT is able to connect using the browser function, which allows it to handle a large portion of the communication for me. This Chrome plugin is extremely important because it is what finally allows ChatBT to gain access to the various programs that you are working with on a daily basis. And together with your Second Brain, this makes for the ideal combination, allowing ChatBT to take on more of your workload in the future for you. Now we arrive at one of the most underrated features, and that is the remote access functionality

20:52which is available on your mobile phone . Because now you have to imagine that here on our PC or laptop we have more or less configured the perfect setup with JGBT Work, with the desktop version, with Second Rain, which possesses all the knowledge, and with the Chrome plugin, through which the artificial intelligence is also capable of accessing our programs directly. The only problem is that we are not always sitting in front of our laptop or PC, and this is exactly where the remote access function comes into play. You can use the ChatGBT app on your mobile phone to log in to your existing ChatGBT sessions that are currently

21:21open on your computer as well. The connection works as follows: first, open the sidebar at the top left, then navigate to your profile and into the settings, and finally, under the Programming tab, you simply select the option to Connect. As you can see, at this point I have already gone ahead and connected my iPhone here. And what you will do next is go ahead and click on the Add button right over here. And now you can see that this QR code is being displayed right here. You then need to scan it once using your mobile phone camera. This will then establish the direct interface between your

21:44mobile phone and your PC. And if you now open the sidebar in the ChatGBT app on your mobile phone and then select the Remote option, we are now able to see every one of the chats that we also have available to us on our computer at the same time. That means I can now go ahead and open the Second Brain folder and then navigate to the chat where we were just evaluating those LinkedIn collaboration requests that we received . And now I can write here on my mobile

22:03phone, what is worth pursuing further? And just take a look: if I send the whole thing now, you will see how it gets synchronized here and also sent directly to this specific chat on the laptop. I personally find this function quite brilliant, because it means that you could, for example, take a walk around your office or perhaps not even be at your laptop, and JetGBT can continue working on your computer for you while you remain in control of the

22:22entire process using your mobile phone. For example, what I like to do is to keep the laptop connected to the power cable at all times, so that it does not automatically switch into standby mode. Most of the time, for example, it happens that I might be out for a walk or otherwise away from my computer when an email from a customer comes in. I then instruct ChatGBT via my mobile phone to log into the CRM, perform some research, revise some of the code, and so on. And you will also notice over time, as you work more extensively with the Second Brain and the Chrome plugin, that the desktop version of ChatGBT is

22:47significantly more powerful than the mobile alternative. Unfortunately, those two functions are not available for you on your mobile phone, but as you have just seen, you are able to log into your laptop by using your mobile phone as a key. With that, I want to show you how you can automate your tasks. And the most important function

23:02for this is the scheduled tasks. Scheduled tasks are simply recurring automations in ChatGBT. That means, up until now, we have always worked directly with ChBT. As I said, the Second Brain and Chrome are able to perform the work much better, but up until now, we have still been required to provide every single instruction to it manually. With scheduled tasks, you can set up an automation in Chetchbt where the system will consistently execute a specific task for you at a pre-determined time that you have defined. As of now, I still have many

23:28of my automations running in the cloud. But it works exactly the same way there . As of now, there are around 20 automations running for me, and I have generally noticed that the more boring and dull an automation seems to be, the more helpful it usually turns out to be in practice. One automation that relieves me of a lot of work is, for example, a YouTube archive routine. So all that this routine does is, as soon as I have uploaded a new video, it archives the transcript, the title, the metadata and so on, and then it saves these files once in my Google Drive account and once in a separate local

23:55folder that is located on my personal computer. Or this routine right here, which is more or less the entirety of my bookkeeping. Claud logs into my tax software, where he pulls the latest transactions from my various accounts and checks immediately to see if he has the appropriate receipts on file. If he finds that the necessary receipts are still missing, he either pulls them from Gmail or logs into each of the individual apps using Chrome MCP to download the receipts directly from those sources for me to have them all in one place. The whole thing is then imported into the tax software once, and I also maintain a mirrored version

24:18locally stored on my personal computer. The whole thing then flies in there too . So as you can see, scheduled tasks are an extremely important function and it is essential that you learn how to handle them properly in your daily workflow. Fortunately, it is quite simple to create a scheduled task for yourself, such as a ChatGPT for this specific purpose here. All you have to do is simply tell ChatGBT that you would like to have a certain task performed by the system at specific intervals of time. For example, you could say here, please greet me every single morning at 7:30 a.m. with the specific words hello Jonas. And when I

24:44send the whole thing off, what it does is that it proceeds to set up the daily greeting for you just a single time. We can see that here as a morning greeting , and you can view all of your scheduled tasks if you open the sidebar once and then navigate to the tab that is labeled as planned to see your overview. Here we can now see our morning greeting, and the description of the planned task is set as follows: Greet Jonas during every single planned run by using exactly this specific text : "Hello Jonas." With this, we can now be absolutely certain that this scheduled task is going to be carried

25:08out on a daily basis every morning. That was of course just a test, which means I can go ahead and delete these items here now. You want to create planned tasks now that are meant to genuinely take the workload off your hands. That's why you have to provide the appropriate context here. What I prefer to do at this stage is to manually work through the entire task with the AI, subsequently verify the results, and iterate on them. Once Kim is able to demonstrate that she can handle this assignment independently from start to finish, only then do I instruct her: "Okay, using the information provided in this chat,

25:34please set up a recurring scheduled task where you carry out this entire process autonomously on a daily, weekly , or monthly basis, depending on what is required." So the simplest rule when creating planned tasks is this: make sure to work through the process manually with the AI at least one time. Ensure that it has mastered the task, even when dealing with various edge cases, and then you are usually ready to say in the chat: please make this a planned task now. So, let us now move on to the very advanced techniques of AI automation, which probably only a few people have on their radar, and for

26:01which I certainly do not expect that you will immediately implement them for yourself here. So, you are now aware that we can already automate quite a lot with the planned tasks, but what happens if we process sensitive data within that automation that we cannot simply pass on to ChatGPT without further consideration of the security implications? In that case, I would say that there are two possible ways to still create this automation with the help of ChatGBT. And these two different paths depend entirely on whether you want to use AI within the automation or not. So, the first option is: if you only require software without any AI, you can have ChatGPT

26:29write the script for you, and this script will then run locally on your own computer, which ensures that your personal data never actually leaves your machine to go to ChatGPT. In this case, ChatGBT is simply the developer who sets up the entire structure for you, but never actually sees the data. The second option is that you also need to implement an AI within the automation process that is currently handling your sensitive data. In that case, I recommend that you use a local AI, so that the workflow again runs only on your laptop and ChatGPT, which is hosted on US servers, never gets to see the actual data that you are

26:57processing. Also, the key is always that ChatGBT creates the structure for the automation, but never sees the data itself. Let me show you a short example for both of these variants right now. The first variant, as I already mentioned to you, is that ChatGPT only writes a script, but you do not need any AI for the actual automation process itself. This is the case, for example, if you want to create invoices from sales contracts. So, I've prepared my folder here. There are now two

27:19examples of sales contracts in here. They look exactly the same in both cases. In this specific case, these are merely demo purchase agreements, but ChatGPT is tasked with building an automation from them. This uniformity means, in this particular instance, that we are able to simply write code which serves to automatically generate invoices derived directly from the information contained within these various purchase agreements. So what we can do is take this invoice automation folder and drag it into ChatGBT. And now I can go to ChatBT, for example, and say that I want to build a small tool which automatically creates invoices for us, directly from our existing sales contracts. There are two

27:51examples in the folder, without any real customer data yet. The structure will be exactly the same later on. When I send the whole thing, ChBT knows, okay, we're working with demo data here , but the structure of these sales contracts will be exactly the same later on in production. So now he simply needs to write the code here one time, in order to automatically extract all the data from the sales contracts and then insert it directly into a fixed invoice template. And he can now use the demo data to prepare everything , including the code, the invoice template, as well as the error handling

28:18and the user interface for the system. And the actual app, which ultimately processes the data at the end, is one that we run entirely on our own local computer system. Chatt was then just the developer. Theoretically, we could then also turn off the Internet, because in that case all data would be processed exclusively on the local PC instead of online. And now he's finished. This means that he has built this little tool right here, in which we are now able to input our sales contract on the left side of the screen . We can try that out too. Yes, and as we can see, the system has now also

28:41gone ahead and automatically extracted the contract data for us here. And this tool now runs entirely by way of code only on our local computer system. This has nothing to do with ChatGBT anymore. I think you can tell from the speed. And if I now click on Create PDF, the system will generate the corresponding invoice here, which we are then able to download directly to our computer from this location as well. And of course, we are still able to customize this entire application by using ChatGBT here. It’s about how the invoice looks in the end and how the app functions. It can also help us by

29:04allowing us to open this app directly with just a single click on an icon. It can do all of that here. The point that I simply want to show you here is that although ChatGPT is indeed the developer of this application, it never actually gets to see any of the real data. So that's the difference. ChatGPT has now developed this software, which we can use offline and theoretically without an internet connection, and all the data that is processed within it will never be transmitted to or stored on any servers located in the United

29:25States. So, that's the first option. You simply need the code and can then let ChGBT develop this specific automation for you. I recommend that you work with demo data in this case. Provides ChatBT with the entire structure so that it can build the automation cleanly for you and you can then test the whole thing locally on your computer. The second option, and this is where it gets a bit more complicated, is if you want to process sensitive data with the help of AI. In other words, you want to integrate AI into your local automation. As I said, what you need here is so-called local AI. So imagine, for example, that you

29:55want to evaluate your meeting notes, but you also want to ensure that the data does not end up on the servers of ChatGBT at the end of the process. Then ChatGPT, acting as a powerful AI, can provide assistance to help you set up this small local AI system. It can provide you with the prompt and the corresponding environment, and can build an entire web app around it without ever having seen a meeting. So ChatBT is the developer here again and everything is ultimately processed locally on your computer. So the AI that ultimately evaluates the meeting also runs locally on your PC. In this case, I would write to ChatGBT here,

30:25please create a small tool where I can enter a meeting transcript and a local AI model creates a helpful, concise summary from it. So, send the whole thing off. And basically it's the same game again. ChatBT will now handle the entire setup for us, installing and configuring the local AI model, and it might ask you a few additional questions. Ultimately, it will orchestrate everything so that a program runs on your computer, providing you with a local AI system that is fully capable of processing all of your meeting notes automatically. If you are feeling unsure, I would always recommend that you simply switch off your internet connection and then check

30:56whether the program continues to function as normal. Then you can be certain that it is running in a one hundred percent local environment. So, he's done again. I would say let us take a look at what kind of an app he has created for us. Just a quick reminder: ChatGPT operates in the cloud , but this software, specifically this app with its local AI model, is currently running entirely and

31:13exclusively on our own computer system. I can now enter a transcript here on the left, select the local AI model here, and set the scope, i.e. to the point or with more context. Let's get to the heart of the matter here. And if I click on summarize, it will start formulating the summary. With the local AI model, it may take just a moment for this to process. And then it's done. So now the local AI has generated the summary here, and we can see our summary on the right-hand side, which we are able to copy directly or alternatively save as an MDEI file if we would like to do so. So, as you can

31:44see, this app is working perfectly now. Even with the local AI, everything worked. I could have turned off the internet now. Everything would have worked exactly as it did. And what I think is cool is that this application was a one-shot experience. As you can see, I simply entered the task and then it came up with the app. So it's really impressive here. And these are the advanced applications of ChatGBT, ensuring you know how to build automation processes even when sensitive data is involved in the workflow. If you are currently planning something similar for your company on a larger scale, you will find a contact form on my website where you are

32:15welcome to reach out to me, preferably under the section for AI Agents and Automations. Congratulations on making it this far in this video. We have now covered the most important points. You have an AI Burgflow that will bring you noticeably better results in your everyday life. And if you are now saying: "Hey, you find this interesting , but you do not quite understand how to incorporate all of this into your daily work yet," then I have something for you: the AI Compass. This is my AI product developed specifically for professionals who want to integrate AI into their tasks. In over 6 hours of video material and seven structured

32:43modules, you will learn in much greater depth what we have discussed today. For example, you will learn how to work correctly with skills, connectors, and of course, second brains. So, if you no longer just want to try out AI, but instead want to incorporate it into your work using a clear system, then I highly recommend that you click on the first link provided in the video description now and secure your own AI Compass. With that, I would say that we will see each other again in the next

33:02video. Take care, until then. Tiao.