Transcription
Most people are still using AI as a consultant. You get the advice and then you go do the work. That's already changing and most people haven't even noticed. Cloud Co-work can now open your files, build your documents, and update your systems. I've built a three-level setup where I drop in a file, walk away, and I come back to finished work. No prompt. It just knows what to do. Today, I'm going to walk you through that exact system. And by the end, yours will actually get smarter every single week on its own. So, let's get into it.
The primary shift that's occurred with Claude Co-work and Codeex when compared against ChatBT, Claude and Gemini is that those two tools, they can not only just read things, but they can actually go off and take actions on your behalf. Because when using ChateBT, Claude or Gemini, they can read files that you give it and even sometimes read stuff from other systems you connect them to, but they're usually limited to just reading. But when you use something like cloud co-worker codecs, they can read and write files on your system as well as in other systems and that's the primary driver of value.
Now what I want to do is I want to walk you through a simple framework. There's three different ways that you can use cloud co-work or tools like it. And each one of these build on each other and the value increases as you move up the levels. I'll walk through each of these in detail, but as a quick summary, we start with do. So do is simply going to cloud co-work or codeex and giving it something to do and it takes that action for you. This is somewhat ad hoc. The next thing is make where you're giving the task to cloud co-worker codeex and all you have to do is drop in a single input and what the AI does is it goes off interacts with multiple systems and then brings back to you finished work and then finally the most valuable of these three is no and no contains everything from one and two but in addition to that the AI also remembers different things from previous sessions. So you're transitioning this from a tool to an asset because it builds insights over time on reoccurring conversations, activities, actions, functions, etc.
Now, before I dive into the use cases and walk you through each one of these examples, I probably should show you what cloud co-work looks like. Okay, so if you're enjoying this, you're going to enjoy two other things. First off, Blow is a 30-day AI insight series, completely free. You'll get 30 insights in your inbox so I can apply AI to your business and your work. The second thing is if you'd like to work with me, blower a series of offerings to see if there's a good fit between the two of us, such as a private community for business owners and leaders or private one-on-one coaching. With that being said, let's get back into the video.
So, here we have Cloud Co-work. So, if you've downloaded the Claude desktop app, you probably land on something like this when it gives you the basic look that you would see when you use Cloud in the web app. But in addition to this, you have two other tabs. You have co-work and you have code. We're going to talk about co-work. So, if I go into here, you can see things are slightly different but very similar. And if I expand the menu here, you can see we have three options. We have new tasks, search, and plugins. Over here, we have folders. And this is really important. So, anytime you use co-work, you want to make sure you're using in a specific location on your computer. So, you have folders on your desktop somewhere, you want to open up one of those folders for the AI to take those actions. And to do that, all you need to do is select the drop down and then go choose from different folder. And you can choose any folder that you'd like on your computer. In addition to choosing the location, the next thing we want to do is potentially add files and/or most importantly have connectors to this. So connectors allow the AI to reach out to other systems, not just reading, but writing to them as well. And an important caveat here is that all connectors aren't created equal. Some allow you only to read, some allow you to write. So you have to actually go to the manager manage connectors button here to read through what's possible. So if you go to browse connectors, you can see when clicking into each one of these, it gives you a description of what it can and can't do. So in notion you can see that it allows you to update things mean meaning that it can write to notion and when choosing your model for the task that you're working on if it's a complex task with many steps you probably want to use the highest end model available and right now that's opus 4.6.
So now that we have a general understanding of how to get started with cloud co-work I'm going to show you some use cases and we'll learn more about cloud co-work as we work through these use cases. So the first one is do. So this is the first level of value and in this use case what we're going to have is our miscellaneous folder. So, I'm sure many of you have a folder like this where you have a bunch of receipts, contracts, PDFs, images, all types of things you've scanned and you've put into a single folder and they're likely named horribly based off of whatever device you pulled it from. So, it could be image some random number. It could be document parenthesy number parenthesy or a variety of other things. Most of these titles mean nothing to you initially and you have to go through and change all the names of these titles over time. I personally just went through this for taxes. So, I had a bunch of documents for taxes and they're all named weird things because I scanned them all in. So, what I did is I dropped them all to a folder. I gave it to Cloud Co-work and I asked it to basically organize and rename every single one of the files based off of what's in the file. So, it then went off and put all my invoices in different files, all my contracts in different files, and all my receipts in different files. And they all had a very specific naming structure that I prefer. And the way to do this is very simple. All you have to do is have a basic prompt like this, put it into cloud co-work, and let it do its thing. So the prompt has a clear structure. At the very top, we have a role. In that role, we're giving the AI the role of a file organization assistant. After that, we've given it a clear task of processing all the files in this folder that I've opened. And I'm asking the AI explicitly to read the content inside the file, not just the file name. And then what I want you to do is I want you to rename each one of the files in this specific format. So we're going to do date, type, description. And then then I give it clear instructions on what the types are. I ask it to sort it by subfolders. And then if a subfolder doesn't exist, go ahead and create it. And then in addition to the task, I also give it specific constraints, things never to do and or to avoid. So the first constraint is to never delete any files, only rename them. And importantly, I ask it to log every single rename action into a change log.txt file. And having the AI create a file like this is very important, especially for really big tasks. Reason being is the AI's head is limited in space. It can only hold so much. So when the AI's head fills up, what happens is it compacts previous conversations and it summarizes previous things and keeps going. And what we want the AI to do is always log everything that it's done in this file. So when it's in memory gets wiped and it keeps going, it can check this file and see exactly what's been done and not redo work and work on the very next task that's available. And in this case, that's going to be renaming files. So it'll see all the 26 files it's renamed and it sees that there's eight more to go and it only works on those eight and not the 26 that it's already done. And for this use case and the other two use cases, we're going to use a really important file called cloud.MD. But the thing is, you've likely used this in other tools not knowing. So if you've created GPT projects, if you've created custom GPTs, cloud projects, or Gemini gems, you've likely created system instructions. These are instructions that the AI looks at before it ever interacts with your user, making it a tailored AI to a certain task. And let me show you what this looks like in Cloud Co-work. So if I simply just say test here, so I can get into a different screen, you'll see a few things. So on the right hand side we have a progress and a context and a testing memory folder. So this is the folder that I'm inside of. And then right here my instructions. So if I click this you can see that these are our folder instructions. These are the things that the AI looks at before it does anything else. And this is also the same exact prompt that I just showed you. And the reason this is valuable is that anytime that I come into this specific folder within cloud co-work and I've already dropped in files that need to be renamed, all I have to do is type go and it'll know exactly what to do because it looked at these instructions. And the reason I said that cloud.md is important here is because you can see it says cloud.md. If I actually open this folder and go here, you'll see that I have this cloud.md file here. This is automatically created by the system once you add any text into here and select save. And the AI knows to look at this file anytime it starts any task. So that's an important feature to know about.
Now let's move on to our second level which is make. And I'm sure this use case is something that many of you are going to be excited about because what we can do here is we can drop in a file. And in this case, we're going to drop in a transcript from a reoccurring meeting with the client. I'm going to drop the transcript into cloud co-work. It's going to do a series of things for me without me having to do anything. So after I drop this transcript in the cloud co-work, it's going to go off into a series of systems, find information, as well as create a series of files for me based off of the instructions. And in this use case, we're going to have a recoccurring meeting with a client. So, we meet with them every single week. And the first thing it's going to do, it's going to go off to Gmail and see, okay, have I sent any emails to this client in between those sessions. In addition to that, also, if I wanted it to, it could go to a CRM that I have. So, it could be HubSpot or something else and see if I've updated anything in the CRM in relation to this client in between those sessions. Now, the question we want to ask oursel is, how can the AI do this? How do I set this up effectively? Well, this goes back to the point that I mentioned previously of data connectors. With Cloud Co-Work, we can connect it to our calendar, our Google Drive, our Gmail, our CRM, etc. And as long as these connectors out of the box allow you to write and read different things to their systems, you can then attach them to your cloud co-work and sure they're enabled and allow the AI to do what it needs to do.
Now, let me walk you through the instructions for this specific use case. So, as always, we have a role. So, the role for this AI is a postmeating processing assistant. Its purpose is to handle any follow-up work after any client meetings that are reoccurring. And the task is detailed below. So I'm saying when a meeting transcript is given to you, I want you to complete these steps below. And I'll walk you through these briefly. So the first thing it needs to do, it needs to go to my Google calendar and check to see when the meeting occurred and pull all the attendee emails from that event. After that, I'm going to have it actually do a variety of things simultaneously. So the first thing it's going to do is it's going to draft a follow-up email. There's going to be a certain tone. There's going to be a certain structure, writing level, etc. associated, as well as the inclusion of the emails inside of the event above. so I can easy easily copy and paste that into Gmail. In addition to the email, it's also going to do a few other things. It's going to create an Excel sheet that covers off on the action items that I need to do and others need to do. So, it's going to have certain columns, statuses. It can have conditional formatting and all types of things because Cloud Co-work can build those into Excel for you. And it'll also draft an executive summary in a very specific format that I prefer for PDFs, including certain information that I care about when it comes to the sections relevant to that conversation.
A few important things that I'll call out here in the constraints area. One important thing I've listed here is that everything that the AI gives me has to be strictly from the transcript and nowhere else. So the AI can't fabricate any information. The other interesting thing here for the constraints is I've been explicit in this last statement asking the AI to create multiple sub aents to do all these tasks simultaneously. So what do I mean by that? That's it's a lot of jargon there. What is a sub aent? Well, sub aent is basically just a baby a AI. So we have a parent AI here in cloud co-work and it basically spawns or creates a bunch of baby versions of itself. So, we have one version here focusing on email, one here on the spreadsheet for action items, and one here for the PDF summary. The benefit of doing this, of having multiple sub aents working simultaneously. There's three benefits. One, it's faster. Two is that the AI's context window isn't filled up immediately because each one of these are going to have roughly 200,000 tokens of context that they can use for each one of the tasks without them being filled up immediately, which correlates to higher quality outputs from each one of these AIs instead of trying to do it all in one AI. And the third valuable thing here that I found when running sub aents is not just running it for this task, but running it for research. Say you have a really important research task that you want AI to take on and you want to make sure that it has a really inclusive perspective on researching this. Well, you can give it a question. You can say, I want you to spin up five to 10 sub agents and research different aspects of this question. So when it gives you a response back, you know that it's thorough. And let me actually show you what this looks like inside of Cloud Co-work. So inside of Co-work, I dropped a fake transcript. And here the AI started going off and doing a bunch of work for me. But if I scroll down here and I expanded this drop down here, you can see that we have these little icons. They're kind of like squiggly lines with circles in them. These represent sub aents. So each one of these sub aents are working on a different task. You can see this one's writing a follow-up email. This one's creating the action items in the Excel sheet. And this one's writing the summary PDF for me. And also you can see previously the primary task that occurred first is the AI looked at my Google calendar before doing anything. And that's what sub aents look like in action inside of Cloud Co-Work.
We've done do, we've done make. So now let's do no, which for me is probably the most valuable of the three use cases. And it's valuable for this very specific reason. We're taking AI and transforming it from a tool to an asset that compounds in value over time. And the primary reason this happens is a little file that we have the AI create, which is a memory file. The use case I'm going to walk you through is a reoccurring engagement that I have with a client on a weekly basis. But you can apply this for not just engagements for conversations, but also you can do it for functions, for activities, etc. basically anything that you do on a recoccurring basis that if the AI had compounding knowledge about would benefit you in the future and that's a lot of activities.
So I mentioned this memory file and how this acts as the primary catalyst for an AI tool going from a tool to an asset. So what is this? Well, it's simply just a text file and we're going to have the AI incrementally look at this and also add to it over time every time I engage with the AI in that specific folder. And the use case I'm going to share with you like I mentioned is a recoccurring meeting I have with a client. So here we have some fake inputs where we have the AI noticed that the client prefers bullets in one of the first sessions. The AI also noticed that the client has a budget cycle reset happening in March and knows that the client prefers speed when it comes to onboarding. Then also it actually gave us a critical insight that I'll talk more about later. So how does this actually work? Well, here's the system instructions and we're actually tacking on to the previous instructions that I drafted previously. So in these brackets here, the instructions that I showed you for level two, these are the same. they stay the same. What we're going to do is we're going to bookend additional instructions on top of that prompt on the beginning and the end. So the beginning edition that we're going to add is before the AI does anything, it needs to first read the memory file because these are all the insights the AI has derived over time. After it's read the memory file and it's done the tasks, immediately after that, it needs to update the memory file with any additional insights it's derived from this specific action it's done for the session I've engaged with the client on. And I'm specific on what I want the AI to add to the memory file and what that should look like. So I'm talking about client preferences with communication, any reoccurring themes that it's come up with, key decisions that were made in the conversation, and an important constraint down here is never to remove any previous entries, only add stuff to it. So this is our basic prompt. What does it look like inside of Cloud Co-work? Well, the example I just showed you here showing you sub agents is actually where I've run this multiple times as an example. And you can see here we have a memory file. So when I open this up and I scroll up, you'll see that we have multiple inclusions. So here we have three sessions and here's the recurring themes that it's noticed through time. And if I scroll up even further, you'll see the file is pretty long. And what we can do is we can ask the AI to condense it over time if necessary or just not to include as much each time it adds to the memory file. So here we have session three. Above that we're going to have session two. And above that we're going to have session one. And these are the cumulative insights the AI's compounded over time.
And these are the three simple ways you can start using cloud co-work today. And as a recap, we started with do, which is the simple use case where we just give the AI a task and it goes off and does that for us. And here we learned the importance of setting system instructions. So you can just come in and drop a file and it does the task without you having to reprompt it every time. Similar to system instructions inside of custom GPT, GPT projects, cloud projects, and Gemini gems. After that, we have make, which is taking this to the next level. So, we're not just giving it a single input and having it do stuff for us, but it's going off into multiple systems, reading and writing into those systems, as well as running multiple sub aents simultaneously to have it happen faster. And remember, the important caveat for sub aents is this is only useful when you're doing a task where they don't depend on each other. Meaning, task B doesn't rely on task A to get done. And finally, we have the most valuable use case in my eyes, which is the no category. And that's where we combine level one and level two with level three in addition to adding a memory file. So, the AI actually compounds its knowledge over time and transitions from a tool to an asset.
And that's it. So, a reminder of two quick things. First off, Blow is a 30-day AI insight series completely free. You'll get 30 insights in your inbox so I can apply AI to your business and your work. The second thing is if you'd like to work with me, blower a series of offerings to see if there's a good fit between the two of us. Now, I just handed AI the keys to my files, my calendar, my CRM. And if that makes you a little nervous, good. But most people are thinking about that risk completely wrong. There are actually two risks with AI. And the one that you can't see is the one that's costing you the most. I break it all down right here in this video. So go ahead and click it. Go ahead, internet, click that video. I'll see you next time.