Part 2 | Step by Step AI Personal Assistant Automation Setup! Part 1 @Dean | AI Automation #ai #automations #aiautomation #aitool #aiforbeginners #aiforbusiness

Two, I'm building an AI personal assistant. So in this video, I'm gonna take you step-by-step through how to do that with make.com. So bear with as this is gonna be a fairly long video. So I made a few assumptions before we get started. One of which is that you are familiar with make.com. Second of which is you're using a business Gmail account. It makes the connection process a lot, lot easier. The third of which is that I have a filter set up in Gmail to make sure that only the emails that I want a reply to actually come through into this automation. And the fourth thing really is more note to say that this is the simplest way to set up this automation that I've found. Last thing is you will need an open AI developer account to actually set this up. They're free, obviously you have to pay for the credits. If you want separate videos on either walking around make.com or setting up an open AI developer account, let me know in the comments. Okay, so first we're gonna get stuck into this. So let's click into scenarios, new scenario, and we are in a blank canvas essentially at this point. So I'm gonna go ahead and click on the plus icon here. Then what I wanna do is add the Gmail app. But within that, I wanna add the watch emails module. So as we can see, that's now our trigger module. Now this clock means that it will be set to run on intervals rather than an instant trigger. So what we've done here is we now need to, so we'll have made the connection. You can click add connection, and then you simply sign in with your Google account to do so. Obviously mine's already connected, so that's why it's showing. But that's all you'll need to do. The little pop-up comes up, you sign in, and then it'll look very much like this. Add this folder to all mail because basically it'll go into a labeled section in my inbox, but I want it to search my entire mailbox for it. The filter type is a Gmail filter. As I said, I'm looking to bring up label. Thing that you need to set up here is the label. So again, this will depend on the filter. If you haven't got a filter set up, then you could simply just go with simple filter, and your criteria could be unread emails, for example. You could keep it as simple as that, but for me, it is having the label set up because I'm filtering off newsletters and things on Gmail before they come here. So I have a label called reply. So the code that you need here is label, colon, reply, and that sets it up for me. Then once the email's been fetched, I want that to mark it as read in there so it's nice and easy, and then you can go ahead and click OK. What I wanna do is, for the purpose of this example, I'm gonna click all emails, but you could have from now on because obviously you don't want it to go back through any of your historic emails. The test email that I've got set up for you, I'm gonna click on all emails. Okay, so that's our first module. Then we need to add in the chat GPT module. So we wanna go over and navigate to OpenAI, and then set up an assistant in OpenAI, very similar to a custom GPT, with your knowledge base so you can get more detailed answers. This example, we're gonna use a chat completion anyway because it's simpler than setting up an assistant. So we're going through this one where it says chat completion on the GPT and O1 models. I use the O1, sorry, 4.0 mini system, this one here. Reason being, it's just cheaper, and I find that with repetitive tasks such as like replying to emails and things like that, it's very much a case that you don't need 4.0 even to get a good reply. And obviously the mini is more cost effective. And the whole idea here is that we're keeping the cost down as well as automating as much as we can. So you're gonna need two messages for this part. The first bit is gonna be your developer slash system role. So this is essentially the prompt. This is the prompt field of what we're gonna do. And I'm just gonna grab the prompt, and we're gonna paste that in over here, and I put the prompt in here. So as you can see, I've put, you are a professional virtual assistant helping to draft polite, concise, and relevant email replies. Your tone should be friendly, professional, and conversational. Always acknowledge the sender's message, address any questions or concerns, and provide clear, actionable next steps if applicable. To use British English, because I'm from the UK, do not include a subject line or signature, and format the response in HTML. The last bit about being formatted in HTML is really important because you'll find that if you don't put that in, Gmail will not pick up it as formatted text. It'll pick it up as plain text, and then it'll clump together and make for a very unreadable reply. So that's very important to make sure that we get that bit right. The next bit is that we wanna put the role as user. This is basically gonna be the information we're passing through, so the context from the email. Put some extra context in to make sure that we've got some great replies. So we're gonna put sender name in here, and then we're gonna map the sender name section. So that's gonna be this one here. And then I'm gonna put in email body, and we're gonna map the text in there. So then when you do chat GPT replies in just a normal GPT chat window, it'll come back with bracket name. So if we don't give it the sender name, it will do that, and obviously that's just gonna put a delay in the reply. So this way, by giving it that information and being really clear about this is the person's name, and then this is the email, we'll get a more polished response to a point where it's only slight tweaks before sending. Okay, so the next thing really is the tokens. So I have, because my emails are typically shorter from a lot of people, I set the max tokens to 650. This will depend on you individually, and to be honest, I would just say monitor it and tweak it. But I would say anywhere between 500 and 1,000 is pretty typical for an email reply, put it that way. So that's all you need to do here. Once we've got this bit done, we're gonna click okay there, and we're gonna add in our draft module email app, and then we're gonna go to create a draft. So again, it's loaded up my inbox. So the folder that I like to put them in, I mean, you can choose a different folder, but I just put them in the drafts folder because it's super easy for me. So who's it going to? So we're gonna pick up the email address from the sender from the first email and drag that in there. And then we're gonna go over to the subject line. I keep the subject line the same just because it's easier for me to search in my inbox, like to see all the emails chain, but obviously you can put whatever you wish. In the content, that will be the result of module two. So your chat GBT result, which we'll just drag and drop here. As you can see, really intuitive, like just dragging and dropping. And then that's it for this module. We click okay here. That's now all set up and ready to test. So what we would do here is you could go and navigate down to run once. Now, what that means is it's gonna run the entire automation. Now, just to give you an insight on my inbox here, I sent myself an email, just a test email. So that's marked as unread and it has the reply label applied due to my filters. So basically this module is gonna pick that up. So we're gonna click run once. So then obviously we can see that it's checking for emails. You can see that we've had a response because we've got used one operation. It's now using the chat window and that's been successful because we've got one operation there. And we can see that the draft was successful because we've got one operation there also. My Gmail, we'll just have a little refresh and we can see that the draft has appeared here and this has been marked as read. So if I go here into my drafts, me and email, it's giving me the same subject line and it's replied to by email and it's formatted it exactly how we wanted it to be as well. So that's essentially it ready. I can just read through this and be like, yep, few tweaks and then I can hit send. So you imagine all of your emails coming through on a daily basis and essentially what you can do with this automation is just make sure that replies are pre-written, ready to go so the bulk of your typing is taken care of. Hopefully that's a really useful walkthrough for you. Like I say, step-by-step on how to put those modules and do all the mappings together. If you want any further follow-up videos on how to use make.com or potentially how to set this up in a different way, let me know in the comments and drop me a follow for more.