268 | Get Your Brand into AI Search Engines Without Spending a Dime on Ads with Pietro Montaldo

28:56In this case, the most common type of trigger is a scheduled trigger. So you tell them just like you would tell yourself at 8am every morning to this specific action. This is also the one we will use for this specific workflow. So if I set up a scheduled trigger and I want for this specific workflow to have one trigger once a week. So I want it weekly on Thursday. This workflow will only run every Thursday at the hours which I set, in this case, 8 a.m. So once you have done this, you click done, and you have the alarm set. Now we have to tell him what to do when he wakes up with this alarm. I'm going to add nodes, right?

29:42But you don't need to add them. It's just so that when we build, it's more clear. You don't have to do this stuff. They don't change anything in the workflow. So feel free to ignore them. It's just so people who follow along and see the video can have a better idea what I'm building at each stage. Okay. So the first thing we're going to do is identify the subrack that you want to monitor. Okay. And this is different for each single one of us, you have multiple options. If you have, for example, a Google Sheet where you have already stored this information, you can connect a Google Sheet. What we want to have is a basic list.

30:17How you create this within this platform is you go to utilities and you create a table of constant. What this is is a simple table. Imagine as an Excel, as a Google Sheet, but within the platform so you don't have to get out. And what we're going to add here is two simple colon. Here is subreddit name and here is just subreddit. Okay. Pardon, I'm a bit dyslexic and you only need to check my spelling, but ignore it for now. Doesn't matter. Yes. Again, for those of you not seeing, we're literally just creating tabs by clicking a plus button and adding two different columns inside a simple table that is within relay so there's nothing unique about it it's a table it's exactly a table nothing must imagine as creating a worksheet i'm giving basically i'm just creating a list of the subreddit which is interesting for me okay i created two tab each subreddit and this is just a technicality of reddit is identify as an r slash the name of the subreddit okay and then to have it clean i also want to have it without the r slash because i need to use it later in the flow without the r slash so what i'm gonna do is i'm just gonna create two tables with the name of the subreddit which i want to monitor okay okay these are subreddit and i'm using product marketing and product hunters which are interesting for b2b sas company right each single one of you before setting this up we look for the subreddit which are interesting for them.

31:58We have identified the subreddit. Again, don't do it now. Now use sample once, but do it. This is an important search. You do it once, it's done. You check it maybe once a year that you're still following the red-side edit and then it's done. The second thing we want to do is we want to give information about our company. In this case, we're going to use a very similar step. We're going to insert another table. We're going to do the same exact things another time. This time we're going to insert our company link, right? So the first thing, and I'm going to insert my company name. Then I want to have my company URL, my website basically.

32:42Then let's say we also have a blog, a lot of companies have a blog. I also want to have my company blog. I basically am trying to give as many information as possible about my company right and i'm also going to give my linkedin url okay so i gave all the information which are publicly available on the internet about my company if you have file you can also upload files right this question would you would you also write like a like you do from a context perspective like when i do anything on any ai i have like a full document that explains about my company, what we do, what services I provide, who I work with.

33:21Would you add something like this or is it less relevant for this kind of a workflow? No, it is very relevant, right? It's a way to simplify. When you're bidding a lot of agents, if you have the documents, you can skip this part or you can have a Google Sheet or an Excel and that skips this part, right? All the information are there. You don't have to write them every single time. Every time you just reference the same file, okay? For the sake of this workshop, there will be people that are familiar and comfortable with doing that because they have built a lot but there are often people who are being the first thing so i try to give it an approach which is generic right this is very simple it works very well it doesn't consume extra credit so for example i build a lot of workflow like this even if i could have a document right i need this for information point company name company url company blog and linkedin url and I use this along the way.

34:09Okay? Perfect. Company info. What I'm doing this is, what I'm doing next is basically exactly what you were doing, right? Based on all of the company info which I have created on step three, I'm going to ask an AI to create a company profile. If you already had the company profile built, you could skip this part, right? So you would save a bit of time every time you build a workflow. Okay? And to do this, I want to click on plus. I want to go to add an AI step, right? And I want to add a custom AI step, okay? This is basically me prompting whatever LLM I want. You can select anything that you can find within the platform.

34:53You can select ChatGPT 5.2, CloudSonet, Gemini 3 Flash, anything you want, right? In this case, I'm really liking Gemini this day, so I'm going to use Gemini 3 Pro for this specific task. So what I want him to do and what I will now prompt him to do is to take all of the information I have defined in the previous tab and to create a quick company profile, right? So I'm going to type here and then I'm going to paste the prompt in the chat so those of you who are following and maybe you can also post it on LinkedIn can copy paste it if you're building with me. I'm going to tell him based on and then I'm going to reference.

35:29When you want to reference information you have used in the flow, you can type at, okay? Then go to company information and then tell him there will always be always one row in this list, right? Because you have bidded once and there will only be one company, our own company. So it's very simple because it's the only option. You have company info, first row, company name. And then I'm basically going to reference here all of the information, okay? I'm going to reference a company name, company URL. So while you're adding the other ones, again, for those of you who are not watching, The way this works, and it's the same on all these platforms, whether you're using Relay or NA10 or Make or Zapier, it doesn't matter.

36:08They're all the same. Every previous step or basically every step in the workflow provides different parameters into the workflow. And then you can reference these parameters in other steps. So in this case, what Pietro is doing is he's saying based on, and then the parameter is company name, and then also company URL, and then also company blog. and these are parameters he created in the previous step when he created the table. So it knows how to reference information from the table, but it's the same thing if you get an information from somewhere through the automation, you can reference it in future steps and so on.

36:44So it's the same exact approach in all these tools, which basically means when you're prompting, you can create prompts that will be dynamically updated depending on that one time it runs because it knows how to pull information from previous steps. Exactly. That's absolutely correct. And after I have referenced this information, I give this information, I told them, okay, take the information from here. I'm going to prompt and say, create a description of my company, who it serves, what niche is it in, what our target customer is, what the main messaging and style are, and what type of product it offers.

37:19So basically everything that I will later need to draft good comments and to understand which post I want to comment on. I'm giving the AI the necessary information for it to work well. So whenever we create an AI step, and this is true here, but on any other platform, you have to make sure of two main things. Number one, you have to make sure that the AI has access to the right set of information. So if I reference something that the AI doesn't have access to, it will not work. Okay, in this specific platform, how you do it is you can see here that it says AI has access to, and then what he has access to, in this case, company info, which is exactly what we want.

38:01If you had to add something, you would click to add data from previous step and reference the right step, right? So one thing you have to always think is, okay, does this have access to the information it needs to give me a response? And the second thing you always have to do is to make sure that the output is in the correct format, right? So do I want one paragraph, one paragraph of rich text that tells me all this information, or do I want several different items in a response? I want the company inscription customer served. Cool tools like this one, they already auto-generate responses based on what they think will be the best output.

38:38In this case, what I'm looking at is something more simple than this auto-generated output. So I'm going to click on this three step and I'm going to change the output format to text. So instead of having a complex and formatted text, I want a simple plain text. I want one page about my company basically with all this information. Okay. So what we have done, we have given subreddit, relevant subreddit with the name of the subreddit we are looking for. We have given information about our company. the next step is going to be extracting the post from Reddit. To do this, we want to extract the post from each single subreddit which we have identified.

39:21In this sample, I included two, product marketing and product hunter. You can have one or you can have 10. Absolutely up to you, depending on how many you can find that are actually relevant for your company. The more you do, the more credits it will cost. So you have to be relevant as well. right? So what I want, what I need to do now is I need to add an iterator, okay? An iterator, for those of you who haven't been, it could sound as a complex word, it's actually something which is very, very simple. It loops over a list, okay? So if I have a series of subreddit I want to analyze, I will tell the AI for each of the subreddit which I have in this list, go to line one, extract the subreddit, perform this set of tasks for this specific subreddit, and then once you have finished this, go to the next item in the list, go to the second line, find the second subreddit, do the series of tasks for that second subreddit.

40:14This is called an iterator. Whenever you iterate, you have to iterate through a list of things. So I'm prompted here to insert a list and I will reference to the subreddits. I want to iterate through every subreddit which have defined here on step two. Once this is done, I will ask to... I will connect Reddit. If you haven't, I think you have to connect one of your Reddit profile. This only extract doesn't have permission to write on Reddit, so it can't damage your reputation. It can only query information and it's a super simple connection. Maybe it's not even needed. I don't recall now. But once you have here you can type on the bar above reddit and if you wanted to do this on link and you just type linkedin you'll be able to find all of the actions which you can um which you can do with reddit right depending on the platform you have a set of tasks which will be very easy to do right they are pre-compiled you don't have to do anything technical right and here what i really like is that here you can literally read and understand so this one third item is get comments on post getPostByURL, or this last item, which is the one we're going to use, getPostBySubreddit.

41:30This is exactly what we need now. We type Reddit, we look for what we want, getPostBySubreddit, and then we basically just set this up to reference the right set of information. So I'm going to tell him, getPostBySubreddit, what subreddit do I want? Make sure to reference the subreddit without the r slash. Here, and it's just technicality, and it will tell you if you do it wrong, so don't worry, even if you do it wrong, it will stop and tell you, okay, this is not the right format. You will reference that specific subreddit, and then you will ask it, and it will ask you what do you want to sort that subreddit for, right?

42:10You can sort, and you can do it as well on Reddit, it's the same if you do it manually, you can sort by new post, top post, hot post, best post, or depending on what you are doing and what activity you are doing in that specific timeframe, you can select a different one. For example, if you run this every day, you'll probably want to do the rising, right? Because if you comment closer to the posting date, you have better odds that your comments actually become as viral as the post, okay? But if you run it once a week, you might want to do, for example, the new or the hot test. In this case, let's just select the new, and then I can select the time period which I want to reference in this workflow.

42:54Since we run it once a week, we only want posts which are part of the last week. I don't want to post another one that I might have received already last week. But if you run it once a month, you will have monthly here. Okay? And I also want to tell him if you don't find any relevant posts, pause the run and send a notification. This is just bureaucracy, right? I just want to do a quick recap of the overall and then where we are right now. So the first thing was just set up. Like here is my company, here's the subreddits I want to follow. This is a very static step. The next thing is because we've defined several subreddits, we want to be able to do the process for each and every one of them.

43:38And that's why we're using this iterator. And what we're building right now is the steps inside the iterator. Basically, what is going to iterate through for each and every one of the subreddits. and what we did now is basically tell it what information to pay attention to when it's getting the subreddit posts what Pietro said earlier that you need to be mindful of is that every time this thing runs because it's going to take the information from the subreddit and it's going to send it to your favorite LLM you're going to be paying tokens to the API so don't go and build this with 15 different subreddits each and every one of them has 150 ,000 posts a week and then you'll be paying a lot of money to your LLM in order to process all of that.

44:23I'm sure there's other ways to build filters and things like that just to make it more efficient, but we are aware of that, that the more you ask it to do, the more credits you're going to pay. And to be fair, since this thing runs always, then you will need to go in your Excel file afterwards through 150 comments that you have to verify and copy and paste and edit, which you may not have the time to do. And so there's other reasons why to at least start small and learn how to do this and then perfect from there. Absolutely. And this is a question of design, right? It might sound complex now that Eiser mentioned, if you haven't ever built, but the reality is once you learn how to design and you learn how to test them, and we'll also see while testing this one, themes become much, much clearer, right?

45:10You learn how to make sure that you consume less and platforms like this also give you a great summary of what you're spending and you can set your own limit, right? So it's absolutely true, but it's something you learn in the first week while you start, you know, dedicating a bit of time to this. So now that we have extracted all of the subreddit, we closed the iterator. This was the only item which we're going to put in the iterator for every subreddit. We extract all of the posts of the last week, which are new. And then what I'm going to do next is I'm going to ask an AI step, an AI, to look through all of the posts which we have extracted for each single one of this subreddit and identify the one that, based on my company information, the information I gave about my own company, are relevant for me to give a comment on, right?

45:58So we look for a comment opportunity and we also ask him to explain, okay, what's the angle, you know, why do you think I should comment year. So you also have that piece of information. To do this, we'll create another custom AI step. And one more time, we want to select the model first. Again, I'm going to use Gemini 3 Pro because I really like it these days, but it will do a great job as well with Cloud. I think Gemini has the perfect balance within this type of workflow between precision and cost in terms of credits, but you can also test with different ones and see which one performs best, right?

46:35In the same platform, just by changing with one click. This is another very nice thing. So once we are inside here and we have selected the model, I'm gonna copy paste the prompt and then I'm gonna drive you through it, okay? I'm gonna have to, when I copy paste, it won't keep the reference. So remember to reference an information which we mentioned before, we have to type at and then look for the specific company information. Okay, for example, in this case, I'm referencing the company name and I'm referencing the AI profile which you build about my company, which is this AI output, which we created on step four, right?

47:12So I'm telling you, based on my company name and the AI output, which we created about my company, I want you to identify which Reddit posts are relevant for me to comment on. Sorry, my grammar is acting up. They might be interesting to leave a comment on my company and what we offer. Okay, so I'm explaining what you want him to do. And then I'm telling you my goal is to leave a comment and mention our service and how it can help you without directly promoting or disclosing us being part of the company. I want him to basically answer the question, how can company name help in this case? Okay, select the post.

47:50And I want to be sure, like I'm basically remarking that I want to select him, the AI, to select the post purely by topic feed. Specifically, those where my company's value proposition or solution is relevant and could have meaningful insights. Okay, so I'm doing a very simple thing here. I'm telling you how I want him to select the post which I want to comment. In a more complex flow with a bit more design and a bit more time, you would also be able to tell them, okay, I want you to draft comments for this post. I want you to specifically use this tone and go much deeper into how you want these comments to be selected and drafted.

48:31Like we said before, every time we build an AI step, we need to check two things. Thing number one is, does he have access to the right set of information? In this case, I need you to have access to the right post. It's fundamental without those, you won't be able to reference it. And I needed to have access also to the AI output. Okay, so I'm going to make sure that also AI output is added to the type of information that my AI has access to. Okay, and the second thing which I want him to do is I want to check that the response output is in the format which I want. In this case, the auto-generated one works very well because I might want to reference the post URL and post title in different sections, but you just remember to make sure that you're using the right set of information.

49:21Again, just on a very, very high level, every step with AI, by the way, it doesn't matter whether you do this automatically or manually, needs three things. What is going to be the input? What's going to be the output? And what needs to be the process in between? right that's how like every thing that you're building is how it's going to work and it's the same thing here you just tell it what the input is it's getting it from the previous steps tell it what to do and then tell it in what format you want the output and the the way you choose the format of the output is whatever you need there is no right or wrong like if sometimes uh simple text is great sometimes you want a table sometimes you want uh json code sometimes whatever it is that you need for either the following steps or for an output that you can process, you can pick the right output.

50:10Yeah, exactly. And I also pasted the prompt here in the chat in Zoom. Maybe someone can post it also in the link. Yeah, I will do that. If there's someone who's building with me. We're going a bit fast today because we wanted to cover a lot of ground. But again, you can re-watch this and go at your own pace, okay? The last thing which we want to do is we want to take this information, these results which we have gotten and we want to receive it via email. This runs once a week. once a week, I want to receive an email in a more complex and with more time, we could also put it on an Excel file, right, and receive an Excel file via email.

50:42In this case, to make it simple, let's just send it to ourselves via email. To do this, we're going to type send and we're going to find the voice, send email to yourself. This action basically takes a set of information which you define and send it to your own email with which you created the account on the platform, okay? Like for an email, you have to define a subject, which is going to be post to comment on this week, for example. And then I'm going to add a current date and time just to make sure that the email reference the week and time. And then what I'm going to do is I'm going to create a body of the email that says, here are the top posts identified for strategic commenting this week.

51:22and I'm going to reference the result of step seven, which identifies the specific posts which are interesting for me to comment on. So I'm going to do data, I'm going to go to step seven responses and I'm going to add a bullet list with all of the responses. So what this will do, it will iterate through the post and it will give me a response with the three top posts of the week because I asked for three top posts in the previous prompt. okay uh when i add it and i go to data here on the bottom right uh and then tell them bullet point of the responses i'm gonna reference the specific information which i want and in this case it's gonna be uh the reddit post title so that's the first thing i want then i want the link of the reddit post okay and again i click on data i go to responses reddit post url and then i go or why is relevant, which was the second type of information which we requested from the post.

52:23Relevant topic, suggested comment is this. And the last thing is, okay, sorry, I made a mistake here. Why is relevant? It would be a relevant topic. And then, sorry, getting stuck here. data reference to relevant topic and the first thing would be suggested comment insight so why is this post we should comment on insight and again to explain how this works you can write some of the stuff yourself and you can then add parameters into it and one parameter could be three pages of data because it's the output of whatever the AI gave you this could be very very long but this is how you create the email.

53:16So the email, basically what you type manually is going to be if you want like the headings. These are the ones I want to comment on and then we'll give you a list. And then why do I want to comment or what do I want to comment? And that will be the output from the AI. And it's really, really easy to set up. And then you're going to get an email like this every single week with the summary that the AI created. Perfect. Yeah. And let's test it. So we have maybe also five minutes for a Q &A if there's any question to test it, you go up at the beginning of the workflow and you click start test run and you click on start now.

53:47Okay, while this act, let me explain you in brief what we did, so it's a quick summary. We gave information about which subreddit you want to monitor. We gave information about our company so the AI know us and know what kind of comments are interesting for us. We extracted all of the relevant posts of the week from the subreddit which you selected. We fed it to an AI, we gave this input to an AI and we told them identify the ones that are relevant for me to comment on and explain to why they're relevant. And then I'm telling him, okay, now send me an email which explains why these posts are relevant.

54:24And if we see it running, it goes through every single step. You can go in and check what it did for every single step. It created a profile about my company. It extracted 25 posts for each of the Saturday which I selected. And now it's thinking about, which of these posts are the relevant ones for this specific company, relaying this example, are relevant for me to comment on. And then hopefully we send us an email with the result. In a second, we'll see the results. In the meantime, since we don't have so much time, either do you have a question? Is there any question in the audience? Maybe you'll link it in.

55:00I know this is very quick, but we covered a lot of ground. It's mostly to show you what the potential is and what you can achieve. And hopefully some of you have also managed to build this along. Together with the email of Samarit, I will also send this template so you can test it for yourself. You didn't build it together. I can export this template and you can import it directly. Yeah. So a few very high-level questions. One, the way this works in Relay is I understand you don't need to bring any of your API keys. like you connected, you know, Gemini 3 and so on. How much are you paying relay?

55:43Are you paying per automation? Are you paying per month? Like, how does this work? So you see it here. For example, whenever you run a test run, it's free, which is very nice. You don't pay credits, right? But at the end of every run, I can see how much run it would have used, right? In this case, it would have used 285 credits, right? I can see every single step how much it takes. For example, this Gemini 3 Pro, it's basically everything I paid for 241 credits right it was the most complex one right if I don't like that while I test I go back and I use geometry flash a cheaper model okay so you can really metric yourself on okay what do I really need uh while you test once you deploy honestly there's a there's a little bit of a markup uh if you run gigantic volume of operation you're probably better off having something self-hosted with your own LLM but for everything else for me for technical operators, it makes a lot of sense, right?

56:35It's super generous, the free plan they give. So you can basically test everything before putting into production, putting live, and then you can track your spending. Can you bring your own API keys into it or you got to use there? You can also bring it if you want, and that's significantly cheaper. So you can also do it if you're a more technical and advanced person. but just before we go, okay, I received this email. You see post a comment on this week. This is the result of the workflow, the date which I asked him to put. And then here are the posts. Title of the post. Here I can check the link.

57:14Why is it relevant? It talks about burnout, task automation, AI research agents, which is exactly what Relay does. And then the comment inside, the list of tasks mentioned, market research, competitive analysis, and brief are prime candidates for AI automation. Instead of doing them manually, you could set up AI agents to gather the raw data, draft the initial version, allowing it to act as the editor strategy. So this is great. Honestly, think about it. An intern wouldn't have done it so well. Next step, you build a comment, you put it on a Google Sheet, and you can edit in the Google Sheet, and then you go and post it manually.

57:47But I asked him for the post, and he returned the post with the exact logic for which he selected. So I can say, okay, this week I'm going to comment on this one because this is perfect, honestly. and then I want to talk about this one, SaaS competitive intelligence. So I will invest time into just posting meaningful comment, SEO and geo-optimize for those specific posts. So I want to do a quick, again, 30 ,000 foot recap because I think it's important after we were in the weeds and the details, which were fantastic, like people now know how simple this is to create. The first thing is you see this is a great combination of human AI working together, right?

58:26Pietro didn't create a system that automatically posts on any of these platforms. It finds information, suggests what to post, explains why, so you have the reasoning, and then it's up to you to decide what's going to be really meaningful to the audience, what's going to be really meaningful in the context of that particular subreddit, or as I mentioned, LinkedIn, Twitter, like wherever it is that you're posting this, you want to be able to contextualize it. So I think it is very, very important to have the human in the loop, because otherwise I can tell you that there's four or five people that comment on everything I post on LinkedIn and 100 % of their posts are AI generated.

59:05And it's horrible. Like, yes, it promotes my post, so I don't care because more people comment on it. But it doesn't provide any meaningful value. And hence, those particular people I don't really care about because they, you know, it's just why are you doing this? Like if you're not really providing any real value. And so I think that's a very important part of this. The other really important part of this is while this was quick and maybe not as easy to follow over sound without looking at the screen, all these platforms and relays definitely from a user perspective side is really, really easy, is a simple step-by-step.

59:43everything is clearly on the screen and it's not hard if you have zero technical skills to go in and start building these kind of automations because you don't need to connect any ipi keys you don't need to know what mcp servers are you don't need to know authentication o of two or whatever none of that like you log into your account somewhere like on this case on reddit but could have been linkedin and you just select things from there so it's very simple and straightforward to start building. And I highly recommend to start experimenting with these kind of tools. And the last thing I will say is it's a great example of combining linear process.

1:00:20So step-by-step that is linear with an AI that does the thinking aspect in the middle to be able to provide more valuable outputs than just old school linear workflow automation. With that being said, Pietro, this was fantastic both the introduction the knowledge about Reddit why is it important and how to think about building processes it's exactly what we said we're going to do and you delivered exceptionally if people want to follow you, if they want to take your course, if they want to connect with you what are the best ways to do that? Yeah, I'm going to post it here in the chat, my main channel is LinkedIn, I also have a large newsletter, but if you want to chat, if you have any questions, if you want to debrief on this.

1:01:05More is very available. Send me a connection reference to this specific session. So I know you have taken part into this and I'm going to 100 % respond. I'd love if this is a topic which is interesting and you enjoy the session, however fast it was. I run a course where we start from the very basics of building and designing automation to how to improve, evaluate and maintain agents using no-code platforms. And it's very much applied, right? So we build things like this with Calm, with one-to-one coaching. So it's very much applied that at the end you live with system which actually already work for your company, for Reddit, LinkedIn, and many of the things related to growth.

1:01:48So I also live here at this concorder, send it to Iser. Maybe you can send it in a follow-up email with all of the materials and the template too. And in any case, if you have questions regardless of the course, please connect with me and send me any question you have. I'm always very happy to respond. Yes, thank you. Thank you so much. Thank you for everybody joining live. I know William and Shahed have been really active on the chat on LinkedIn, but there's been like 10 or 12 other people on LinkedIn as well. So thank you all for being there. And obviously Andy and Dean and Derek, who's been with us and some other people who left, joined and left in the middle on Zoom as well.

1:02:25So thank you all for joining us. We appreciate you again, spending the time with us and asking questions that make this more interesting. and again Pietro thank you thank you thank you bye everyone I'll see you next it was a pleasure bye guys