Introduction

All right, so welcome, everyone. Nacho Pujol here, co-founder at FlowHawk, an agency and consultancy that helps teams optimize their internal operations.

My co-founder over there is a technical guy. I'm more of in the commercial side. But I can also do some stuff.

Overview: Simple AI workflows for daily productivity

Today, I want to show you how to use AI for your day-to-day. with some simple workflows that you're going to be able to build on your own by the end of this little 15-minute talk. So we're going to be going over just three little workflows, two when we have time, a third one.

The first one is one that I like a lot.

What did I do with my phone? I don't know. I might need my phone for that.

Well, in the meantime, I'm just going to get started. Oh, there you go. Thank you.

Workflow 1: Voice-to-Post capture on the go

So for this first example, we're going to be creating an idea to post that I like, that I like to mention, which is what is this workflow going to be about? On the go, for any of you who are creating content either in LinkedIn or for blogs or even for your upcoming meeting tomorrow, we're going to be creating a little workflow that turns your voice into that content, and it's going to be saving it on a database, whether you use Notion, Google Sheets, whatever.

In my case, some of you might know, most of you might not, I post on LinkedIn on a weekly basis, like two or three times a week, and I have to come up with a bunch of ideas. Some of them happen whilst I'm on the train, sleeping, or whatever.

So we're going to be creating a little workflow just right now, that will turn a voice message from WhatsApp into an idea in my content calendar or content database.

So on the go, oh, yes.

Trigger: Emailing yourself a voice note

So how we would start it is this little message, sorry, this little idea, we're gonna be sending it to email to the workflow. So we'll start with a little Gmail trigger so that every time that we receive an email, The first email, we've got a little workflow here live so you can then do it on your own.

So we're gonna be receiving an email always from the same person, which is gonna be, there you go, you guys have my email for whatever marketing you wanna send. I'm exposing it right here. That's going to be it.

And also another filter, in order to make it easier for the workflow to identify when I want the workflow to run, we're going to make sure that it only checks for emails sent by these address and that have an attachment, which ideally will only be for this situation.

I'm going to zoom in real quick. So I'm actually going to keep it this way. And then I'll zoom in to see the whole thing.

So now we have an email trigger.

Step 1: Transcribe the audio

What we want to do is to turn the voice message into text so that the AI can then turn it into a post or whatever we decide to do. So we're going to do a transcribe audio from. We're going to be choosing the file. Select from attachments.

A little context here, which I went very straight into the demo.

Tooling: Relay app basics and why use it

We're using a Relay app.

Relay app, if you may not heard, is like an or Make. But for me, it's a little more intuitive to use.

It's pretty easy to get started, and it's pretty simple to build day-to-day workflows. It's very intuitive.

You don't need 30 or one hour videos that I had to go through when I started playing with it. And this was very straightforward.

They also have a very good free plan with a bunch of credits per month. I'm not linked to it, but I really like it.

Step 2: Turn transcript into a post with a solid prompt

So now that we already have an AI transcribing our voice message, we're going to turn this into a post. To do this, we will have an AI to write within a format.

1As Manuel was mentioning earlier, the prompt is the most important part of the workflow. A bad prompt is gonna make this workflow crappy. A good prompt is gonna make it a lot more actually useful.

So we have a list of resources that I will be able to share with you by the end of the demo, where I have literally the step-by-step of how to build this prompt, how to use this workflow, and also the prompts for everything.

So we are gonna be using for this one, hold up, we're gonna be using this little prompt over here with, you will have time to check the structure for the sake of the demo, we're just gonna go straight forward and you then will be able to to go over the prompt and get the sense of the structure.

So now we have a prompt, but it's not linked to the voice audio, so we need to do that. In order to connect the previous step, so the audio that has been transcribed, we are gonna be using this. So it's now connected to the transcript that was generated earlier, and just for the sake of the demo, and to make sure that it doesn't forget about anything, we will do a little

header here so turn this and now a little tip if you do command add you can actually refer to the Thanks to the transcript. You can actually find whatever step goes before this.

So now we already have a very solid prompt. Here you would put an example of a LinkedIn post. I don't have it here because it comes from the demo that I prepared for you, from the resources that I prepared for you guys. So it might not have my exact structure.

But yeah, we're just going to go for it.

Choosing the right model and managing token cost

So now, actually, another interesting thing that I don't want to forget about, when you run a workflow here, it's very important to choose, because this is going to be sending the information to a general LLM that you're familiar with, and it's important for us to decide, depending on the task, What type of model do we want to use? Because as you can see here, you can choose any model from any of the big LLM providers. And they recommend the ones.

And you're going to be spending tokens, the ones that are given to you every month, you're going to be spending tokens every time this workflow is run.

So I would recommend, depending on how important the quality of the content needs to be, I would go for a cheaper, more expensive option. So I always go mid-level. I don't want to go too cheap nor too expensive, token-wise.

So we will go for the ChatGPT-5, which is pretty solid. So now we have all of this set up. Now what happens? We have already the idea turned into text and the text turned into post.

Step 3: Store the post in your content database (Notion/Airtable/Sheets)

We want this to be stored somewhere. So whatever you use, it could be Google Sheets. It could be Airtable. In my case, I use Notion, as you might have seen.

So we're going to be adding this, the LinkedIn post to Notion. So by tabbing here the app that you want to connect it in, Again, it's super simple, super self-intuitive.

I want to add a page. So it's going to be add page to database. I already have my database with all my content ideas that I actually want to show you before we go in.

Here's my entire LinkedIn-like content board. So we're going to be adding a new one here. So we will do here a content tracker. The content, which will be the content of the LinkedIn post in this example, we're going to be adding it here.

So again, I'll go for the ad, the ride with AI, which is the third node, let's call it. So we will have the output here. We will add a filter because that might be useful for another workflow that we're gonna show later.

So as a filter, I want the status to be, it's gonna be idea. There you go. So we have an idea of a content.

Run the workflow: From WhatsApp voice to LinkedIn draft

Let's try this. This now is actually a workflow completed. We're gonna be turning it on. So if the demo works out, hopefully we'll have a LinkedIn post ready to post for whatever day comes next.

So how I usually go, like go in the flow, I'm in the train or whatever, I just came up with an idea of a great post that I want to post on LinkedIn. So for example, I don't know, you over here, do you want to tell me a fruit? A fruit, whatever fruit.

Mango. Mango, I love mangoes. All right, so I just got a great inspiration eating mango. I wanna write a LinkedIn post on how great mangoes are and how they can actually trigger the inspiration in your brain.

So make a super cringe post on LinkedIn of how great are mangoes to closing more sales. All right, so now I just send it to my own personal WhatsApp. You guys have your own chat.

I don't know about you guys. I have my own chat where I send all my own stuff. It's called actually Brain Box because it's whatever goes through my brain, I put in this chat.

I'm going to be, since I cannot trigger WhatsApp directly to the workflow, what I'm going to be doing is forward it via email. So with one, just two clicks. I'm going to put here the email where my Relay app is connected, which is my business one. I won't be sharing all my emails, so this one I might keep it for me for now.

Now the email has been sent and received. There you go. We have a run going on there, so we're actually going to be showing it. the workflow was automatically triggered, and now the audio is being transcribed.

This might take a little bit, so we're gonna move on to the next workflow for now, and then we will go to the database and see if the post is ready.

Workflow 2: Weekly AI-curated newsletter from your unread emails

So, for the second workflow of today, we, okay, I don't know if it's me or maybe someone else in here, how many of you subscribe to newsletters about whatever topics you're interested and then you never read them, but you still see them in your inbox? Okay, a bunch of you guys. Okay, thank you, I'm not the only one.

Moreover, it's stuff that I wish I had the time to read, but every newsletter is just so long, and most of them, they come up with pretty similar topics, so those take a lot of time.

I used to actually read those newsletters while I was having coffee every morning before breakfast, but it was taking a lot of time, so I came up with this workflow, which is, pretty much grabbing different newsletters that I like, putting them together in one newsletter, and by the end of the week, having a summary of everything that happened in the space that I'm interested in. In my case, it's AI, of course.

So we create a very short, very down-to-earth, practical newsletter. For that, which takes, it's a bit of a more complex workflow, so I'm not gonna have the time to do it live, but again, it's gonna be my resources with every single step that you have to do for the notes, but we're gonna go over it step by step so that you can actually understand how you could actually do this on your own.

Trigger: Weekly schedule

So, in my case, we're going to start with a weekly trigger. As we mentioned, it's a little trigger like Manu said in about an event, it's the same. So, I decided that every Friday at lunchtime, I want to get whatever happened that week in my inbox.

I usually like two newsletters a lot. One is Neuron AI News and the other one is the Rundown AI.

Inputs: Fetch recent items via RSS

So in order to connect, to gather every single newsletter from every day of the week, I'm gonna be calling their RSS. The RSS is something that you can find in your newsletter of choice. Not all of them have them, but many of them do. And if you go at the bottom of the page, you can actually find this little icon here.

which is all the content, but well-adjusted for LLMs or for AI to understand the content a lot easier. So we copy this, you would copy that, and then once you add a little node, I'm just gonna show you, you would add a node that would be, you type RSS.

So this fetch recent RSS item, that's what we have here. So this is the node that we chose.

We did one RSS for Neuron AI. We only want to grab the five last blocks, so these five last little cubes, because that usually is associated for the five last days, so the five last newsletter, considering that they do one newsletter per day.

We're going to be doing the same for the random AI.

Qualify, scrape, loop, and summarize the best articles

1And then we want Gemini or an AI to actually qualify these news because some of these daily newsletters might be relevant or promotional or whatever. So we have a node that qualifies these news with a little prompt that is here that, again, it's in the resources that I will share later. And once we have qualified them,

I want to, like at high level it qualifies them, and now I have a web scraper that once we have decided the articles that are, the newsletters that are relevant for us, we're gonna scrape those and get the content in order to then summarize and create the final newsletter. So here it's just a scraper in a format that is actually a very interesting, way of working with workflows, which is called a loop, which means that it's gonna be doing the same action for different outputs given before.

So per se, before we extracted 10 different blocks, 10 different newsletters from different days, and we want to scrape those one by one and get the summary to then put it in the one newsletter. So this means that this web scraping is going to happen every time for every block that we have from earlier.

And then with all those summarized newsletters, we're going to put them together with, again, AI.

Again, for the sake of the demo and because we don't need a super highly well-written newsletter, we're going to be using ChatGPT-5. to turn all the scraping from the different newsletters into one.

So again, long prompt, very important. You have the structure and the resources, so this is ready to go.

Assemble and deliver the newsletter

Once the newsletter is written, we're going to be receiving it in our email.

So now that we have seen this, let's not forget, ah, let's actually close this one, and once this one is loading, we will go see the previous one.

Pro tip: Use test runs to save tokens

So we're gonna start, actually, a little trick here. If you ever start using RelayUp, If the workflow is active, so it runs on its own, it's gonna be consuming tokens. However, if you're running test runs, they are not consuming tokens. So what does this mean?

I could either or automate it and have it every Friday without thinking in my inbox the newsletter, or I could come manually into Relay and hit the start test run, so I'm not consuming tokens every day, every Friday. It's a little cheat code here. So we're gonna be running it now.

Okay, yeah, so it found the newsletters. Now it's doing its thing with reading different blocks. So we're gonna go back. Actually, it should be somewhere here, right?

Yeah, it's gonna be this one. Let's see. One mango, three close deals. It's a good starter. I'd stop and read it.

So I was mid-slump at 3 PM. I bit in a ripe mango, brained it up, whatever. It's got a solid format, maybe a little short, not too long.

Honestly, here's what I noticed. Sweet bite gives quick focus. Ritual makes me show up. It's actually pretty solid. It's not bad.

I might post this tomorrow. No, I don't think so. We're going to give it a little title. I forgot in the demo to give it a title. We're going to give it manga post.

This one is done. This one is still going. Almost there.

We're going to be receiving it on our... How are we doing on time? Tied? Okay.

Workflow 3: Auto-generate images for LinkedIn posts

There was a third demo, a third little... I'm actually going to show it live, but we're not going to go over it. A third little workflow that I wanted to show, which is we have the LinkedIn text, but we don't have the image.

And right now, most of the posts you see on LinkedIn They require an image for you to stop and actually get to read it if it's only text.

Personally me, I never stop and read if there's only text. I need an image to hold my attention.

So I created also a little workflow that is going to be creating an image based on the text of the workflow that we provided earlier.

I'm going to make sure that this is running. I have it here. I'm just going to turn this on. Publish.

Trigger from Notion: Use a "generate" tag

So basically what it does is it goes to my Notion page. It finds the tag generate.

Process: Summarize to headline and render with NaNoBanana

So here, if I switch to generate to the mango post, if I switch to generate, the workflow will trigger, and what we will do is summarize the post, turn it into a headline, and that headline, it will be encrusted, encrusted? That's not even a word. That will be inside a picture that it will be generated with NaNoBanana.

You might have heard of NaNoBanana. If not, type it, go try it, it's super fun.

And it will be updated in our, automatically updates it in here.

So it's as simple as if right now it's running, if I hit generate and we wait for a little more, we will see on a little column on the right how the image appears ready to then be posted to LinkedIn.

This being said, I should have received by now a newsletter. But I don't think so, because this is only the audio. So let me go check real quick what happened there.

Runs. It's still on it. Oh, my bad. I had to actually. There was a human in the loop involved, and the human wasn't there.

Resources and wrap-up

Okay, so yeah, this is pretty much done. Again, just to close it up, you have everything from this conversation somewhere, yeah. You have everything that I just showed, pretty well explained, I think.

In this resource that is open to everyone, actually, in order for you to access it, feel free to scan the QR code. This is how you're going to have access to the instructions you'll have to do every workflow and with also the prompts needed to execute all these workflows. Yeah, and if you want to try this tool, which is pretty simple, again, I also have a... There's the relay code up top. You will find it here.

So yeah, this is, let me actually show if we managed to have the image. Yes, it's here. Mango, oh no, not yet. Not yet, not yet. But no, just wanted to show.

This is still running. Oh boy, you're slow. Where is this going? It's thought of 10, so it might take a little bit. The workflow is just running. It's taking its time.

But I can actually show a result of a previous one, weekly newsletter. So this format is pretty tough to read. But in the phone, it looks a little nicer.

But again, it's a three-minute summary of everything that happened. And also, the two newsletters that I usually follow, they also have a tool section that I like, because I get to stay up to date with the new tools and upgrades. So we also have it there.

Time saved and final thoughts

So yeah, that's pretty much how to save an hour of reading in total per week to go down to three minutes. And yeah, pretty much it.

Templates and getting started

You will also have, just to close it in here, really has a really good template kind of dashboard with a lot of workflows that you can actually implement and start using day to day.

So yeah, this is pretty much it. I'm just gonna leave the QR code here in case you missed it.

Conclusion

Thank you very much.