Introduction

Hello, my name is Rauno Varblas.

I'm sorry, I cannot be there with you at the moment. At the same time, I'm not sorry either because of the sunshine.

But yeah, next time for sure I will try to be there in person. The event seems really great and the topics are interesting for me.

About Me and Technopol

So, I am the head of AI at Technopol and we work with companies from the very early stage until the corporates and it's a pleasure to be of kind of help also to this MindStone Event and TorGate with this event series.

Everyday Automations: Three Use Cases

And I, in my work, use also automations and I will show you three use cases that I use every day that help a lot with some of my duties.

this.

Why Relay.app Is My Automation Platform

So my weapon of choice is relay .app.

I don't know how many, if any of you have used it before. It's similar to Zapier or Make or yeah these are the main or N8n if I'm sure many of you have used to use these tools before.

Why I use relay .app? It's because it has a very intuitive interface and I since I hate the syntax errors this they have built it this way that the errors are kind of the user interface doesn't let you to make errors basically so you can point

and click and you don't have to worry if you forgot a bracket somewhere or dot is in a wrong place or there's a I don't know a white space somewhere.

Model Access and Billing Flexibility

It allows you to bring your own keys for AI models and of course all the latest AI models are there and they don't mark up on the token credits either, so you can use and you can get your invoice from OpenAI or Atropic or

Google or whichever model you use.

And also it has some free open source models like DeepSeek. I'm not sure which providers...

Oh, actually, if you bring your own keys, then I guess you can connect to replicate or some of the hogging face models as well.

Pricing and Team Philosophy

Anyway, it comes with quite a cheap fee, and it has enough integrations for me to do some of my most annoying tasks. And I would say also that they have a quite inspiring team.

Their CEO is He's at the same time their chief marketing officer and his team consists of 40 plus AI agents that are running on their own platform. So in that sense, they're, so to say, they're eating their own dog food. They make the product and they use their product to run their operations.

Agentic Helpers Overview

And so I have three different agentic helpers that I use every day. They're not doing the full work for me, but they are easing out some of these annoyances that we all have.

The titles are in Estonian, but the flow I will go through in the tool itself.

Use Case 1: Calendar Invites to Structured Calendar Events

So the first one is Calendar Invites.

So what it basically does is when I attend some event that consists of sessions like conferences or, I don't know, even evening events such as this one, the invite is always

in a in a different format so there is like either it's a web page or it's a image or it's an email and it's it's always in a in a format that is not uh well that requires manually uh working through

and i like all my events all the sessions of the events that i participate in my calendar so i can add notes there i can uh i don't know mark attach some attachments or or images or something like this so i like them the all the sessions of my events to be in my calendar individually not just

one big like full day conference event so and this does it for me so let me try to open the the app for you.

Workflow Overview and Trigger

So that's the relay .app in the workflow editor mode.

So as in any automations, what it does, it has a trigger. So something that is actually executing or starting the workflow.

In this case, it's a mail hook. So whenever I get the email, confirmation email of an event that I will participate, and it includes the agenda, then I just forward it to the mail hook.

And, of course, it checks that it's coming from my address so that nobody else can send stuff in my calendar.

Parsing Agendas from Emails, Attachments, and URLs

And then what it does is it will go through all the attachments of the email. It will basically convert all the attachments into LLM readable format. It also allows me to send, if the event agenda is in a web page, and it usually is, then I just put a URL inside of the email and it will scrape the agenda from the URL and then

Then it will extract basically all the session information from this data and put it in a kind of JSON format, structured format.

Creating Events Across Calendars

And then what it does, it will add the event into my calendar. And here I have both Google Calendar and Outlook, my work Outlook Calendar. So it will actually add the sessions into the calendar that I sent the original email from.

So if I send it from my private email, then it will add to my Google Calendar. and if I send from my work email, then it will add to my Outlook calendar. So it's kind of a workflow.

Use Case 2: Pre‑Meeting Briefing Dossiers

So the second workflow that I use this for is actually doing some briefing for the events or meetings that I have.

Like I have one -on -one meeting or some smaller event where there's up to 10, or let's say yeah 10 participants then what i what i like to have is uh uh like some time before the

actual meeting uh kind of a briefing into my inbox of uh previous uh conversations or or or basically the context that i have i have had with this uh with the guests of this meeting or the participants participants of this meeting.

Identifying External Participants and Enriching Data

So basically what it does is when there is a upcoming meeting coming next day and it has participants that are not from my organization, so I'm extracting the email addresses from the meeting invite, then I will find the emails that are not from Technopulse or not my colleagues, and then I will extract the domain and I will look up

the data about the company that they they are from then i will look up their linkedin in url and the data from the linkedin page and then i will have the llm to make a kind of a summary about the profile and the company that they they find from linkedin it's all basically basically public information.

Delivering a Concise Briefing to Inbox

And then this gets sent to my work address, all this dossier. And in the end, it will look something like this.

Let's see, there's an event, our startup, or our AI incubator, next batch kickoff event. and it will send me, let's see if I've got it here, such kind of a summary about the event. So it will be, okay, this one is broken. Sometimes, yeah, sometimes it doesn't work.

What the Briefing Includes

So there's an agenda, there's the external participants participants that are participating, who they are, what are the LinkedIn URLs, and also the participants of the startups that are participating, some summary of previous interactions with them, and some preparations that I would have to do.

So in this case I'm just participating as a guest as well so I don't have much preparation. But quite often times it actually helps me not forget some important detail that actually wanted preparation and then I have time to prepare for it.

Use Case 3: Travel Expert Agent

And the third one is is a travel expert so basically this one is a true agent in that sense in the sense of ai agent so it decides itself which kind of tools it uses so basically what it does is i give it a

kind of open text input or chat input i want to travel somewhere some certain dates or some some destination and then the agent is using actually uh uh airbnb and uh kiwi mcp servers to look up uh accommodation for me according to my uh preferences and also the

flights most appropriate flights so and the way i started is uh is uh this at the moment i started it manually but it can be also triggered by by email or or chat or or i don't know slack message

Starting a Trip Search

or whatever so let me give it a try so i want to travel to run canaria for a new year's eve

Let's start it running and then I'll show you the steps. So it starts going.

Orchestration with GPT and Tool Selection

First it will go to a GPT -5 as a main request. So this will be the kind of main orchestrator agent.

So he's my travel agent, he will look for Airbnb listings and flight tickets according to my preferences. is if it's a family trip it will look for a four room accommodation and tickets for four and if it's a business trip then it will just look for me and i have set some some

kind of context for that and let's see so okay so it has encountered some error let's see that's right so sometimes it's uh yeah a little bit picky so previously here i have run a a similar query so

Example Output and Itinerary Options

a family trip in vietnam for three weeks in february and march and so it has looked up some airbnb listings i've done three airbnb searches so if it's not happy with the first option and then it does give me search flights and in the end it will give it has given me a such report it's i'm sorry it's an estonian but basically

it gives me two options in first week in danang then first 10 days in danang and then 11 days in Phu Quoc. This looks like pretty good options for family.

I've been in Da Nang, it's a very nice city. Phu Quoc is a nice tropical island.

So it has all the information that I have set up for the preferences of my family and then the total sum and links to the direct listings as well as uh as links to the to the actual booking of the of the flights

Conclusion

there we go and uh yeah that's uh that's my three use cases