Speaker unverified: evaluation

Transcript context

…I'm mindful of talking about a thing without showing it. And we already have the setup to show it. Why don't we jump into a screen share? For listeners, you can jump on the YouTube and like and subscribe. But also, let's have a look at how you show off Lindy. Yeah, absolutely. I'll give an example of a very simple Lindy and then I'll graduate to a much more complicated one. A super simple Lindy that I have is, I unfortunately bought some investment properties in the south of France. It was a really, really bad idea. And I put them on a Holydew, which is like the French Airbnb, if you will. And so I received these emails from time to time telling me like, oh, hey, you made 200 bucks. Someone booked your place. When I receive these emails, I want to log this reservation in a spreadsheet. Doing this without an AI agent or without AI in general is a pain in the butt because you must write an HTML parser for this email. And so it's just hard. You may not be able to do it and it's going to break the moment the email changes. By contrast, the way it works with Lindy, it's really simple. It's two steps. It's like, okay, I receive an email. If it is a reservation confirmation, I have this filter here. Then I append a row to this spreadsheet. And so this is where you can see the AI part where the way this action is configured here, you see these purple fields on the right. Each of these fields is a prompt. And so I can say, okay, you extract from the email the day the reservation begins on. You extract the amount of the reservation. You extract the number of travelers of the reservation. And now you can see when I look at the task history of this Lindy, it's really simple. It's like, okay, you do this and boom, appending this row to this spreadsheet. And this is the information extracted. So effectively, this node here, this append row node is a mini agent. It can see everything that just happened. It has context over the task and it's appending the row. And then it's going to send a reply to the thread. That's a very simple example of an agent. A quick follow-up question on this one while we're still on this page. Is that one call? Is that a structured output call? Yeah. Okay, nice. Yeah.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.