The Exchanges

Q I love it. Um, would you be able to go back to the template section? And the reason why is I want to scan through some of these just to get people's minds primed for the types of stuff that you can use.

A Yeah. Let me open the full list of templates here. Uh, you know, lead enrichment, lead qualification, customer support. Like, and customer support over everything, because we just released last week, like, 1600 integrations. Like, again, well, by far right now, the top AI agent platform in terms of number of integrations. And so, wherever your customer support exists, if it's like Telegram, or Slack, or WhatsApp, or Zendesk, or Intercom, like, you name it, like, we can, we can automate your customer support over all these platforms. Um, the focus group here is, is a really interesting In this case, because LLMs are not super good at reasoning, but they are really good at pretending to be human. Because that's, that's what they all, right? They've been trained on, on so much text that's human written. And there's actually a lot of people that find that the end sales that LLMs give you are actually very correlated to what, to the end sales that a human would have given you. So if you ask, like, do you like this or that? Like LLMs are actually doing a very reasonable job at emulating a human. And so that means that you can, like marketing companies and like big firms, they spend fortunes on focus groups. And we use it ourselves as well. Like, we've created a Lindy that simulates our user. So we've prompted it to be like, hey, these are the different personas of users we have. And …

AI assessment note: “Let me open the full list of templates here. Uh, you know, lead enrichment”

Q And that's, and that's why, you know, I saw on Twitter someone, someone say, isn't this just Zapier? And so the difference between Zapier, which is automations, and this is there's intelligence, correct?

A Yeah, there's a huge difference between an automation on the one hand, which is what Zapier is really good at, and an agent on the other hand. Right. And so, yes, I think the two biggest differences are going to be that, uh, one, um, is that you can use AI at any point in, uh, Lindy. So I'll give a super concrete example. Like I have, uh, some investment properties in France, which is terrible financial decision. Um, and I, I, I received these, uh, confirmation emails. I've put them on like the French Airbnb. And so I received this confirmation email. It's like, Hey, you just made 200 bucks because you, you rented your place on, on Hollywood. And I have a Lindy right here. What she does is, okay, when I receive an email, which is a reservation confirmation, you go to these Google Sheets, and you append a row to the Google Sheet. And you can see here, the cells, each cell is a prompt. So I was like, okay, please extract from the email the day of the reservation. Extract the total amount of the reservation. And you'll notice I don't have to, at any point, inject data from a previous step, like In Zapier, each step is an island. It's isolated. Here, it's an agent. It's an agent that is aware of the entire context of everything it did, and so if you want to do this with Zapier, you're going to have to write, like, an HTML parser. You may not be able to do it. It's going to be hard.…

AI assessment note: “there's a huge difference between an automation on the one hand... and an agent”

Q And while you bring that up, what is an agent swarm?

A That's a good question. An agent swarm is the ability to send a list of You have things to do to an AI agent, and for this AI agent to do all these things reliably, quickly, and in parallel. Because AI agents are awesome, and they're super powerful, but over long-term, like long-time horizons, they fail. They lose coherence. And so, for example, if you give an AI agent a list of things like, hey, take these 500 people and send personalized outreach to them. Like personalized lead outreach is a very big use case for AI agents. First of all, it's going to take forever. It's going to be pretty expensive. And by the 200th lead, it may fail. It may just become unreliable. If you use an agent swarm instead, it's going to, basically, you know, like in the matrix, you have like an agent Smith that duplicates himself. It's like the agent Smith thing. It's like the agent is going to duplicate himself and send one copy of itself to each lead. Yeah, so this meeting prep Lindy here uses, it's actually, it's funny, it's a swarm of swarms. So what I do is every morning you wake up, you check my meetings for the day, and then you deploy an agent swarm for each meeting that I have. And then inside this swarm, you deploy another swarm for every attendee of every meeting that I have. And then this is really good because this Lindy actually uses the meeting notes that the other Lindy brings togeth…

AI assessment note: “An agent swarm is the ability to send a list of”

Q I love it. All right. Is there anything else you wanted to cover?

A Um, let me think. I'll, okay, I'll show one last example that the conference is showing is my, my CRM manager. So I have this Lindy that helps me, um, manage my, my, my network. And so I, every so often I talk to her and I'm like, Hey, I just met this guy. He's awesome. I really wanted to stay, stay in touch with, with him. So, um, I'll show you an example here. Um, well, we're going to have to, we're going to have to blur it out, but, uh, I added this guy, uh, was like, Hey, like this guy is, is awesome. You know, apparently he's the best that other guy has ever worked with. I did him to a spreadsheet. Another really cool thing is that I can talk to this Lindy and I can be like, Hey, who should I hit up when I'm in New York? Or who are, like, really good salespeople that I know? And then it finds me a list. And then this Lindy, I actually also made her observe my inbox and notice when I have booked a flight. Which, by the way, soon is going to be a Lindy booking flights for me. And so she sees when I booked a flight because I receive an email confirmation. And she's like, and I can, I can actually show you because right now I'm going to, to New York. So she's like, Hey, like you're going to New York. This is just like a test email. I sent it to myself. And she's looking at my CRM and she's like, Hey, these are the contacts that you should meet up when you're going to New York.

AI assessment note: “I'll show one last example that the conference is showing is my, my CRM manager.”

Q Do you want to show one, while, while we're still waiting, do you want to show one last use case that people could, could use Lindy for?

A I, so we're currently looking for this, uh, designer. And so, so this one looks very complex, but again, like once you get the hang of it, it's basically just like add nodes and, and branching and conditions, right? And so what I did here is like, okay, we're looking for a designer. I'm going to send you random ass information about designers. It could be, I spend an embarrassing amount of time on Twitter. So like most of the time it's going to be a Twitter link, right? It's like, ah, I really like this guy's work. I just saw him on Twitter, right? Sometimes it's going to be a portfolio. Sometimes it's going to be just a name. Sometimes it can be anything, right? And so you can see at the very root here, it's like, what did I give you? Did I give you a website, an email address, a LinkedIn link? And here its job is basically, hey, and you can see this is why it goes on this like site quest here. There's like multiple branches, like find the email address. So I have a Google Sheets where like I already log a bunch of email addresses. Um, but if it's not there, then there's going to be an AI agent here. And again, this is the beauty of AI agent. I'm just giving it a high level goal. Like browse for a bit. If it's a Twitter link, maybe it's there. Look for the portfolio and give up after four attempts. Like if you've, cause sometimes AI agents get stuck in the loop. They go on lik…

AI assessment note: “we're currently looking for this, uh, designer. And so, so this one looks very”

Q head right now. Cause what I'm doing is I'm thinking, what are all the tasks in my business that are recurring like this? Like what, what you're showing me right now, my EA does this. Um, My EA does it. So like, what do you recommend to people? Should they, should they take stock of all the recurring tasks and then See if the, you know, Lindy's can do it?

A I think so. So what I do recommend to people is actually to start with these, like, personal assistance use cases, because it's just so easy, and everybody can use a meeting scheduler, everybody can use a meeting prep, everybody can use a meeting recorder, and so just, like, you know, dip your toes into the water like that. Like, just, like, get the reps, put a W on the board, and you can onboard this in, like, five minutes. It's, like, super easy. We've got templates here. We've got, like, hundreds of templates. If you go to the, The home tab here, there's a bunch of Lindy's that are pre-created, and these ones, we literally advertise them at the top because, like, we call it, like, the personal assistant startup pack. It's just, like, everybody can use these things. And then, little by little, once you understand how the platform works, which it's really pretty easy, then you're going to very rapidly, like, to your, exactly what you said, it's like a light bulb is going to go up, and you're going to be like, oh my god, like, this can be used for everything. So I'll give a more complex example, and what do you think, like, should we just, like, build a Lindy live right now?

AI assessment note: “I think so. So what I do recommend to people is actually to start with these”

Q way, I'm loving this, but where my mind is at is great flow. You're really good at designing these really smart kind of workflows, these AI, you know, agents. How do I get started? Like, how do I know where to start? How do I design it? Like that to me is a really, really important step that I think a lot of people are going to miss, including myself.

A No, 100%. There is, there is a little bit of a learning curve, but you know, you mentioned Zapier, you mentioned, you mentioned Make, like, I think that the, the, the folks here who listen to us, I mean, most of them are pretty tinkery. They can figure it out. We have like a Lindy Academy. We have like a bunch of templates on this tool, but yeah, I mean, let me, let me just show you, I'll create a Lindy right now, just to, to give you an example. Um, I'm going to create a new Lindy and I'm going to make it so that she, Um, what's an example of this thing I want to do? I want to, when I receive an email that is very time sensitive, uh, I want that Lindy to ping me on Slack, for example. And so I'm going to go, um, Gmail, so it's just a trigger. When I receive an email, I'm going to create a condition. It's, again, that's what I mean by like, it's AI throughout. So right here, I'm like, uh, the condition is just natural language. So go down this path if this email is urgent. And time sensitive. Okay. It's really that simple. It's like a small prompt. And here, what I'm going to do is, okay, at that point, you Slack, you send me a direct message. Okay. The user, I'm going to enter my own email here, but even if I didn't do that, it would probably still infill that I wanted to message me. And here, I can enter a prompt, and I'm going to be like, um, send me a message about This tim…

AI assessment note: “let me just show you, I'll create a Lindy right now”

Q gets me thinking that not only could you use Lindy to, like, make a scalable business on way less employees, but couldn't you build a business around Lindy? Meaning, couldn't you create the design recruiter and just, like, cold call, cold email using Lindy? Um, you know, say, hey, like, are you looking for designers? I'll give you every single day, like, email, three emails of, of high quality designers.

A 1000%. So you can see here, and again, especially once you get your Lindy's to work together, you basically end up like an entire team of Lindy's, and yes, you could, you could build, and that's our vision, like you could build an autonomous company with Lindy. So you can see here, for example, I have this like lead generator Lindy, right? So I'm going to ask it, I want five product designers who work at Zapier, for example. Uh, I don't want to, I don't want to hurt my friends at Zapier, but, um, So it's going to kick off the search, and I think this one, I've asked this Lindy, hey, once you've found people, I want you to add them to Google Sheets. It's just more convenient for me to review them there. Even if it doesn't add them to Google Sheets, it's just going to give me like a markdown table. All right, so it's giving me a Google Sheets. Now it's going to talk back to me. It's going to be like, I found these people, and I guess we're going to see who she found. Um, oh, I know. I know some of these people. Uh, so I'm going to click on their LinkedIn links right here, and I think there's a problem with the scrolling, but I believe, I believe there's more information here. These are personal, no, they're like Zapier emails, which kind of sucks. Yeah, these people, director of product design, design at Zapier, product design at Zapier, director of product design, product design…

AI assessment note: “1000%. ... yes, you could, you could build, and that's our vision”

Q Yeah, that's amazing. I love that. That's hilarious. Cool. The, uh, is the proposal, uh, finished?

A Should we check in on that? Yes, it is finished. Okay, so now she wrote this Google Doc. Let's see what the Google Doc looks like. Damn. Ok, so executive summary, vehicle specifications, base cost breakdown, extended range battery, a thousand dollars. I mean, look, it's an LLM, so LLM's are going to LLM, like they're going to fill in the blank, like given like how few, how little guidance I've given it. Ok, so then it got back to the time-sensitive email notifier, and it received the message back You see, so it received this message back. It sent it in an email. Hi, Florent. I've prepared another proposal for the Sabre truck. And, and then it, it kept me posted on Slack. So now, oh yeah, because I asked for several proposals, like she, she created several proposals. So both of them ended up going through. Um, and so now if I go back to my email, Yeah. So because I asked her to create two proposals, like she just received two proposals. Um, it did take a good amount of time, didn't it? Uh, I sent this at two PM. Yeah, it took like six minutes. Yeah.

AI assessment note: “Yes, it is finished.”

Q Yeah, that's amazing. I love that. That's hilarious. Cool. The, uh, is the proposal, uh, finished?

A Should we check in on that? Yes, it is finished. Okay, so now she wrote this Google Doc. Let's see what the Google Doc looks like. Damn. Ok, so executive summary, vehicle specifications, base cost breakdown, extended range battery, a thousand dollars. I mean, look, it's an LLM, so LLM's are going to LLM, like they're going to fill in the blank, like given like how few, how little guidance I've given it. Ok, so then it got back to the time-sensitive email notifier, and it received the message back You see, so it received this message back. It sent it in an email. Hi, Florent. I've prepared another proposal for the Sabre truck. And, and then it, it kept me posted on Slack. So now, oh yeah, because I asked for several proposals, like she, she created several proposals. So both of them ended up going through. Um, and so now if I go back to my email, Yeah. So because I asked her to create two proposals, like she just received two proposals. Um, it did take a good amount of time, didn't it? Uh, I sent this at two PM. Yeah, it took like six minutes. Yeah.

AI assessment note: “Yes, it is finished. Okay, so now she wrote this Google Doc.”

Q And from a prompt perspective, like, do you have any advice on how to create optimized prompts for Lindy?

A None. I just, like, the agent builder takes care of all of that for you. I mean, you've seen me create this agent. It's like, it's not rocket science. It's like, I just, I just talk to it like I talk to an intern and just figures it out. And again, that's the beauty of it is like, oh, okay, it's found your profile. And now it's going to. Uh, ah, fuck, you don't have DMs. You're like in the photo mode. Let's see. Yeah. Let's see if you can figure this out. But this is a good example of what I mean by, like, that's the advantage is, like, you have this double mode of, like, this double loop of, like, you can edit the agent and the agent's instructions, and you can see the agent operate. And it's sort of ping pong between these two modes. Because if the agent screws up when it operates, you can just go in these instructions, and you're like, okay, like, keep this in mind moving forward. Okay, so now it's, um, it's looking at your, like, profile to, like, know how to pitch you.

AI assessment note: “None. I just, like, the agent builder takes care of all of that”

Q And from a prompt perspective, like, do you have any advice on how to create optimized prompts for Lindy?

A None. I just, like, the agent builder takes care of all of that for you. I mean, you've seen me create this agent. It's like, it's not rocket science. It's like, I just, I just talk to it like I talk to an intern and just figures it out. And again, that's the beauty of it is like, oh, okay, it's found your profile. And now it's going to. Uh, ah, fuck, you don't have DMs. You're like in the photo mode. Let's see. Yeah. Let's see if you can figure this out. But this is a good example of what I mean by, like, that's the advantage is, like, you have this double mode of, like, this double loop of, like, you can edit the agent and the agent's instructions, and you can see the agent operate. And it's sort of ping pong between these two modes. Because if the agent screws up when it operates, you can just go in these instructions, and you're like, okay, like, keep this in mind moving forward. Okay, so now it's, um, it's looking at your, like, profile to, like, know how to pitch you.

AI assessment note: “None. I just, like, the agent builder takes care of all of that for you.”

Q way, I'm loving this, but where my mind is at is great flow. You're really good at designing these really smart kind of workflows, these AI, you know, agents. How do I get started? Like, how do I know where to start? How do I design it? Like that to me is a really, really important step that I think a lot of people are going to miss, including myself.

A No, 100%. There is, there is a little bit of a learning curve, but you know, you mentioned Zapier, you mentioned, you mentioned Make, like, I think that the, the, the folks here who listen to us, I mean, most of them are pretty tinkery. They can figure it out. We have like a Lindy Academy. We have like a bunch of templates on this tool, but yeah, I mean, let me, let me just show you, I'll create a Lindy right now, just to, to give you an example. Um, I'm going to create a new Lindy and I'm going to make it so that she, Um, what's an example of this thing I want to do? I want to, when I receive an email that is very time sensitive, uh, I want that Lindy to ping me on Slack, for example. And so I'm going to go, um, Gmail, so it's just a trigger. When I receive an email, I'm going to create a condition. It's, again, that's what I mean by like, it's AI throughout. So right here, I'm like, uh, the condition is just natural language. So go down this path if this email is urgent. And time sensitive. Okay. It's really that simple. It's like a small prompt. And here, what I'm going to do is, okay, at that point, you Slack, you send me a direct message. Okay. The user, I'm going to enter my own email here, but even if I didn't do that, it would probably still infill that I wanted to message me. And here, I can enter a prompt, and I'm going to be like, um, send me a message about This tim…

AI assessment note: “We have like a Lindy Academy... let me just show you, I'll create a Lindy right now”

Q One of the things I noticed, by the way, is that, uh, clode. 3.5 is default selected. Should people keep that default or should they change it?

A I think you should keep it. I think that's like a great default. Um, basically start with Cloud, and then if you find your agent is too dumb, graduate to, we've got Gemini, we've got O-one. O-one is, is really good. It's really expensive, but it's really good. Um, I use O-one for research tasks. Um, and if your agent is too expensive, use Gemini Flash. And most of the time, Cloud does a good job and Gemini Flash does a good job. Yeah, so right here, I just, I mean, you can see it took me, in the time I spent to answer this question, I just reached out to 20 unicorn founders about, about what we do. And it, it, it, it's actually, uh, personalized, right? So here it's, it's reaching out to the founder of Epic, like Tim Sweeney, and it's like, hey, like, this is how we, Lindy can help you, uh, streamline developer support, manage and organize data from your free games program, and Epic first run initiative. So it's, it's really super, super customizable and super personalized.

AI assessment note: “I think you should keep it. I think that's like a great default.”

Q One of the things I noticed, by the way, is that, uh, clode. 3.5 is default selected. Should people keep that default or should they change it?

A I think you should keep it. I think that's like a great default. Um, basically start with Cloud, and then if you find your agent is too dumb, graduate to, we've got Gemini, we've got O-one. O-one is, is really good. It's really expensive, but it's really good. Um, I use O-one for research tasks. Um, and if your agent is too expensive, use Gemini Flash. And most of the time, Cloud does a good job and Gemini Flash does a good job. Yeah, so right here, I just, I mean, you can see it took me, in the time I spent to answer this question, I just reached out to 20 unicorn founders about, about what we do. And it, it, it, it's actually, uh, personalized, right? So here it's, it's reaching out to the founder of Epic, like Tim Sweeney, and it's like, hey, like, this is how we, Lindy can help you, uh, streamline developer support, manage and organize data from your free games program, and Epic first run initiative. So it's, it's really super, super customizable and super personalized.

AI assessment note: “I think you should keep it. I think that's like a great default.”

Q When should people have a human in the loop, and when shouldn't they have a human in the loop?

A Great question. If an AI agent could embarrass you, you should probably insert a human in the loop, at least in the beginning. For the first few cycles, and you're going to work on it. It's the same thing as training a human. It's like you just, think of it as like you just onboarded an intern, and like you don't, at first you're kind of on their back, you kind of watch what they're doing, that's how you think of it. And soon, we're building a feature right now that's making it so that the agent can learn from your feedback. So as you insert a human in the loop, if you correct it, little by little it's going to learn. Um, that is it. Right here, I have like an MVP recruiter. I'm gonna rename her and just call her the recruiter. That's it. I've got, I've got a small recruiter. And again, that's the beauty of it is like, you start small, like this took me two minutes and then you add to it and add to it and add to it. So for example, here I could make it send reminders. Like if the person doesn't reply, I could make it send two, four, five reminders. Right. Uh, and actually I'm going to add a last step here. So it's like, Hey, average that to everyone. Um, I'm going to start here, and I'm going to be like, I want to hire, like, find me five account executives, uh, working at, what's a good, what's a good place to hire from for account executives? HubSpot.

AI assessment note: “If an AI agent could embarrass you, you should probably insert a human in the loop”