Automating business workflows with Lindy AI agents

[00:00:00] Flo: Yeah, absolutely. Yeah. Thanks for having me, Ben. Uh, everyone, great to meet you. I'm Flo. I'm the founder and CEO of Lindi. Uh, we're building an AI platform that lets you build your own AI automations in no code.

[00:00:11] Flo: Um, and so, you know, what I'm talking about today is, uh, how Lindy is helping me day to day with my own personal productivity, uh, and, and the parts of my business that I have automated with Lindy.

[00:00:21] Ben: Awesome. What, what led you to this point of starting Lindy? I'm just interested in that story actually.

[00:00:28] Flo: Yeah. I, I, I always, I'm actually always surprised when I get this question because I'm like, how could anyone work on anything but agents right now?

[00:00:35] Flo: Um, it was, I just got very sucked in. It was like after our previous startup, You know, it was, it was clear that like the previous startup, we needed to pivot. And he was like, Hey, what's the most awesome, exciting things we can work on. And it was just very, very clearly AI agents. It was before chat GPT, by the way.

[00:00:51] Flo: Uh, but it was just like, Hey, like LLMs are getting to a point where they can just like, not just talk to you, but actually perform actions for you. Um, so we just, in a way we just got sucked in. It wasn't like the most strategic thing ever.

[00:01:03] Ben: Oh, interesting. I love that. Actually, a lot of people, including myself.

[00:01:07] Ben: Like, the claim to say that you're in AI before it was cool is like, we launched, we were out before ChadGBT. So, Ben Spice launched like, I think it was like six weeks before. So, it's always like a good stamp to say, look, I was here before that happened. So, uh, yeah, it's nice to hear you do the same.

[00:01:23] Flo: Yeah, which, which in hindsight, it was a bit crazy to try to make this happen before GPT4.

[00:01:27] Flo: Because like, it barely worked. It was like, not be pretty usable.

[00:01:32] Ben: Yeah, exactly. Um, but yeah, excited to see what, um, what you're going to show us today. Um, and yeah, show some workflows and for everyone else, if you've got questions during the, during the talk, then put them in the chat and I'll try and squeeze them in.

[00:01:44] Ben: And then we'll do a Q and A at the end, 15 minutes or so, um, for any questions you have. Um, but yeah, I can pass it over to you now to share what your, what you've got.

[00:01:56] Flo: Yeah, absolutely. Um, I just have a few slides, [00:02:00] uh, as an introduction and, uh, you know,

[00:02:04] Ben: Oh, I think you've just gone, your sound has gone.

[00:02:12] Flo: Hello. Hello. Yeah.

[00:02:13] Ben: Okay.

[00:02:14] Flo: Uh, great. So, um, you know, I'm talking about like AI agents for the rest of us because I won't tell you a thesis about AI agents is that, uh, agents are awesome. They're very exciting. Like that. That's why everyone's here today. Uh, they don't really work. Uh, and, uh, when they work. Uh, it's very hard to make them work, right?

[00:02:33] Flo: It really takes an engineer today mostly to truly get value out of AI agents. Um, and so the entire thesis of, of Lindy as a company is like that, that's too bad. Like AI agents are awesome. They're super powerful. Uh, you shouldn't have to be an engineer in order to leverage AI agents. Um, and by the way, I, I don't mean this to like throw shade on the rest of the category.

[00:02:53] Flo: I think it's, it's all like a very natural evolution of the industry. Uh, I always compare it to the personal computer. Right, the first personal computer, arguably, was like the Altair, uh, in 1978 or something like that. And it was, like, not really doing anything, but it was very exciting to a lot of people.

[00:03:12] Flo: Uh, and then three years later, that's when you had, like, the Apple II. And, like, the Apple II was really the personal computer that worked and that was useful. You had spreadsheets, you had games, but it was still, like, command line. It was still really hard to use. And it took seven years between the Apple II and the Macintosh.

[00:03:28] Flo: For like the personal computer to actually become useful and usable by mere mortals. So that's, I think today, AI agents or between phase one and two, it's like, again, it doesn't really work most of the time. And when it works, it's super hard to use. We're working on building what we call like the Macintosh of AI agents.

[00:03:49] Flo: We want an AI agent that works reliably all the time and is super easy to use by any non technical person.[00:04:00]

[00:04:01] Flo: And so, when I say that, people often ask, like, hey, how does it compare to existing solutions? You know, today the biggest solution, if you're building AI automation, is maybe Zapier. Um, and so the way we compare today is, one, we're truly AI native. Um, and so by that I mean that, um, AI in, in Indy, uh, automations are not, is not merely like tacked on to like a legacy automation product.

[00:04:27] Flo: It's really infusing the entire product. And so I'll, I'll give a very concrete example now. I, um, I have this use case where I have got some investment properties in the south of France, which by the way I don't recommend anyone do, um, and uh, I, I received these emails to confirm like, hey, someone booked your place.

[00:04:47] Flo: And I want these data to go into a spreadsheet when I receive these emails. If you wanted to do that with, like, a legacy solution, you would have had to write an email parser, right? And so it's super hard, like I said, not very easy to create. It's very brittle, you may not even be able to do it. The moment the email changes, the entire parser changes.

[00:05:09] Flo: Creating this with Lindy, I obviously have Lindy doing this, and creating it with Lindy is like two nodes. So I have two workflows here. I have one handling reservations and one handling cancellations. So it's literally two nodes. And it's like, hey, when you receive an email, and you can see here I have a filter that's like, if the subject contains resolution confirmation, so I only pay for the task if it's actually satisfying this criteria, then you append a row to a spreadsheet.

[00:05:36] Flo: And this is the part AI throughout the entire product. Let me show you. Oh, I think I need to reload the page because I was on like an old account.

[00:05:54] Flo: There you go. So you can see here in the values of this spreadsheet, these fields here are in [00:06:00] purple because I have filled the value of this step in the automation with a prompt for an AI. So here literally what I can say is like, hey, extract from the email the date the reservation begins on. Extract the amount of the reservation, extract the number of days, the number of travelers, and so on and so forth.

[00:06:16] Flo: Like it takes like two minutes to create this. Right? So again, I don't have to create another node in between for like, extract or like a parser. It's just truly AI native, AI is like infusing the entire, the entire thing. It's maintaining context and you can pick the task back up at any point. So again, if you think of legacy, uh, automation platforms, it's like a ZAP is just a thing that runs and then is over.

[00:06:42] Flo: Well, it's like most tasks in your business are like long lived. You need to pick them back up at any point. Um, here's the example is, uh, I'm going to showcase a much more complex Lindi because I've been adding to it over the, over the months. And it's my meeting recorder Lindi. It's actually really meta because this is the meeting that's in this big thing right now.

[00:07:01] Flo: Uh, and so you can see, you know, really it's actually, it looks very complex, but it's actually pretty simple. It's like when a meeting begins, she records it. She sends me the summary. She sends me coaching notes. That's the part I'm going to talk about. And then after sending me the coaching notes, you can see after the email is sent, she goes and she does a bunch of other stuff.

[00:07:20] Flo: Mostly she disseminates the notes on Slack. So for example, you know, we have this integrations channel on Slack and whenever we have a meeting with a customer and they ask for integrations that we don't have, we, uh, it reports them. It's like, oh, hey, during the call, they asked for Zendesk and Airtable and we actually built these integrations since then.

[00:07:47] Flo: And so here's the part that I want to highlight is so you'll see it sends me coach coaching here, uh, in the in the after the meeting. So I received these emails from Lindy that gives me feedback. And it's [00:08:00] pretty good because, um, I received these emails where Lindy tells me, Hey, like, for example, here's the explanation of Lindy's new features was a bit technical at times.

[00:08:10] Flo: And I'm able to follow up in the email thread. What did you find technical about my explanation? And so when I do that, Lindy picks the automation back up here at this branch after reply received, even though she had moved on, she's able to like pick it back up and say, okay, send a reply if you receive a reply to your email.

[00:08:30] Flo: And so here she was able to say like, oh, I found the explanation too technical. And so she's restoring the context of the entire meeting that she just attended. So it's like, oh, you know, you said deterministic, you said context window, you said agent state, that's, that's too technical. And that can happen not just over email, but over any channel.

[00:08:47] Flo: So for example, this happened just yesterday. Lindy also posts the details of the onboarding sessions. That we have with customers in like our Slack general channel. And it's really handy because that way the entire team can see everything that happens and the customers that we onboard. And so here it was like, Oh, you know, um, we had issues with, uh, Salesforce opportunity creation, and so I'm able to go into the Slack thread and follow up.

[00:09:13] Flo: I'm like, Oh, what happened? And you can see, you can see the timestamps. It was like in less than one minute, she responds and she's like, Oh, this is what, this is what we saw. Right, we successfully created contacts and accounts, but like creating the opportunity, we had an issue. And so it's, it's really awesome because now there's like a bunch of meetings that I don't need to go to anymore.

[00:09:32] Flo: And indeed, just post these messages on Slack and all the team can have a conversation with with Lindy about the meeting.

[00:09:38] Ben: This is, this is insane. This is really good. I mean, I've never seen anything where there's something further than the automation. So I think that's something that you sort of leaning into, right?

[00:09:49] Ben: With the AI native stuff, which is, okay, it sends me an email saying you didn't mention this, or you said this in this way, but be able to respond to it. And then it [00:10:00] actually pull out more context is something I've not never seen actually. And I think that that piece now I've seen it, I can't unsee it. And now I've seen it, I think.

[00:10:11] Ben: That should happen all the time, like it's so much, I think AI summarizes and sort of gives you context on so much information that you're then supposed to do something with it and often you might go copy that to somewhere else to then go and do something, maybe, but like me, I record all my meetings. I don't think I've once gone in.

[00:10:32] Ben: And checked anything, um, or like reread any of the transcripts. But I think there's things still like doing these workshops, for example, how, how to get better at something. Like, how do I, I guess if this is the prompt you're showing us, or you can talk about this maybe is if it's like coaching notes specifically, if it's like a customer call, I assume there are certain things you're trying to look for, which is, did I explain this properly?

[00:10:57] Ben: Did I get the benefits? Did I get the features? Right. Did we talk about pricing? Did we do the talk about next steps? So. No, you missed some of these pieces and almost like helps you then create what could even be like a sales script for like, these are things that we need to be mentioning on every single call.

[00:11:13] Ben: Right.

[00:11:15] Flo: 100%. And here, that's really what I want to emphasize is like, we at Lindi did not build a sales coach. I, as a user of Lindi built a sales coach using the legal blocks that Lindi gave to me. And it's super easy, right? Like you can see, it's just a prompt here. It's like, Hey, I set the body of the email here as a prompt.

[00:11:33] Flo: And it's like, Hey, right. Feedback and bullet points. By the way, that could have done better. You know, I'm French. So it's like, was I unnecessarily confrontational? I'm not confrontational at any point. Um, and, and then I just told it, Hey, if you receive a reply, you send a reply back. But again, you can customize it as much as you want.

[00:11:50] Flo: So for example, we've had customers in sales teams, what they have is like before Lindy sends the coaching email again, because every step is aware of the entire context of everything that happened to [00:12:00] before you can insert a step before here. That's like, Hey, we've got our best practices, our sales best practices, for example, In Ocean or in Google Doc.

[00:12:11] Flo: And so before you send me your feedback, check these notes. Pull this document. And now here, in the prompt, you could, you could refill to these notes. You could be like, hey, you know, refill to the notes that you just uh, pulled from Google Doc. Uh, and send me feedback uh, following them. That's literally this simple.

[00:12:33] Ben: That's awesome. And then, yeah, with Alex, I just asked about the, the send replies step. Is that it says a default state or is it something you have to sort of set up yourself?

[00:12:44] Flo: Yeah today you do have to send it to set it up yourself We are reconsidering that we're thinking of making it just like a checkbox on this step here That just is enabled by default.

[00:12:55] Ben: So is there a prompt then in that lindy mail send reply? Step as well. Yeah in that node

[00:13:01] Flo: No in this one not even because it's just pretty like lindy will be small enough She's like hey when you send a reply you send a reply back and so she'll figure it out Um, we also do offer uh templates Um, and so if you go to like the home tab here on Lindy, if you go to, I believe it's in the meeting category, we have a meeting coach template.

[00:13:20] Flo: And so if you add the meeting coach template, it's a much simpler template than the one I just showed you, because it's really just that. And so the send reply node is all these there by default.

[00:13:30] Ben: And then Vincent's asked, what, how does the context window work? Like if you've got 30 nodes and there's different documents and all the rest of it, like how, how is that getting pulled through, um, context?

[00:13:43] Ben: Okay.

[00:13:43] Flo: Well, it kicks off a new note, uh, sorry, a new thread with a new context window every time you have a new calendar event. And if you want to pull context from previous events, then you've got to, you've got to do that after this. So it's like, get a Google Doc or get a memory and whatnot. [00:14:00] Uh, we also do a ton of stuff behind the scenes to manage the context window, but realistically, modern LLMs, they have 200, 000 tokens, 1 million tokens of context window.

[00:14:08] Flo: You can, you can manage here on a pure node basis. You can manage which model you want to power each of these nodes. I use 3. 5 for everything now and I almost never run out of context window.

[00:14:20] Ben: Oh, interesting.

[00:14:26] Flo: Um, and so I think that's the third point about like how we compare is like super easy to use. Like this is not a secondary concern for us. That's, that's the whole reason we exist. It's like the Macintosh for the agents. And so, um, Our goal is for any non technical user, and we achieve that goal all the time, we see non technical users, jump on a call with us and try out Lindy.

[00:14:48] Flo: And from zero to, I get a Lindy that works for me, that solves my problem in like five minutes. And we invest a ton on that. So it's like the ease of use, yes, but also, uh, what we call the Lindy Academy. So we've got this, uh, uh, hub on the, uh, website, uh, where we have a bunch of help sections here. So we've got videos that explain to you how it works.

[00:15:11] Flo: Uh, we've got templates, we've got step by step guides. Um, and like I mentioned, we've got all of these templates on the homepage for the most frequent use cases that people want. And so very often that's enough. These templates are enough to bring you to 90 percent of what you need, if not 100%.

[00:15:28] Ben: How, how are you sort of experiencing, experiencing what's.

[00:15:35] Ben: I guess I came up with, so my previous company was like a no code education company, and I mistakenly thought everyone who can't code is my market. Um, but it wasn't that it was much, much smaller than that. It was still people need to learn the building blocks of these things because it was back then it was.

[00:15:53] Ben: You Zapier to connect everything, have air table for your database, have web flow for your front end or whatever. Obviously these, [00:16:00] these sort of agents run in your systems where you are and they're more chat based. You think it's a different. Do you think it's almost like a different time for people to be able to spin up stuff easier?

[00:16:11] Ben: And I assume you have, like, there's a Lindy bot on the site that can help navigate the academy and all that kind of stuff, the knowledge base to help you through it. And I think things like that are making it easier. I think Zapier has created Zap from a prompt and that kind of stuff. Like, do you think it's different these days than a few years ago with that cold start problem for like targeting the no code audience?

[00:16:35] Flo: I think it's different for two reasons. It's like one, um, the market has been educated. Now there is a no code audience that didn't really exist 10 years ago. And so people are increasingly comfortable. building these things because of solutions again, like Zapier, Airtable, Notion, and so forth. Two, the upside has become much, has become much greater.

[00:16:55] Flo: It's not just about building like a little automation that like, Hey, when you get to lead on Typeform, it's easy to slack anymore. It's really like, Hey, you can automate entire parts of your business and you can save an insane amount of time. And so now people are even more motivated to figure these things out.

[00:17:08] Flo: Uh, but yeah, I will say like, it's a very long term vision is like, you don't have to worry about the DNA or no code that you do, or, and you can just. Talk to Lindy and tell her what you want to do in English and she does it for you and that's it. And I, I actually think this is closer than people think. I think we're just a few years away from that.

[00:17:29] Ben: I hope so.

[00:17:35] Flo: Yeah. So, you know, what I had next is like, I wanted to paint a picture of what I use Lindy for all day. And, um, the scenario that I want to talk about here is recruiting. So this is an actual scenario that happened, um, in the last few weeks, actually. We have been hiring a product designer. So I'm going to talk about how Lindy's helped me, uh, outreach, schedule the interviews, uh, analyze the interviews, sit [00:18:00] with me in the interviews, and then negotiate the offer with the designer.

[00:18:04] Flo: By the way, we also hire product designers and we're hiring, uh, software engineers as well, so this is my email if you know anyone extremely talented. So, step one, outreach. Um, I have this Lindy right here. Uh, it's a design your recruiter and so you can see I give it any information I have about the designer so it can take really any input.

[00:18:26] Flo: It can take a LinkedIn link, it can take an email, it can take a link to a portfolio, but you can see here I gave you the LinkedIn link and so it's like, okay, hey, it's a LinkedIn link. I'm going to look for the email address. So I have a spreadsheet that contains a bunch of email addresses, but if it couldn't find it on the spreadsheet, it's going to go online, it's going to use prospecting API, it will find the email one way or another, and then it's like, okay, Hey, I found the email.

[00:18:48] Flo: Now I'm going to list your emails in your inbox just to make sure you haven't reached out to this person before. Okay. You haven't reached out before. So now I'm going to draft the outreach email. So it does this here and then I'm going to wait. And after the wait elapsed, if I have not received a reply, I'm going to send a follow up message.

[00:19:12] Flo: And then I'm going to wait again. So it, it manages the entire page. So you can see, I use this. I mean, you can see, this is how many times I used it yesterday. It was like five times and I've used it for hundreds and hundreds and hundreds and hundreds of design emails. So that's how she's helped me with the outreach.

[00:19:30] Flo: And then. So she reaches out, and then people reply. And so here, the next step is to schedule an interview with the person. And so at that point, what I do is I have another Lindy that's my meeting scheduling Lindy. And it's really simple. It's like, hey, you know, sure, I'd love to chat. And I'm like, awesome, Lindy will send us some, will send us some times.

[00:19:51] Flo: And all I do is like, I cc to the email my scheduling Lindy who has her own email address. I could also just like provision [00:20:00] another email address for this link under my own domain so people don't even know it's an AI, but obviously I want people to know it's an AI, uh, but you may not. And so, here Lindy replies, again you can see the timestamps, like this is in the same minute, like she replies in like 30 seconds or so.

[00:20:14] Flo: And she's like, happy to help you get this on the books, here are a few times that work. And then the person replies, it's like, okay, let's do July 23rd at 11am, and she's like, okay, awesome, I just sent the invite. And so if you want to see behind the scenes, you don't have to, but like, if you, if you want, you can like open Lindy, uh, and you can see behind the scenes how this is working.

[00:20:37] Flo: And this is the, this task, right? And so it's like, Hey, she received my email looping her in. She looked at my calendar for the best times to schedule an event. She sends the email reply. She received the reply saying, sure, let's meet at this time. And then she sent the calendar event. That's it.

[00:20:59] Ben: Could you walk us through the part where it uses different APIs to search like for an email address, for example, and, um, Danny's asked, and I'm also interested with, with the email part is, do you get like a email per Lindy that you set up? How does, how does all of that kind of work to be able to, I don't know.

[00:21:21] Ben: It feels like one of those. You don't have to build an interface if you've got email and you can like cc in different bots that you had or different lindys Um that you have to do these different tasks and like fire off different workflows Yeah.

[00:21:37] Flo: So the, the, the philosophy here is like, we want Lindy to meet you where you are.

[00:21:41] Flo: So ideally you never even opened it in the app. It just works for you in the background. Um, this is how it works. Like, so, um, the trigger here is we could eat like a Lindy mail trigger. And so I've, I've just provisioned, um, Her own LindyMail address for this Lindy. Um, but you could also, again, you could [00:22:00] just create your own trigger here and make it a Gmail trigger.

[00:22:03] Flo: And then just create your own Gmail account to this Lindy. So then people wouldn't even know it's a Lindy, right? And then you would just connect this guy right here to this guy right here, and boom, now you've got the Lindy that's connected to your Gmail. And this is how it uses the API. So it's like, hey, if I've cc'd you to find times, You look at my calendar, the best times to meet, then you send a reply.

[00:22:26] Flo: Then once you receive a reply, did they accept the meeting? If so, victory, just send a calendar invite. If not, well, you're going to have to go back and send a few more times, right? And even once the calendar event is sent, she handles like, hey, are they asking to reschedule afterwards? If so, delete the event and go back here, offer some more times.

[00:22:48] Flo: And wait for their reply and schedule the calendar event.

[00:23:00] Flo: So that's scheduling. Um, then interviewing. So this is the whole meeting, uh, recording use case I showed. Again, the really good thing about it is that you can ask follow up questions at any point. I'm not gonna go over that, uh, again too many times, but like, You can ask them by Slack. You can ask them by email.

[00:23:17] Flo: You can ask them in the app. And so it's really handy. Um, I have also, again, set up my meeting recorder so that she, so you can see when she takes the notes, she adds them to a Google doc. And then when I send a message up here, so independently of a, of a, of the other trigger, which is, Hey, when the calendar event starts, you record the meeting.

[00:23:39] Flo: When I trigger Lindy and I send a message, if I'm asking a factual question about a past meeting, What she does is that she searches the Google Doc. And so this is basically doing RAG over my Google Drive. And so I've configured it with my Meeting Note Google Drive folder. So you can see here, there's like 150, 000 And [00:24:00] then it's just pulling up like the notes from a Google Drive folder.

[00:24:02] Flo: And so at any point, I can go back and I can be like, Lindy, what did Ben say about this again last week? And boom, she just finds it and then sells the message.

[00:24:14] Ben: Yeah, that's a really good way to do that, I guess, is it organizes it for you as well, and then uses that organization to find the information itself in a natural language. Because I think Having all of these pieces before the workflow. And I think this is generally why people talk about AI wrappers and the better version of AI wrappers, which is if you think of like a legal workflow, you need all of the specific legal workflows, the knowledge and everything else baked into it before it can be valuable with AI on top, whereas with what you're doing, you're building, allowing the user to build these workflows or enable them whilst also baking in AI every step possible.

[00:24:59] Ben: So the whole workflow instead of. me doing this with fireflies and then taking that and going or using his app to then send that somewhere else and do that somewhere else. It's all like, it's all in one place. Um, yeah, it's really impressive. 100%.

[00:25:20] Flo: Again, Lindy sends me feedback. He sends me coaching. And then step forward negotiation. So, I mean, if there's folks here in the, in the zoom who are salespeople, they know that time is of the essence for any deal. And so you really want to close the deal very quickly. And I think recruiting is sort of a sales job really.

[00:25:38] Flo: And so I have this other Lindy here that pings me on Slack anytime I receive an email that is time sensitive. And again, it's super easy. It's three nodes. It's like email received, is it time sensitive, ping me on Slack. And so you can see here, I did this other message and don't need to give you an idea of the kind of messages I received from this Lindy.

[00:25:56] Flo: But here the candidate, the interviews went super [00:26:00] well, we were all about to make an offer. And then they send me an email at some point and I'm like, hey, can we chat right now? Can we chat today? And, you know, like, I don't know if he received an offer from someone else and whatnot. And so Lindy alerted me.

[00:26:12] Flo: And so I replied and I'm like, oh, absolutely. Again, I loop in my scheduling Lindy. Because I was like, Hey, can we meet in the next two hours? And I don't want to have like, keep an eye on my inbox for the next two hours. So, Hey, Lindy, like I'm adding you to the thread. So if he says yes, then you can send the invite.

[00:26:27] Flo: He said yes for a meeting, literally half an hour later. And so then we met and happy ending. We made an offer. Um, and, and he actually just started, uh, two days ago.

[00:26:42] Flo: So again, that's how Lindy has helped me through this entire journey. It's like outreach, scheduling, interviewing offer. I mean, at this point I, I use Lindy 20 to 30 times every single day. It's like, it's just like my co pilot, like all day she's next to me and I'm just using her to automate a ton of tasks.

[00:26:57] Flo: Like I always say, like, don't do the same task twice. Like if you find yourself doing the same tasks twice, it means you've got to create an AI engine for it.

[00:27:08] Ben: What are, what are instances where. you have tried to do that and then have struggled. And I suppose, I guess there's two versions of that. One is you've struggled because the capabilities aren't there or something, and then you've added them. Or what about another version where it's like, Oh, that that's not quite possible.

[00:27:27] Ben: It doesn't quite work that way yet.

[00:27:29] Flo: Yeah. Well, you may notice here, we're missing a step zero, which is the finding the candidates in the first place. Right. And, and here, this is. Missing for a few reasons. One is Lindy today can't yet scrape the Internet and we're adding that ability in the next few weeks, so I'm super excited about it.

[00:27:49] Flo: She'll be able to go online and dribble and whatnot and just like LinkedIn and just find a bunch of candidates and then loop through each of them. So that's another thing we're building is this idea of like [00:28:00] looping through each of like a large group of data and then reach out to them and so forth, right?

[00:28:04] Flo: That's the first thing that was missing for finding the candidates. The second thing is, um, um, selecting the candidates I want to reach out to. Like, today, I still want to be in the loop of like, hey, this is a candidate I like, this is a candidate I don't like. We're building another system, um, it's basically like an RLHF system.

[00:28:24] Flo: It's really cool. It's like, your Lindy is going to be in training mode, and so you're going to be able to insert a node in the, in the Lindy that's like, hey, at that point, decide if this is the kind of candidate that you want to reach out to. But you're in training mode for this node. And so you decide, but you check with me as you decide.

[00:28:42] Flo: And so basically the ninja is going to go online, find a bunch of candidates, and she's going to be like, I think this is someone we want to reach out to. And then you can be like, yay or nay. And you can provide an explanation when you say yay or nay. And so you do that a few times, and it's actually really surprising.

[00:28:56] Flo: It doesn't take that many examples for Lindy to start to get it, like 20, 30 maximum. And like little by little, you find yourself saying like, overriding Lindy's initial decision less and less often. So she's like, Hey, I'm about to do this. And you're like, yeah, that sounds good. Yes, yes, yes, yes, yes, yes. And at some point you just take her off training mode because she's just got the idea.

[00:29:17] Ben: Oh, that's interesting. Yeah. Cause I've seen people talking about just having approval steps. But then I haven't seen anyone look at this problem in, in terms of you are, which is, if it's an approval step, it's because I've got maybe a taste or a specific thing in my head that is deciding that yes or no approval.

[00:29:37] Ben: But actually that could be an opportunity to train it specifically. And like you said, you, it's not fine tuning a model. Effectively giving like 5, 20, 30 prop, 30 shot examples to, to that node to then use that as a way to filter or go through the steps before it. Um, yeah, I've not seen [00:30:00] that before either.

[00:30:00] Ben: That's awesome.

[00:30:02] Flo: Yeah. And we already do have like the asking for approval. It's like this toggle on every, uh, node that you have. But yeah, we're building this like RLHF on top of it as well.

[00:30:16] Ben: And when, when you like respond or can respond to. Lindy, when it's in, say it's in, within its workflow, it's like halfway through and it's checking something with you. Can you respond and say, Oh, actually update your instructions or like do something different here? Like, Oh, actually you are sending me stuff that I don't want to be doing.

[00:30:36] Ben: Like, it's not the right flow for me. Like. Instead of checking this document, check this document instead. Um, do you have anything like that yet? So, no,

[00:30:46] Flo: and by design, because again, we want to make it reliable. And so, today, Lindy is following strictly the workflow that you define here. And so, within the confines of this workflow, you can impact their behavior through just English.

[00:31:02] Flo: So, I'll give you the example of the meeting, the designer recruiter that I have. Here, if I send her, I'll send you an email, Ben. Um, And I'll show you, uh, do I have your email address right here? So Ben Tussle, and then it's like, in your email, in your email, mentioned that I'm a big fan of his newsletter. So that's going to work because it fits within the workflow that I've defined for her, but I'm not able yet to override that workflow.

[00:31:31] Flo: That's something that's going to come, but not, not quite yet.

[00:31:34] Ben: Yeah. Cause I think there's other instances I've tried where you can update the instructions by messaging saying, Oh, and also add this to instructions. But it's not clear, I think it's the same with building a GPT, right, it's not clear how much of that is going to override what I've already spent God knows how many hours trying to, like, trying to specifically get perfect.

[00:31:55] Ben: And then which bits of that might mess up other parts of that flow. And if [00:32:00] it's a big flow, you don't want to mess up with that, right?

[00:32:04] Flo: Exactly. Um, oh, look, this is funny. She's, she's refusing to reach out because of the workflow that I've defined, which is like, hey, she's not reaching out twice to the same person.

[00:32:12] Flo: And so this is good. She's like, oh, regarding the expert workshops, this is very meta. Um, yeah. Good, Lindy. Yeah. Yeah. Um.

[00:32:24] Flo: I mean, there are like more exotic workflows that I also support. Like, this is really more experimental, but you know, if we're, if we're having fun, why not? Um, there's this Lindi I call the AI console. Um, and so it's, again, it's experimental, but under the hood, it's actually two Lindis. They're talking to each other.

[00:32:43] Flo: And so I can say, for example, anytime I take a decision now I have like, I face like a Thony business decision. I talk to this Lindy, and actually what what it does is like it, it talks to another Lindy. So a little bit of, a little bit of a schizophrenic Lindy. And one Lindy is arguing in favor of the decision and the other one is arguing against the decision.

[00:33:02] Flo: So it's like for like binary decisions. So like for example, I can be like, Hey. We, oops, we have been approached, uh, by a big enterprise customer. Uh, we're very excited that they're asking for a lot of custom features. Um, I'm wondering whether we should pursue them or drop them. Okay. Again, it's, it's in beta, but you'll, you'll get a feel for how this works.

[00:33:28] Flo: Um, so what she's going to do is she's going to send a message to the other Indians. So you're going to see here in the sidebar right here, there's going to be another task that pops up because she's going to send this other message to the other Lindy. And so boom, it was like, Hey, I'm reaching out on behalf of flow, trying to make a decision.

[00:33:43] Flo: And so it's all going in favor. And so here, this other Lindy now received this other message and is going to respond arguing against.

[00:33:55] Flo: And so she receives the reply and she's like, Oh, well, well, the potential benefits of [00:34:00] pursuing this big enterprise customers are here. And so they can, they can go back and forth for a long time like that, right? Sometimes like they run for like 30 minutes or whatnot. And then once they've reached the conclusion, they send me a message.

[00:34:10] Flo: And they're like, Hey, we've deliberated for a long time. And this is the conclusion we reached. Right. Um, so again, I don't like outsource my thinking to, to this AI, like I still, I still make my own decisions, but you'd be surprised, like it's, it's good, like the debates are good. Very often they make very good points.

[00:34:25] Flo: I hadn't considered it. It's good.

[00:34:27] Ben: That's, I love that. That's, I can just think of times where we were doing, should we hire an agency to help with our growth? And I sort of, prompted Claude with a bunch of financials and everything else. And sort of did like a, should we do this? And it always leans on the positive side, I suppose.

[00:34:48] Ben: It's like, Oh yeah, you could. And then you forget that in a co founding situation or in just like a group situation, there's always different differences of opinion that actually would make some valid points that you wouldn't necessarily consider. And same with, so if, if we've taken certain sponsors on, Sometimes they come with, Oh, we actually need to have tracking pixels and where we've got payment terms of 60 days.

[00:35:16] Ben: And I kind of had to have a checklist of how I like to run my companies. And I could almost, I assume, and you can show us this in a moment. Like if there's. If you can sort of give them their own context of like, why, which, which lens do you look at certain things through? So it would be, okay, we've got another, um, invoice from this customer who wants to sponsor, but they want 60 day net payment terms and they want to put a pixel in and they want reporting after every email.

[00:35:45] Ben: And they want to see the email before it goes out, all of those things. And if in these like defined checklists, I was like, these are all the things we don't want to do and the reasons why we don't want to do them. And then that could sort of, I don't know if that is worse actually for this conversation, because then it's me.[00:36:00]

[00:36:00] Ben: Me being the arguer kind of going between them, but I don't know if that's how you've looked at this where each each sort of Council has their own baked in things to go off based on like your principles.

[00:36:15] Flo: Yeah, so i'll show you I have these other use cases like an AI focus group So here it's, it's different because it's actually, so we were redoing our landing page right now.

[00:36:25] Flo: And so the way it works, you can see it's a very, very long debate here. And so the way it works is like, I I'm giving it like an initial draft for the new landing page. And then it's actually three Lindy's in conversation with each other. It's, it's one product marketing manager Lindy. So it's expert in drafting landing pages.

[00:36:42] Flo: So it's drafting the outline of the landing page. And then he talks to two Lindy's with two different personalities. So one of them, okay. You can see here, there's this guy who is a 30 year old tinkerer and solopreneur running a small travel agency. And so he rates his excitement on a scale of 1 to 10, and he rates his confusion on a scale of 1 to 10.

[00:37:01] Flo: And then there's another, uh, person who's like a 38 year old small business owner who can figure out software but isn't an engineer. Alright, so like, less tech savvy. And so, rates for excitement and rates for confusion. And so, you can see she's quite confused. Like, 10 confused. And so, they basically keep going back and forth.

[00:37:17] Flo: And she's like, oh, dear Focus Group, thank you for your valuable feedback. And so, They're just, and you can see this is a long conversation, they just keep going back and forth and at some point, uh, it's like, oh, victory, right? It's like, the final positioning has been successfully refined, both of them are maximally excited and minimally confused.

[00:37:36] Flo: And so then I'm like, okay, summarize the exchange in bullet points, let me know the main changes you've had to make. And it's actually really good. Like there's, there's actually some really good points here, um, about, about the new landing page.

[00:37:47] Ben: Yeah, that's interesting. Cause again, I've, I've done a similar thing there and sometimes it's a case of I mean, I don't know if you've heard of, like, there's lots of different startups looking at sort of synthetic data or synthetic [00:38:00] users and you, like, lots of people can do this and I think we've got some tutorials on it where create personas of the types of people who you already have signed up or you're looking to target and then rate your copy, your landing page, whatever it is, Based off of these personas, but it seems so much more interesting and probably a better outcome to have them talking together in one place and coming to some conclusion rather than you almost trying to coax that out of these different, these different pieces that you're, you're looking at.

[00:38:34] Ben: Yeah, that's really interesting. And so with this, I assume, can you show us actually what it looks like to build one of these kind of, one of these Lindy's? Sure.

[00:38:44] Flo: Yeah. Um, right here. So, oh, you mean like one of these like multiple societies of Lindy's? Or just like a simple Lindy?

[00:38:53] Ben: Yeah, no, one of these, yeah, one of these multiple ones.

[00:38:56] Flo: Yeah, so there's really multiple ways to go about it. Uh, this one I hacked together by like what we mentioned before, which is like this ability to give them, uh, an email address each. And so you can see here, it's like three Lindys. It's like, Hey, I send you a message and here you are an expert product marketer.

[00:39:12] Flo: You draft the outline for landing page. And then you begin your message by their focus group. And then, um, why is it sending a message? No, I think I think this one was like a I think this one had been like messing with Let me show you the other one. I mean basically it should have sent an email, but it's it's the same principle Actually, i'll just go back to it.

[00:39:33] Flo: So it It should have just sent an email. I think I'm just in the process of refactoring it. Um, so it sends an email to the people and, oh, in this one, I basically just changed them like this. And so it's like, hey, you are part of a focus group giving feedback about the landing page. You are a 30 years old thinker and a slow pruner.

[00:39:54] Flo: Give feedback, rate how excited you are on a scale of one to 10. And then, you And this one is like the other person's like, [00:40:00] you're still the eight years old, small business owner, give feedback and so on and so forth. And then, and Vincent asked like, Oh, can I set a limit to how many times they go back and forth?

[00:40:09] Flo: And like, the answer is yes, because like you can just insert a condition here. And it's like, um, well, in this case, you didn't do it, but you could totally have conditions like, Hey, if you've gone back and forth for more than 20 rounds, then you, you stop immediately. Right. And otherwise you just keep looping.

[00:40:25] Flo: Um, yeah. Uh, and then there's a condition here. It's like, hey, if both people in the focus group are 10 out of 10 excited and less than 2 out of 10 confused, then you send this final message that's like, victory! And then you explain the final positioning and summarize the whole conversation. I love that.

[00:40:44] Flo: That's one way to go about it. Um, The other way, which is in beta is again, we're building disability for agents to really send messages to each other across, across like a Lindy channel. And so you can see here, it's like this thing is the one arguing for this one is the one arguing against. And so you can see it does this agent send message action here.

[00:41:04] Flo: And so it's sending a message to the AI console in D. And so it's, this is the message it sends. This is the prompt I've given it for the message to send. And then this is that AI console and he's like, Hey, you received the message. And you send a message and here you are arguing against the decision, right?

[00:41:22] Flo: And so because you're sending a reply, then there's like a reply received here, and this is where there is a condition. It's like, all right, either the debate ends or the debate continues.

[00:41:31] Ben: Okay, yeah. So then why isn't this also created like a loop system?

[00:41:40] Flo: In a way, it is a loop. Because you can actually see like, okay, the debate continues, then you send a message.

[00:41:46] Flo: And then after the reply is received, you go back to the condition. And then it just keeps going back and forth like this. Okay.

[00:41:51] Ben: Okay. Yeah. I think, yeah, maybe I just thought, cause they're two separate flows. It doesn't look like the other one where it's like loop back into this [00:42:00] conversation, but I guess that argue against is just to kick off the argument against it.

[00:42:05] Ben: And then the actual conversation happens. That's right. That's right. That makes sense. Um, yeah. Vincent's also asked, can he connect a fine tuned model to one of the Lindy's? Yeah.

[00:42:21] Flo: Not yet. I'm actually surprised because I had never gotten this question until yesterday and this is the third time since yesterday.

[00:42:28] Flo: So I'd love to hear from Vincent, like where is this coming from and why would you like to do this?

[00:42:34] Ben: Yeah, Vincent, we're in the last 15 minutes anyway, so if you want to come on and ask or explain, I can ask you to unmute.

[00:42:45] Guest: Yeah, sure. Can you hear me? Yeah. So connecting this to the synthetic user story. If I'm talking to my consumer, I'm going to probably fine tune a model of my consumer based on what they've been talking about on social media and so on and so forth.

[00:43:00] Guest: So I want to be able to use that model. So I'm talking to the, the person of my consumer based on their conversations, you see what I mean?

[00:43:07] Flo: Yeah, absolutely. Right. That makes sense. Thank you guys. Um, Yeah, so we don't feel that yet. Like the way you would do this is instead of fine tuning, you would use rag and so you could very much like you could say like here, for example, before you send the message, you search a knowledge base.

[00:43:21] Flo: So this is the right thing. And then you connect it to wherever you're having knowledge about what your consumers said. So if you have a giant Google Drive folder or Google Doc with all what your customer said, you can connect it here. Or notion or like a website or whatever.

[00:43:40] Guest: Thank you. So in my case, in my use case, this is going to be for a stream type use case where it's going to be a live conversation. So it may not work because of the latency it's going to introduce. But I suspect that most of the use case here are not dependent on latency. It's like you can just wait a few seconds.

[00:43:59] Guest: But [00:44:00] in my case, I'm going to have conversation probably over voice. So it's going to become an important factor. This is why fine tuning will be a better scenario here.

[00:44:07] Flo: Right, right, right. That makes sense. Yeah, we're actually looking into voice use cases. It's not a huge priority just yet, but we're looking into them.

[00:44:13] Flo: And so, yeah, once we've, once we really want to invest into voice use cases, then we're going to have to like figure out the latency. In the meantime, you know, if you look at the model like GPT 4 oh, like the latency is really, really, really low. Um, We're also about to introduce, um, Gemini Flash, which is ridiculously fast.

[00:44:32] Flo: So, yeah, thank you. Have

[00:44:40] Ben: you seen anyone build Lindys like this that are going back and forth that are like, um, say you wanted to, this might be a bad use case, but say I wanted to create a blog, a blog share image for a specific blog. Um, and it's almost like design, creative, Back and forth saying actually no that image is too AI focused.

[00:45:03] Ben: It's got a robot in it. Let's like make it more Subtle or things like that and then seeing that sort of back and forth to iterate on Something creative. Yeah, I mean it could be a piece of writing could be a blog post right where it's just You're looping back and forth saying, no, this is not how we like to write blog posts.

[00:45:20] Ben: We like to do them based on knowledge base. We've got something Google drive, that kind of thing.

[00:45:25] Flo: Yeah. I mean, if you wanted to do that, it'd be trivial. It's just two nodes, right? So it's like a trigger for, uh, well, there's all the user sent message trigger here, but, uh, it's basically just this trigger followed by I'll delete all of these guys.

[00:45:40] Flo: And I'm just not going to save my changes followed by, uh, an action to reply to the message. And that that's it. Like if you do this. Actually, if I send another message, it will pick back up here. And so, that's what I mean that the loop is implicit in a way.

[00:45:54] Ben: Yeah.

[00:45:54] Flo: Every message I send is just, it keeps appending.

[00:45:57] Flo: And again, because it preserves the context of the [00:46:00] entire automation slash conversation at every step of the way. Like this, this right here is the conversation.

[00:46:07] Ben: Yeah, that makes sense. Yeah, because I would, I would assume that a lot of the, I mean, you see lots of people using chat GPT or whatever, going back and forth, copy, paste, copy, paste, iterating on certain things, but it seems much more, I don't know, it just seems a lot better to have it looping in a conversation that is iterating over time.

[00:46:28] Ben: And maybe you see those changes and you see that actually preferred version two or, um, but then you, you set the, you set the context for what each of those people are. One is a very strict content editor. And the other one is, you know, It's like an intern writing the blog post, um, or they're looking for, they're trying to target.

[00:46:46] Ben: We're looking to target enterprise users with this blog post, and we're looking to target small businesses with this one. Um, yeah, you can really bake in lots of context in each user, user, uh, AI user there. Um,

[00:47:01] Flo: yeah. Yeah, and the way you would do that is trivial is like, so you would have a send message here and here you would prompt it is like, respond to me thoughtfully, for example, and here, you know, there's always sometimes there's two branches sticking out of these notes.

[00:47:15] Flo: It's like one is like, hey, immediately after you perform the action, you do something, or after you receive the reply to this, you do something. But here I'm going to be like, you don't wait, like immediately you send another message. But this time, uh, you answer my question in a silly fashion. And so now, every time I send one message, you're gonna send two messages back.

[00:47:34] Flo: One serious, one silly.

[00:47:42] Ben: My brain's going 100 miles an hour. I don't know about anyone else. Um, okay. Anyone else in the chat have a question? If you do, please raise your hand. Alex,

[00:47:52] Guest: Over to you. Uh, yeah. Hi. Uh, thank you for this. This excellent presentation. Lots of [00:48:00] really exciting insights. And could you please share more on how credit spent is calculated?

[00:48:07] Guest: I mean, For which iterations inside a particular workflow, uh, you're charged and also how, um, credit span, uh, spent in threats is calculated as well.

[00:48:23] Flo: I'm sorry, credit span in what?

[00:48:25] Guest: Uh, in threats, like, like, like, like the one that we see right now, uh, in the window. Oh, this,

[00:48:32] Flo: uh, this AI console one? Yeah, yeah,

[00:48:35] Guest: for instance, the threat, yeah.

[00:48:36] Flo: Oh, it's right. Sorry. Absolutely. Yeah. Yeah, so we told you for token. Really? It's that simple It's like we told you for the llm tokens. We have a little bit of a markup on top of it So we manage the the account and the and the the api, uh key And so you can see here below each tasks that Lindy performs.

[00:48:53] Flo: You can hover and you can get like the cost of this task. And this is cents. Like, uh, this icon here is for our credits. And, uh, one credit is basically one set. It's like credit card points or airline miles or whatnot. And so these back and forth probably cost me of the order of like 20 cents in total.

[00:49:13] Flo: It's like 3 cents here, 4 cents here. If you look at like a medium schedule schedule, um, This one has been really protracted. 'cause this person kept rescheduling, but this is like 20 cents here. This was 20 cents there, and the price goes down by a factor of literally 10 to 100 x every every year. And so I foresee that in roughly one year.

[00:49:36] Flo: I think every, we have a flow of 1 cent per task, and so in roughly one year, everything that Lindy does is just gonna be like 1 cent. Just 1 cent. Unless it's a very complex task and like a very beefy model, it's just gonna be 1 cent.

[00:49:53] Guest: Awesome. Excellent. Thank you.

[00:49:56] Ben: Anyone else got any other questions and want to raise a hand?[00:50:00]

[00:50:03] Flo: Yeah, Denny was asking, can this become your own salesperson? Uh, yeah, absolutely. Uh, actually, uh, I don't know if you're going to be mad at me, Ben, for mentioning one of your competitors, but, uh, the competitor of Ben, who I shall not mention, um, tweeted about us recently. Um, and, uh, where is it? Ah, right. here.

[00:50:27] Flo: Uh, and this is actually like a big use case for us. And so it's like, Hey, intercepts leads that are coming, uh, for, for, for him. And like it, it, it performs research for, for, for, uh, for him. And it sends like an outreach email to the lead. And then it manages the entire backend pools. And so again, big use case for us, like we've got a customer, does they represent, uh, TikTok and Instagram influencers, like people with a million plus followers, and these influencers are harassed with, uh, uh, sponsorship, uh, uh, offerings, uh, in their inbox.

[00:50:59] Flo: And so what that agency does is that they, they, they take over these influencers inboxes and they, they remove the spam, they remove the non serious proposals, and then when a proposal is serious, They engage with the person, they negotiate the whole deal, and then when a deal is mentioned, they just go to the, the influencer and they're like, hey, you just made 50, 000, you have to make a video about this lipstick.

[00:51:19] Flo: Um, they replaced All of that with Lindy, like the entire thing runs entirely autonomously with Lindy. Um, even more funny, we have another customer who is one of these brands that wants to reach out to Instagram influencers and he set up a Lindy to do that. So, I don't know if these Lindys have crossed paths yet, but like in the future there's totally going to be a world where you have a Lindy that reaches out to an influencer to ask for a partnership and then the influencer has a Lindy answering that Lindy.

[00:51:47] Flo: That could totally happen.

[00:51:49] Ben: The most efficient deal in history, I assume. Um, that'd be awesome.

[00:51:55] Flo: Yeah. Yeah. I'm seeing in the chat, uh, Danny is saying like, is, is, uh, you're [00:52:00] trying to build this, you know, with Zapier. Yeah. Uh, this is, this is a very big use case for us. Yeah.

[00:52:06] Ben: Yeah. Luckily I like Rowan and we're not enemies, so it's all fine.

[00:52:09] Ben: You can share that. Um, and I did actually see that, that message that he sent the Twitter. Um, yeah. Is there, obviously that's one piece is like looping in or sending out an email to someone who's. inquired with you. I think a lot of use cases or a lot of the scenarios that people want to use an agent for is, Hey, here's a list of like, and we had this last week with Chris who built a system in Google sheets to basically put it in the company URL.

[00:52:36] Ben: And then you type in, I'm looking for creative directors and senior developers and whatever, whatever the sort of job titles, it then goes and finds, which I guess is the piece that you're talking about, maybe scraping and looking up where people are. Um, is there, And I think you mentioned there's ways to do it with APIs at the moment.

[00:52:57] Ben: Um, would love to see a bit on how that is possible now, if it is to say like, okay, I want Coca Cola and Paramount are my target customers. I'm looking for partnership managers or whoever on there. Um, go find them, pull in some details about who the person is, their email. And then I guess that would set off another Lindy or the same workflow.

[00:53:19] Ben: It's just a deeper piece to send out an email.

[00:53:22] Flo: Yeah, so we do have a prospecting API, uh, in Lindy. It's called, uh, PeopleDataLabs. Um, PeopleData, um, right here. And so you can find, you can find people here. And so you can do search for people. And so here it, it, this actually requires a SQL class, but you don't have to write it because Lindy's, Lindy's going to write it for you.

[00:53:42] Flo: And so you can just prompt it and be like, I'm looking for partnership managers in Northern, in North America, working for companies with more than a thousand people. And, and it'll, it'll just work like that. It works. And also, what I meant by web scraping is really like bulk web scraping. Like if you tell her like, hey, go to Indeed and get me all the job postings, [00:54:00] she's not going to be able to do that.

[00:54:01] Flo: But we do have um, a web browser action right here. And we've got a search the internet action here. And so she totally can. You can be like, Hey, here, like search for the name of the person that just arrived, that just submitted the form, um, followed by the name of the company. Um, and then this is prompting.

[00:54:24] Flo: So it's like examples are always better. Also EG, John Doe, Acme, right? And then search engine, you can be like, well, Google, Google News, Google Maps, whatever. So you can be like Google.

[00:54:38] Ben: Yeah. That sounds like a way to do it. Yeah. So I think Chris mentioned, the way he built it was custom scripts in Apps Script, um and used RapidAPI to pull in some of the API stuff there. Do you have anything that's like a developer node that is writing code or anything like that, that people can use as well?

[00:54:58] Ben: We don't have a code node yet,

[00:54:59] Flo: but we do have an HTTP fetch node, so if you want to hit up any of your APIs, you can just do that. Yeah.

[00:55:07] Ben: Okay, that's awesome. Yeah, I mean that same flow because we lots of people after that call were like I want to see that built like how do I do that? Um, I think tammy who's Doing tutorials for us is on the call and it's probably looking at this thinking.

[00:55:22] Ben: I think I could build this with with lindy So, um, yeah, we might have a look at that. I think that would be good. Um, Awesome. Yeah any any final questions for anyone in the audience?

[00:55:36] Flo: Right. Um, yeah, you can sign up if you go to lindy. ai slash ben's bytes. Uh, you can sign up and we also have a slack community Um, i'm happy to send you the link after that ben so you can send it to everyone. Um, And yeah, so, you know, you can you can ask us any questions We just keep our ears super close to the ground and we provide really good service

[00:55:57] Guest: Awesome.

[00:55:57] Flo: How many of your community

[00:55:58] Ben: managers

[00:55:59] Flo: are lindy's? [00:56:00] I shall not answer this question. We actually provide most of the support ourselves, but we do have an Indie helping us. So we have an Indie. If you post a message in the community channel, we have an Indie that intercepts it, and drafts a reply and sends it in the other Slack.

[00:56:14] Flo: And it's like, Hey, someone's needing help. This is what I think we should answer. Yeah.

[00:56:18] Ben: Awesome. Love that. Well, yeah. Thank you so much for, um, for your time here and showing us all of this. I think we're going to try and go off and create a bunch of content on this as well, because there was so much you can do.

[00:56:30] Ben: Um, and all the tools I've seen with building agents, it seems like a very, a lot simpler one, a lot easier to understand. Um, and yeah, we'll get the recording and everything else. Summary and transcript and maybe we should change our flow to be a Lindy to do all of that. Um, But i'll get that up hopefully this evening and then uh, yeah.

[00:56:51] Ben: Thanks again so much flow. Absolutely. Thanks everyone. Cheers Bye everyone

‍