https://www.listennotes.com/e/28bbe5abbe27461bbb717e3934d44421

People don't realize they think AI agents are just like pipe dreams, they think that's going to happen at some point in the future.

I'm like, no, no, it's happening right now.

I've been saying that, but here it is. People have success with this. We've got one in particular, it's starting to generate like $10,000 a month with his LND agency. It doesn't take a lot. Just go to lnd.ai/templates. We have like hundreds of templates here. I do believe AI agencies are the.

Future because most people, they either don't know how to set up these AI agents or even if they know, they don't have the time to. If someone were to launch an agency, what industry Would you attack first? Wow. Okay. My brain is buzzing. I just finished an interview with Flo. He's the owner of Lindy. Lindy is basically the AI first Zapier of the future. And I'm a user and a fan, and I wanted to find out how can we use Lindy to make money or to save money or to save time? Very granularly, actionably, he shared a screen. My mind is blown. That's period. I'm not hyping this up. AI agents are here. I thought they were just here in theory, and it was just a cool talking point, and it was an easy way to raise a bunch of money for something that wasn't ready yet, but they're here. Agents are here. And this was incredible. So you're going to love it, share with a friend and enjoy. Well, why don't you just give us.

A high level overview of what Lindy is and what problem it solves? Yeah, we are a no code platform that lets you build AI agents. What it solves is that, well, AI agents work for you. They're like AI employees. So it's just the same sort of problem that like a human employee would solve for you, except it's way cheaper, way faster, It works 24/7. It doesn't go on strike. I know I'm French, but we haven't implemented the strike feature yet. People use AI agents to, I mean, save time, save money, make money. You automate your customer support, you automate your lead generation, your lead outreach, and so forth.

Where would you draw the line between an AI automation and an AI agent? How do you differentiate between them personally?

An AI agent is power-wheeled throughout by an LLM, which is at every point aware, you know, the Camara Harris quotes, like, you know, everything that comes before you and the context in which you exist and what that. Yeah, that AI agent versus AI automation. It's like they understand everything that just happened to have this memory. And it's both short term and sometimes long term memories. They understand their history and they learn over time. So, for example, an AI agent can engage in a conversation. with someone, right? Because it understands everything that was just said. And so if you add a message to the conversation, it knows what it means because it's seen the rest of the conversation. And AI automation is more like each step is an island. And if you want some context to bleed over from one step to the next, you need to build that in. Not all the context automatically bleeds over. So like an AI automation is really useful for like really simple like ABC, like once and done things, like, hey, A new lead comes in, you research him, you qualify him, you send him an email, like that's an AI automation. An AI agent is like for more advanced use cases, any use case where you're going to have a conversation with human, you require an AI agent. And any use case where you're less than 100% sure of what should happen next at any given point should have an AI agent because the AI agent can decide what to do next.

So like an agent needs to be more open-ended.

Right?

More variables.

It can be more open ended. Exactly. Yes. Okay.

Maybe this is a good example. I'm running Facebook ads right now for a pickleball club that I'm launching. It's just like a side project. And I have a Zapier automation set up to take my Facebook leads and put them in a Google sheet. Super dead simple. But then people also fill out a type form where I get their phone number and I'd like to call them and get feedback on what they think about this club. an agent might take it a step further, give them a call using 11 Labs or Synflow or any of those other things that you probably integrate with and say, Hey, how much would you pay for this? What's your shirt size? Because founding members get a t-shirt. It would go 10 extra steps for me. Oh, that's accurate.

100%. And phone calls is an excellent example. It's one of the top use cases for AI agents and for us, we do support phone calls. I know you like to keep it tactical, so I'm happy to do a demo later on of a phone call using Lindy. But it's impressive because it's kind of surreal to like, it hits different once you hear the voice of the agent and you're able to talk to it on the phone.

What would you say like 80% of your users are using Lendi for? What type of tasks?

It's like three big categories of use cases. Number one is sales. So it's like lead generation, lead enrichment, lead qualification, lead outreach. Like you can have an AI agent like personalize its lead outreach at scale, like really fast, really affordably. Number two is customer support. And because we have thousands of integrations, and so you can get your AI agents to basically entirely automate your customer support over any channel, like SMS, email, in-docs, Zendesk, Freshdesk, WhatsApp, phone, you name it, it can do it. And then the third really big one is, and it's sort of a universal one, really, is personal assistance. So anything email related or meeting related, so I have Lindy Schedule my meetings for me by email. I have Lindy prepare me for my meetings. I have a phone Lindy that I call. It's like, yo, what's on my calendar today? Can you help me clean it up? I have Lindy record my meetings and take notes and email, you know, like Lindy drafts my email triage, she uses my inbox. She sends me a text messages or.

Gives me a call if I have like a really time sensitive email that they need to take attention to, that kind of stuff. So you will get a call from AI saying, Flo, dude, check this email. The pipe just exploded in your rental house. We've got to get someone over there.

Yeah, hopefully that doesn't happen too often. But yes, like I would say once or twice a week, I receive a phone call from a Lindy that's like, Flo, you really want to take a look at this right now. And, you know, about once a day or so, I receive a text message from Lindy that's like, Flo, you received an email from Bob asking about this urgent thing. And I can actually email message back.

And forth with Lindy. I can be like, what's up?

I'm in the meeting. Like, tell him I'll be right back. And she just like sends the email. Wow.

So she'll move that message over to Bob.

Well, it's cooler than just moving it over though, because it's not just like a copy paste of like, I have to like type my email. I can literally just be like, tell him I'll be right back.

And she's going to draft an email saying that. Wow. Okay. I thought I was using Lindy a.

Lot, but I haven't even scratched the surface. I haven't scratched it either.

I made the dumb thing and it's like every day it surprises me in what it can do. Yeah. I think it's interesting that you said sales first. because like everything I've noticed about business and all these SaaS products is like, it needs to be able to tie to generating revenue. Like that is the easiest, lowest hanging fruit use case to anything. Like if I know that this X dollars per month that I spend on this tool can return Y dollars per month, then I'm gonna be your first customer. Is that kind of what you're seeing?

Yeah, yeah. I mean, that's obviously like the fastest way to ramp up your revenue is to offer other people to ramp up their revenue. And in a way, it's a no brainer all the way because I don't know about you. The first time I heard about the job of an SDR, I was like, why is it a human doing this? This is insane that we need a person doing this. So yeah, all of these jobs are being augmented at the very least and automated actually quite rapidly.

Now, what are the 20% use cases that are really interesting but less common for Lenny?

Well, seeing quite a bit of marketing as well, like people automating content creation, using Lindy. I need to coin a term for this. It's like basically reports, like AI generated reports. Because AI is really good, obviously, at ingesting a ton of content. Like you could upload like a 200 pages long PDF to ChatGPT and ask for a summary and it's basically reading the 200 pages book in like two seconds or something. And so some of my favorite agents, we have an agent that sits at what I call the boundary between the company and the outside of the company. So every customer-facing interaction we have, that agency sees. So every ticket, every email, every meeting that we have with the customer or with a prospect, like that agency sees. And then at the end of the day, it sends a message, and it actually sends it to the entire company. It sends it on Slack. So in the general channel on Slack, it sends a message with like, hey, this is what's happening today. That's the boundary of the company. and then you can engage in like an interactive Q&A like back and forth with it on Slack. You can be like, oh wait, there's this like bullet point in here where you're like, Bob was really unhappy with XYZ. Tell me more about this and I'll reply to you. I love that and we're seeing more and more use cases of that. Like I think people underestimate the extent to which companies are basically black boxes. Like you don't really know everything that's going on in your business.

Agents are really turning on the light for a lot of businesses. Yeah. I mean, I could even see it, like let's say I'm the CFO for a publicly traded company and I see that Slack update, the summary come in what's happening in the boundaries of the company. And it's like, Hey, Bob, the customer Bob had a complaint about X, Y, or Z product. And I see that as an executive and I say, why is that even launched? We launched that product? Like, they're probably seeing major things that's like, holy crap, 30 people in the company knew about this, but no one thought to tell me until this automation.

Another thing that I 100% sometimes it flags oh, customer X was really frustrated with, oh, sales rep during the call. And I'm like, oh, and is that, can you send me the meeting recording, please? And so I get right here. It changes the whole company because again, it's like all of your walls are transparent. Everything sees, everyone sees everything. And as a result, like, learnings spread a lot faster as well, right? You can be like, ah, look, like this customer got really frustrated at the moment we said this. So let's not say this anymore.

You see that, it's like customer got very upset at our employee, John. Well, I want to listen to this. So you listen to this and you learn that John just said what our pricing was in a very nice way, and they were just an irrational customer. Can you type back to it and say, Hey, I don't really want to see stuff like this anymore because that's just an irrational customer. Can you like train it as you go?

100%. And so your question earlier about like AI agents versus AI automations, agents learn continuously. So you could literally just tell it's that in natural language, like, Hey, don't tell me about this moving forward. and it's going to remember it. It's going to keep it in mind moving forward. I do that with, I have an email triage agent that just triages my inbox every day. And I just have three buckets. I have urgent, I have FYI, and I have archive. And when I see an email make it in the wrong section, I just forward it to my email triage agent. My email triage agent has its own email address. And so I just forward these emails. I'm like, Hey, this email does not belong in urgent. This kind of email belongs in FYI. And I can just give arbitrary rules. It could be like, hey, emails about this topic or emails about recruiting or.

Whatever, just do this moving forward. Now, are you still at the mercy of whatever LLM you're using on the back end? Because like ChatGPT, it's like, stop using M-dashes. Seriously, stop using emojis.

Seriously. And it doesn't, right? So are you still at the mercy of that? You are.

At the end of the day, it is powered by an LLM.

I will say, though, are you using GPT-4.0?

4.0 and 3.0, it depends.

I always recommend people not to use 4.0. It's a really bad model, frankly. It's fine for day-to-day conversation, like to power workflows, it's not good. We default to Cloud Full Sonnet, which is excellent. And we also offer Cloud 3.5 Sonnet, which is one of my favorite models. So yes, at the end of the.

Day, it is probably better to buy an LLM, but LLMs are getting really good. Yeah, I said 3.0, I'm at 03. And I'm using Claude Sonic 4.0 for writing a lot, and it is really, really good.

It's so good.

With a great context window. Speaking of context windows, is Lindy also at the mercy of the LLM's context window?

Technically, how does that work? Yes, every agent is just an LLM at the bottom, right? So yes, at the end of the day, we are subject to the same context window limits the LLM offers. I will say though, like modern context windows are huge. They go up to like one or 2 million tokens. That's like, I think that's the equivalent of like a thousand pages book. It's very rare that we hit context window limits anymore. And even when we hit them, what we do is that we just compress the context window. We jam as much... Make it more efficient. That's right. And it works really well. And basically, context windows are for all.

Intents and purposes, they're basically getting infinite. It's been a few months since I've looked for this solution, but I just want something to read all of my emails. Through the past and to be able to just learn so I could train, I could pass it on to an employee for customer support requests. Is there a clean way of doing that today or are there too many privacy concerns that prevent that?

There is a way to sort of hack it together. For example, so like the way you would hack it on Lindy is you would export your inbox in Gmail. There's like an option to export every email you've sent or received. and probably sent is the most important one. And then you would upload that to Lindy as a knowledge base. And then you would ask Lindy to use that information to draft your emails or respond moving forward for you or whatnot.

Okay. But like me, I've got a holding company and we have eight different businesses and they all, a lot of stuff just funnels to my main inbox. How would you be able to sort that out? Like if this business then respond this way, you know, whereas if I was like the customer support lead for one business, It'd be very straightforward. It's like, here's how I answer all.

The random questions, you know? Yeah. So you can set a branching in agents, including in Mindy. So you could just have a condition, like a branching node, is what we call it. And then you can be like, hey, if this business go here, if that business go there, and then each of.

These branches would lead to a different knowledge base. Can you use Lindy to shrink the context window down? Cause that's something that I might actually.

Butt up against the context window. How would that work? Well, you wouldn't. bad up against the context window for this use case, because the point of the knowledge base is that it's got a huge corpus of knowledge, sometimes like trillions of tokens. But then it's using RAG, like a retrieval augmented generation. So it's basically searching, it's like the equivalent of a Google search for this huge corpus of knowledge. So then it actually context, then injects into the LLM and the agents context window is actually quite limited. It's just a couple thousand tokens. So you probably wouldn't run into the limit. That said, if you do run into the limit, yes, we also have a step that you can inject in your agent that's like, when you reach this point, compress the context window, and you can pass it a prompt that's like, compress the context window, and before you do that, extract X, Y, and Z information.

Gotcha. Okay, cool. Has there ever been an automation that it was just kind of like, this is so crazy that it worked. Like, I can't believe this actually worked. And what was that moment or automation like?

Yes, I would say there's two categories of automations where we're like, I can't believe this works. There's one where we're like, I can't believe this works just because the user set it up in such a crazy way.

An inefficient way?

Yeah, you could call it that. It's very often they just have one giant AI agent step with a huge prompt and 50 actions in that prompt. It's like, oh my God, and it works. And the prompt doesn't even really make sense. When I read it as a human, I can barely understand what the person is saying. And somehow the agent is figuring it out. I'm like, how does this work? Again, I think that's the strength of the AI agents is that they are really forgiving. Like automations and coding in general, like you just get one little thing wrong and the whole thing collapses and stops working. AI agents, they leap along. Like you can remove characters, you can remove entire lines. Like the AI agent is always going to try to figure it out. And then there's the second category where we're like, we can't believe it works just because the use case is so complex. It's not as much like the user screwed up, like the user did a good job. It's just like, oh my God, this is an insane use case. I think some of my favorite AI agents are sales and SDR agents. So we have this one that engages with a lead on a web chat interface and it collects information, it qualifies the lead. And then if the lead wants a meeting, the agent books the meeting, so it looks at the calendars of the company, it books the meeting, it sends a confirmation email, it sends a confirmation text. It calls the person on the phone before the meeting to confirm some information. After the meeting, so it records the meeting, it joins the meeting as well as the sales rep takes notes, and after the meeting, it sends a follow up to be like, Hey, thank you for your time. This is what we discussed. And the whole time, it keeps in context all the previous interactions that it had over all these modalities. So like the web chats, the SMS, the phone call, all that stuff. So like if it calls you on the phone and it's emailed back and forth with you before, It will mention that on the phone. It'll be like, Hey, Bob, we've just emailed, we talked about this. How's your day going? So seeing exactly like, holy, that's when I was like, it's getting real.

It's happening. Oh, man. In that case, is it telling the user, I'm AI?

And you can't miss it. It's not at a point yet where you would be confused by it, especially when you have it on the phone, you know it's an AI. The way that users set it up is like, yes, it is obvious that it's an AI. Also because on the web chat, it just replies too fast and all of that stuff. It's actually funny, we have some customers pretty frequently, actually, they ask us, can.

We make the agent slower?

It's like too fast. Yeah, yeah, yeah, yeah.

But yeah, in this case you can tell it's an AI. I mean, I've been setting up like email automations for a while where like, if someone emails me this with these keywords, respond with that. And very few platforms allow me to put a pause in there. Kind of like Zapier, you can put a pause in between something. And I think that's a big part of it because if I send someone a critical email and I get back four paragraphs within 10 seconds, I'm going to cock my head as I look at that, you know? Yeah, okay. I'm going to discount this email. Whereas if it takes a half hour.

To get it, but that's all, I'm sure that's like super easy with Lindy. Yeah, 100%. Yeah. Just like ZPO, you can introduce like a waiting step.

Yeah.

Yeah. I'll give you an example of something that happened actually just a couple of weeks ago. I have Lindy schedule my meetings. So if someone sends me an email and is like, Flo, let's meet. and I do want to meet, I CC my Lindy assistant to my email. I'm like, oh, sounds good, Lindy will help us find times. But then she goes back and forth, she looks at my calendar, she agrees on the time with the other party, she sends the calendar invite. And then a couple of weeks ago, there is this person I was supposed to meet with on that day, actually. And last minute she sends an email in the thread and she's like, Flo, I'm sorry, I have to cancel today's meeting. My son just had an accident, I have to go to the hospital. And Lindy sees the email, she's on the email thread, and she's like, I'm so sorry to hear about your son, which by the way, like, again, that speaks to your point about the AI agent because she's seen the entire conversation. And automation here would have sort of worked, but it would have been dumb. It's like, oh, okay, branch, reschedule request. It's like, here also more times for you. So like here it's like, I'm so sorry to hear about your son. And she did not even offer times immediately. So like, take your time. I'll circle back to you later, you know? And then she waited and she woke up like, I think five days later or something.

And she's like, Hey, how's your son doing?

Is everything okay?

Is now a good time to put some time on the books again?

I know. And that's what I mean.

Like, did you set it up to do that?

No, and that's what I mean. But like, I built the platform, but I'm surprised every day by what it does.

I can't even talk right now. I'm just thinking, like, I can set up this app that does that, and it's like, my son just got hit by a car. I have to transfer or I have to push, and it's like, what date works for you? That just looks so bad. That just, I cringe even talking about the hypothetical.

And this happens every day. Just the other day, someone else added his human assistant to the thread to schedule time. And the email assistant was like, haha, you're adding your AI assistant, am I getting replaced anytime soon? And Lindy replied, she was like, haha, I don't think I'm replacing anyone anytime soon, smiley face.

I was like, that's smooth and cool, you know? Right, right. with like the SDR, the sales example, are there companies like, it just seems like we're still so early that a lot of the sales stuff is like a party trick. Like there's a dozen companies in Silicon Valley that have raised tens of millions of dollars for sales agents, right? That's it. Do you have customers that are like wholly relying on sales agents to actually bring money in the door? Like end to end?

Yeah, absolutely. Well, not end to end. Like you still need like a human to close them, that's for sure.

For now.

For now, yeah. I think it'll take a while to replace that. Be it only because humans don't build relationships. I think sales is truly about relationships. Humans have like an anti AI bias. Humans don't build relationships with the AIs. They build relationships with humans. So I think like that reason alone, I think humans are going to be in the sales loop for a very long time. Unless we reach a point where the AI is indistinguishable from a human, even on a Zoom call. And we reach a point where people are comfortable misleading their customers and not telling them it's an AI. But I think we're pretty far away from that. And I think essentially it would be messed up to do this. But yeah, we do have some customers who basically, like all they have is AI SDRs.

Like they don't have human SDRs anymore.

Yeah.

The example that you gave a few minutes ago of the SDR agent that does everything and it follows up is the call that it's booking. That's still with an actual human then, right? But the AI agent will follow up, will remind them, it will do everything else.

That's right. Including sending a prep email to the human rep. They'll be like, Hey, you have a meeting with Bob. This is what we discussed previously.

This is what you need to know. Do you guys have any integrations with ringless voicemail services? Which one that is? No, like the ringless voicemail drops where there's a cool, that makes sense. Like, Sly Broadcast, does it? Cause like if I'm an SDR agent or if I'm a sales manager, I would want to clone my sales agent's voice with 11 Labs and then send a summary of the call, but I don't want it to actually answer the human. And so I would use something like Sly Broadcast to just drop a voicemail with the sales agent's voice so they think it's him and just say, Hey, thanks for the call. Just want to follow up. These are my action items. These are yours. Hope you have a great weekend. And they're just going to assume they missed the call. Oh, voicemail. Ringless voicemail drops.

Ringless voicemail. I'll look it up. That should be something pretty straightforward we'll to integrate with. Yeah.

So you obviously need a person to like collect the wire info and hop on the sales calls and warm them up and have the relationship. What types of industries do you think that using sales agents, it works the best in today? Because like, if I'm selling like copper pipe fittings B2B, Probably not great, but if I'm selling like enterprise SaaS could be great. What are you seeing works really well for sales specifically? What industries? Or price points?

Yeah, I had the same prioritized use, like pipe fitting, another fit, enterprise SaaS fit. And I actually think it's the opposite. I think pipe fitting is a great fit. And the reason for that is because the people you sell pipe fitting to are not really tech savvy. Like they don't want to jump on a Zoom call. They're sort of like on the gold all day. So like, If your prospect is on the go all day, you should probably use one of these voice agents. If your prospect is not tech savvy, you should probably use a voice agent. We actually even have people deploying voice agents for their own reps who are field reps. They're on the field all day, they're like in their car, and they've deployed AI agents to help these reps update their CRM on the go just by making a phone call. So anytime your prospect is on the go, Or anytime if you're very busy and you can't pick up the phone. So we actually have like some blue collar workers, like plumbers, we have some restaurants as well. Because restaurants, they're busy at peak hours, you know, you see them, they run around and stuff. And that's also the time when they receive the most phone calls. So we actually have some restaurants using us as like an AI photo receptionist because it's always the same questions like, are you open tonight? Are you solving X, Y, Z? And the photo receptionist can also make reservations and so forth. And then like the Plum Bill use case is different. It's more like this guy is plumbing. It's not like picking up the phone and every phone call he's missing is business is missing. So I would say those are like the big use cases where like AI voice agents are really effective.

Yeah, I think there are some use cases where you almost do more harm than good disclosing that it's AI. Like if I'm calling a restaurant to find out, do you have any tables right now? Can I make a reservation a week from now? Are you open? Can I get a gluten-free pizza? As a customer, I'm cool with that being very transactional. I would rather talk to a robot, whether I know it's one or not, than sit on hold or leave a voicemail and I know I'm never gonna get called back. Whereas in the SDR example, something higher ticket or more personalized, 'cause I'm not building a relationship with a hostess at a restaurant, I probably wanna say as I answer the phone, like, Hey, I'm.

An AI agent, but I'm actually a lot more capable. How can I help you? Yeah. And some people, at least at first, they dislike talking to an AI agent. And I think a lot of that is because they've been conditioned by the phone trees, right? It's like type one to continue type two. And that experience is just so awful that people have just learned, I want to talk to a human. Little by little they learned that this is a fundamentally different breed of product from a phone tree of your parents were using. And then they actually also end up appreciating it because they appreciate that, like, yeah, you'll never put on hold, you can talk to me at any time, and you can code it as much as you want. You're not afraid of disturbing it. Like, I have my Lindy. I have a human assistant as well, but I call my Lindy on the phone all the time. I actually call it more often than I would my human assistant. But even though the human assistant works for me, I wouldn't feel comfortable calling him 20 times a day at 11:00.

PM to ask for random stuff. Yeah, I mean, as a consumer, I am more than ready for companies like Verizon or these mega companies to just completely get rid of their call centers and replace it with AI voice agents. We just switched to Verizon, our whole family plan, and my son's phone, for whatever reason, the number would not port over. And we called support four times. We went into the store and they ended up putting me on support while I'm in the store and it never got fixed. We ended up having to get a new number. I was just thinking the whole time, like, I am so ready to talk to a robot here, 'cause I wouldn't have to wait. It knows all the long, 'cause I know our problem was like a very long tail obscure issue. And it would know that, whereas the rep doesn't, they have like four issues that they work through. And so I kept having to be put on hold and transferred, and it's the, it can't get any worse.

Like this can only be an upgrade, you know? So you can actually also use AI agents for this. I use Lindy for this all the time. Like I had also an issue with my internet service provider. a couple of weeks ago, and I just sent my agent to do it. She spent half an hour on the phone with these people. It was Google. I hate them. Half an hour on the phone with them. And in the end, she resolved the issue. And that's half an hour that I got back in my weekend. Wow.

Oh, man. It's like the robots talking to people who act like robots, but the real robots are actually smart.

Yes. Yes.

I would love for you to show some interesting use cases. on your screen that could apply to like a large amount of people.

Yeah, absolutely. Let me share my screen here.

And I'd love to hear about any agencies that are building agencies completely on top of Lindi. Maybe it's a lead gen agency or AI automation agency, et cetera.

Yeah, I'd love to talk about this as well for sure. Let me share my screen right there. This is my chief of staff in Lindi. It's sort of like the main Lindi I use all the time. And so you can see like just on Sunday, I called, I called her four times. I'm going to call her right now. My Lindy number. I'm going to put her on speaker. Hi, Flo, how can I help? Hey, Lindy, what's on my calendar today? You have an interview with entrepreneurship and.

Opportunities, two meetings with the marketing team.

And three interviews with candidates. Yeah, anything else? Hi, Flo, how can I help?

Oh, that's all I think. I think it just keeps saying that.

Better right here. No, oh, it just keeps repeating, hi, Flo, how can I help? No, that's only when you're going to call her today.

Jeez Louise, I don't know why I'm not doing that. I'm using the ChatGPT voice feature a lot, like multiple times a day. It doesn't even hold a candle to this because it doesn't know what's on my calendar.

Yeah. I use it for that all day. I use it for, like, this is the thing I mentioned. Well, like anytime I receive a time-sensitive email, it sends me a text message. And I'll show you what it looks like in the backend. It's actually really, really simple. Flow editor, it's like if I receive an email, you look at my calendar first because you need to know what's on my calendar in order to determine if it's a time sensitive email. And then there is this condition. And this is what I mean by AI agents can see everything that just happened before. Like here, I can write a prompt for my condition. And the prompt is just like, if this is a time sensitive email from someone asking to reschedule a meeting, I have today with them. So again, I don't need to inject the context. I don't need to do any of that stuff. It's like an AI agent that received the email, that looked at my calendar, and now it knows what I'm talking about. And if this is a time sensitive email, you send me a text message. So it's like you received an email message from name about very, very short summary. And here I can just add this kind of, and here I'm just prompting. And you can see I'm prompting in this case, CloudFlare Sonar, which again is.

One of my favorite models ever.

Just awesome. What do you love about it? The vibes all immaculate. That's the power of the youth. It's in the models there. You know, it's smart. cheap, it's like it's affordable, it's fast. I really like its writing style. Like I really usually cringe when I see LLM writing styles. It's like you can tell it's an AI model. Here, like you can barely tell anymore. Yeah. Okay. That's amazing. It actually also sends a message to my human assistant. Like sends the same text again. And again, this is the difference between an agent and an automation. Like here, I can be like, sends the same text again. Because it knows the text message it just sent. because of that memory. But this time, say flow received, not you received. I use Lindy all the time for like, well, meeting recording. I use Lindy to manage my personal CRM. Every Friday, for example, she wakes up, she looks at my calendar, she looks at my CRM, and then she recommends people to add to my CRM. She's like, you met with these new people that you don't have in your CRM. I checked. You don't have them in your CRM yet. Do you want to add them to your CRM? And I can just be like, yep, add A, B, and C and tag them like C is an investor, B is a sales leader. And then every so often I go to Lindy and I'm like, hey, who all, for example, like we just made like a big announcement a couple of weeks ago. And I was like, hey, who are the AI influence fields that I know? And it just goes in my CRM, it's like, hey, these are the AI influence fields. Yeah, yeah. I use it for my daily digest. So it's basically, I'll just show you the flow detail for this one. Every morning it wakes up, it looks at my calendar, and then for every meeting and for every attendee of every meeting, it looks for their LinkedIn, It looks for my email history with this person. It looks for my meeting notes that are brought together by a Lindy. And then it brings together like one nice email with like, this is people you're meeting with today. This is the last thing you discussed during your last meeting. This is the things you discussed over email with them and so forth. I'll give one last example and then I'll shut up because otherwise I never stop.

No, this is good. I want to learn about what you're doing with deep research as well.

Yeah, I have this deep research solution here.

Yeah.

Well, why don't we start with that actually? So deep research. I use deep research for, I really like learning about companies and about founders, and I find that there's actually a lot of knowledge in podcasts. I actually think that's a great way to learn.

And they're not very searchable.

It is not very searchable, Angel Agents. Yeah. And so I'll show you an example of how it works here. So here, for example, I wanted to learn about like Mark Binyav, or like Zeb Evans. Zeb Evans is the founder of ClickApp. And so I'm like, I just give it a name, like Zeb Evans. And here it is searching on Google interviews with Zeb Evans, and it's finding 10 results. And then it's going through all these results, and simultaneously we call this an agent's wall. It's really good. So simultaneously it basically spins up one copy of itself. Okay. Okay. And it transcribes the video, and then.

It summarizes the video.

And then all these summaries.

Doing one video at a time? Well, no, it does all of it.

At the same time. Okay. And then once the summaries are done, it exits the loop and then it gives me the summary of summaries. Oh man. So it's like, this is everything that Zeb Evans mentioned. And I can go and ask follow up questions. I'm like, ah, tell me more about XY.

It's an agent, so I can just chat with it. Yeah. Dude, you know what would be cool is that same thing, except let's say Zeb Evans has 10 interviews. 10 period total. And in each interview, there's like 60 to 80% of overlap. He's kind of saying the same thing. I would love an agent that shows me all the non overlap. Oh, that's what it does. That's what it does. Like the childhood stories.

Yeah, yeah, yeah. So you can see here, for example, like key themes across interviews.

Uh-huh.

Right. So like this is an agent that has seen the summaries of all the YouTube videos and it's giving me its own analysis.

Oh my gosh.

I can ask actually is like here like what are some of the things that he says across the most interviews?

I mean for my podcast prep this would be amazing. I could have done this with you.

You could have, you should have. Looking across the interviews here are the themes that he consistently repeats. One percent better every day, near death experiences, six total. Oh yeah. Yeah, he like I get like quotes from the interviews and so forth. recruiting. I'll show you this one because that's one of my favorite. So I spend a lot of my time recruiting. Like I always joke that a founder is basically just like a glorified recruiter. We all right now hiring, by the way, we all hiring, hit me up. Okay, you heard it. We are hiring across every function, but one of the ones that we're hiring for is sales leaders. So here I have this LinkedIn recruiter. And if you look at the timestamps, look, it's like full 13 p.m. on Tuesday, right? And I'm like, Hey, find me directors of SMB sales from one of these companies. And I give it a list of companies. And she's like, all right, let me look. She's searching and she's like, I found one candidate. And then she says, let me try another search. And again, it's an AI agent. I didn't need to ask it to do this. It's smart enough to know, he's probably not going for one candidate. So let me search for more. It's like, all right, I found these guys here. And I'm like, yeah, keep searching. So it keeps searching and then it finds these people. It's like, okay, tell me which ones of those you'll need to reach out to. And I'm like, yeah, all of them look good. And it just reaches out. And so here again, it's kicking off an agent's womb, where it's like, for each person, I'll just click in the subtasks here. For each person, first it searches my inbox to check if I've already reached out to them. Because sometimes I have. So it's like, okay, you've not reached out to this person before. Since you haven't reached out, let me draft an email for them. And boom, hey, this is Flo, you know, we're looking for a head of sales. Would you be interested in a new conversation? And all of this, I wonder if there's another time stamp I can look at. Well, yeah, I can. Because this was kicked off at 4:13 and this, yeah, at 4:18, the outreach started. So in five minutes, it found the people I was looking for and it sent the email. And then you can see it sent follow-ups. It woke up again two days later and then again two days later and just kept sending follow-up emails.

Oh my gosh. I can't even think of my next question because I just want to log off and start building these.

I don't even think it's real. Like people don't realize, like they think AI agents or this like pipe dreams, they think that's going to happen at some point in the future.

I'm like, no, no, it's happening right now. I've been saying that, like I tweeted a couple months ago, like everyone's talking about AI agents who is actually using them today in their business and how. And like, I just heard a bunch of like third hand stories, but here it is. How can the general public, like I know you have like a repository of Do you call them flows?

I would call them Lindy's, but we're gonna rename it to agents.

How can people just copy and paste.

This with their own use case? Yeah, you just go to lindy.ai/templates. And we have like hundreds of templates here. I think we have one for the recruiter, actually, as just showed. Lead generator.

We should have one for recruiter. I think it's just missing from here. Okay. I would like to build one where I plug in like 20 YouTube channels of, you know, other channels in my niche, and it sorts them from most to least viewed videos. It finds the person that was interviewed and then it finds them on LinkedIn and starts messaging them, Hey, you were on this podcast. I have a very similar podcast.

You want to come on mine? Like that's pretty simple, right? Yeah. I mean, there's like a Scrape YouTube via Lindy chat here.

Yeah, yeah, totally. Oh man. What about building agencies on top of Lindy?

What are you seeing there? We are seeing, we're starting to see people have success with this. Like we've got one in particular, I forgot the exact number that he told me, but I think it's starting to generate like $10,000 a month with his Lindy agency. so what it consists of is like he is posting about us on LinkedIn and he's making videos about us on his YouTube channel and he's like, Hey, check out this like cool Lindy. He's basically doing what I just did, except he's selling his services, right? So he's like, Hey, like check out this cool Lindy. This is how I've set it up. This is how it works. And it's a cool demo too. Like these demos are just compelling. And then he gives the template in the description and he's got a lot of people reaching out to him who are like, I don't want the template. I want you to set it up for me. and he also offers that service. And then he's, I think he charges like a one-time fee. I'm not familiar with his pricing, but I believe he's charging like a one-time setup fee and then like an ongoing maintenance fee of like 500 or a thousand dollars a month, which is nothing. Like if you think of like a pay-SDR, it's just like base price itself.

Like three leagues or something.

Yeah.

And then he's mostly finding customers just by putting out free content. That's right. That's right. Yeah. And most of them will go on to either do nothing or implement it.

Themselves, but a minority of them will say, Hey, just do this for me. That's right. And it doesn't take a lot. If you're charging like 500 bucks a month per customer, like 10 customers, you've got $5,000 a month. I do believe AI agencies are the future because most people, they either don't know how to set up these AI agents or even if they know, they don't have the time to. So AI agents address both of these issues.

If someone were to launch an agency, what industry would you attack first?

I would only say any industry that you're familiar with. any industry where you've got an audience and you understand the industry and the challenges it's facing, just focus on that because every industry can benefit from AI agents. Again, like AI agents are basically AI employees. Every industry's got employees, you know, so every industry could benefit from having AI employees instead because they're just better, cheaper, they're faster, they're more intelligent. So yeah, just any industry that you're familiar with.

What about using Lindi for lead generation, either for your own business or in selling leads as a business? Are you seeing any of that?

Yeah, absolutely. So look, you know, the recruiting use cases just showed that's kind of like.

A gen use case, like recruiting expensive leads too.

That's 100%. So, and it's really targeting, like it's really finding exactly what I asked for here, right? It was like, hey, like directors of sales at like London here, like Dropbox and so forth. So it's yes, we've got a bunch of lead sources integrated in Lindy. the best one is one that we called People Data Labs. It's just like a lead, it's almost like Mercury, but not Mercury, sorry, Apollo, basically better than Apollo in my experience. The lead or better is like less stale. Otherwise we also have like a ton of scrapers. We have like thousands and thousands of web scrapers. I'll give you an example. We have a customer, he has like a, it's funny, he's got like a music studio for video games. He makes like music tracks for video games. And so he set up a Lindi to scrape the Steam store for him. So it gets kind of niche, you know, it's really like a scraper The what? The Steam web store. Have you heard of Steam?

Oh, yeah, yeah, yeah, yeah. Okay, Steam.

Yeah, like the game marketplace. That's right, that's right, that's right, that's right. So, hey, Crepes team find independent studios like smaller than X, like less than X in sales, because they're too big, they've got their own in-house music person. And then for each of these people.

Look on the internet for their contact information and reach out to them.

Yeah.

Oh, man. Okay, this is maybe a difficult question, but if you were forced to nuke all of your partners, like all of your integration partners except three, which three would you keep? Like which are the most, just the most powerful? Doesn't need to be the most common, like Gmail or Sheets, just the most absolute powerful integration paired with Lindy.

Yeah, I think this one, like the people that that apps one is super powerful, like this ability to just find people, I find it is so awesome. I think it's like Scraper, like we integrate with the Apify, like kudos to their team. I think they've done an amazing job and like Apify just gives you thousands of scrapers. It's kind of a cheat answer to your question because it's kind of like, you know, wishing for more wishes because it just gives you like thousands of scrapers, but that's just such an awesome one.

I really like the YouTube one actually. Okay.

Yeah. And what about like voice agents? Would that be one? Oh yeah, I don't really think of.

It as like an integration. I think of it as a capability of the platform.

But, okay. Okay.

What part of all this is still too hard for Lindy, but that you think Lindy will be able to tackle?

It's a good question. I think of Lindy's today and AI agents, they're not really AI employees. Like that's more aspirational. I think of them more as like AI interns. So any simple enough tasks that you can define as a sequence of steps, right? Even if it's not always the same sequence, you can give a playbook to someone and ask them to do it. Like AI agents are really good for this. if there's like a lot of ambiguity or if it's really a problem where it's like, look, I don't really know, you've got to figure it out. Like when you hire VPs at a startup, like a VP of sales, a VP of marketing. I love this quote from Steve Jobs that's like, we don't hire smart people to tell them what to do. We hire them so they tell us what to do, you know? And so if you really need this person that's like, hey, I need you to help me figure out my marketing strategy and like drive it home, like this kind of long time horizon task AI agents are today poorly suited for.

But isn't changing rapidly?

Very rapidly. I think in a couple of years it's basically going to be a full blown AI employee.

Okay.

What feature requests are people asking for the most? It's almost embarrassing. We don't have team accounts yet. You can't really collaborate with your team on Lindy. That's coming soon. They're asking for computer use, the ability for Lindy to use the computer that's coming soon. They're asking for, it's a pain for people to create their own Lindys because you've got to do that. So just a sneak peek here, this is just my account because I'm a admin. This is not yet available in Lindy, but we all building this like, it's basically like a cursor for Lindy. It's like an agent co-pilot that creates agents for you.

Mm-, yeah. Man, okay. What did I miss? Is there anything I didn't ask you that I should have asked you? Like my mind is blown.

No, I think we've got a lot of good ground. It does make your head spin to think of AI agents like today and then where they're going to be in the future. I think people don't totally understand what's happening right now. Well, we know where to find you.

Well, we know where to find Lindy. Where can we find you?

Just email me at flow@lindy.ai. I'm also on Twitter at X. Do people say X now? Altimo, A-L-T-I-M-O-O. Okay.

And we know that if we email you, we're probably going to get Lindy.

Responding, but that's okay. You never know, but yes. Most likely Lindy responding. I'll read the most gust of millionaires.

Well, thank you for your time.

Yeah.

Thank you so much for having me. I would really appreciate it if you used my Lindy affiliate link in the show notes. I'll get a little kickback. You'll get a little kickback. And please share it with a friend. We'll see you next time on the Kerner office.