# Is Perplexity the easy mode for AI Agents?!

Channel: The Next New Thing
Video: https://www.youtube.com/watch?v=FNbpaB9MpMA
Duration: 20 min
Language: English
Words: 4278
Transcript page: https://viewrankai.com/tools/youtube-transcript/FNbpaB9MpMA

---

[0:00] What's your revenue? Levership does about 2.1 million a year. What does Levership do? As an education company for small and micro businesses. At your company, you're using AI a lot to grow. I want to take a look at how you build a business. Let's get right into it. Can I see how you use

[0:13] perplexity computer to get leads? We worked with one organization uh that needed to find fishing vessels and those are very very hard to find. So we And by the way, this is the way that you might get leads for your business too. You can use perplexity computer. We're using your client as an example. Okay. So they why do they need fishing vessels is well so they they sell medical supplies to people who can't quickly get to a hospital and when they went through all the different types of clients that they sold to they found out that fishing boats were the best buyers. Now before we asked them a single question they had

[0:45] no idea that fishing boats were their best and that's I think one of the important things here. The more specific you are with AI the better answers you're going to get out of it. So, as soon as we said we needed fishing vessels, we said, "Okay, let's let's go search for those." And now, you know, you look at the prompt. The prompt is not sophisticated. It's five, six lines, but it drives at where can I find those fishing vessels. So, if you look at all the steps that it went through, it went through six steps to make sure that it was on the right track. Then it went

[1:14] through 57 steps and it gave me a final report of 5,581 fishing vessels. It said, "Here's where they are. Here's where they exist regionally. Who owns five plus boats? Four, three, two, one. What's their email? What is their what is their company name? What is the overlap with me? Because it thought I was selling to them. Then where's your potential warm collision? And then where's a first draft of your first email? And then it gave me an assumptions tab and it said, look, if the assumptions are different,

[1:47] change them and rerun this. Okay. Can we see that that output? Because Perplexity computer is doing what for you? It's actually getting the whole the whole thing including how to contact them. Yes. How does Perplexity even have access to fishing vessels? Well, so it it it it digs in. So the thing the fun with fishing vessels here and you can see it out. It just kicked out into Google. You can see the

[2:09] prospect list. You can see the vest. Whoa, whoa, let me see the prospect list. It actually Let me see what it's gotten you. It's gotten you the names of them, the addresses, the city that they're in. It got you the name of the p the name of the vessel. Vessel count meaning how many vessels they have.

[2:23] Okay. Where are they? We can keep going. Owner type and then you can look and go get them. Exactly. Here. Go get them. What is VR? That's the database it's in. Okay. And then it actually will get you the names, the contact information of these people so you can sell to them. Correct. How how does Plexit even get that? Well, it anything that is public source, if you can point it, it will drive at it. We did the same thing for high- netw worth consultants that a financial planner wanted to get to. Did the same thing for uh for a a cyber security firm that wanted to help private

[2:59] investigators in Florida. Okay. Uh right. So like these are these are it you've got to get really specific on okay the client you want to get to. Okay. And the jurisdiction and what you want to do with Okay. All right. So, when you're when if I were trying to create a business like yours, I might put this type of prompt in and look for the types of customers that I want. And I told you based on my conversation with you that I might go in there and say, I'm looking for businesses that are trading time for money, consultants, coaches, things like that. And you said, Andrew, yeah, you

[3:35] really are not going to get great results like that. How would you rephrase it if you were sitting there looking for your own bis for your own prospects? So, like I'm I'm looking for growth-minded entrepreneurs who run timebased businesses um that want to scale. Those are the three prerequisites, but that's not deep enough. So, then you got to go what are the types of businesses that you normally run in that do that. So, there are construction businesses like plumbing companies, cladding companies, there are massage therapy companies, there are psych psychologists, etc. So,

[4:07] pick one of those. I see. Okay. and then pick. Now you can, what you want to do is work through this whole prompt down to the specific one. Once it works, now you can rinse and repeat that. Just fill in the blanks with something different. I see. You know, before my conversation with you, I didn't know what a cladding company was. These are the people who put the outer layers on business on buildings. And you say to yourself, you know what? they would be good clients for my company because if I could show them how to improve their systems, if I could show them how to get good

[4:35] workflows and then use AI, they are going to save time which then is going to help them grow their business or it's going to help them grow their business by getting better leads. That's that's what you would do. And so instead of saying, "Hey, Perplexity Computer, find me businesses that are trading time for money," it would be, "Find me cladding companies with this criteria." And then it would get you the names of those companies and even email addresses

[4:58] without a without an Apollo connection. Yes. Uh so we did this for the PI companies, but I want I want to say I want to back up just a little bit. So what I will do normally is I will go in and say here is what I need to find. I want to find more leads for leadership. We do well in professional service. So this is kind of what you said here, but I said, "Hey, physiootherapy clinics in Florida." But I'm not just prompting perplexity. I'm prompting model counsel

[5:21] within perplexity here. What's model counsel? I'd never heard of that before this conversation. So what model council does is it actually asks the same question of the three different lead models out there and it rocks down and it says okay you want to build a prompt to find that great here's where the models agree here's where they disagree here's the unique discoveries now here's the comprehensive analysis and it will say here's where the models converge here are the unique findings and then it gave me a version of the prompt now I was a little like look how big that prompt is there's no omen that is going

[5:55] to create that prompt. You're you're just you're not. So, what it's done is given me a first draft of the prompt and then it said, "Run that prompt six to 10 times, one metro county batch at a time." So, what it's done is it's told me Florida's too big. I see. You actually have to get it down to the county if you want reliable results. So, okay, great. So, then in here, you can see that it's told me all the stuff that I need to actually fill in. There'll be a square brackets in your prompt. The the other brackets are fine. Don't worry

[6:23] about them. Then I said, "Wait a minute. Are you sure?" So almost like I'm asking a a a an expert like, "Look, are you really sure?" "Yep, no problem." But then it gives me the complete workflow, okay, that I want to do. And then it goes, "Here's the improved prompt. There is a little bit better that you want to do." Now, I would take this prompt, copy the prompt up here, and put it straight into Perplexity. We could run it live here. I got to go through the square brackets,

[6:48] but you see how long that prompt is? Yeah. Like so if you want a great prompt for AI for prospecting, you need to get really specific and then you need to get AI to help you build the prompt. Okay. And then when you use that council, it was telling you the strengths and weaknesses of each of the three major models. And now you would pick the one in Perplexity Computer that

[7:09] you're asking. Yeah. You you pick whichever one. Well, when I go into Perplexity Computer, now I may choose or I may let Perplexity do it. Normally, I'll run both. I'll let Perplexity choose the model because part of the uh IP of Perplexity is the fact that they know based on your inquiry how to best use the models.

[7:28] Got it. Right. So, I will run one letting them do it, then I'll run another one where I if I have a hunch that you know like look uh Clo is uh I I love Anthropic and I love their moral disposition. Um, but they sometimes don't hunt as deeply as Chat GPT or as Gemini because of some of

[7:48] the guard rails they've put in. Okay. All right. So, first thing you're doing is you're getting leads and you just showed me a couple of different ways that you're getting leads. And this is the more intense version of it using perplexity computer to help you figure out where you should post it, to help you figure out how you should write it,

[8:04] and then even change your your approach. So, instead of going for a whole state, go for a small subset of it. All right. The next thing that you do is you say, "All right, I've got all these names of potential customers. I need to actually get them." And you told me you create a

[8:16] warm collision. What's a warm collision? Yeah. So, cold approach is It won't work. It's got maybe a 3% uh run rate of conversion. It's It's just not a great way to get in contact. Okay. You've got in my world, I want to have a trusted conversation about the health of your business to find out if we can help you with your business. So, I have to find a trusted source to introduce me or I have to find a way for them to trip

[8:42] over me. Okay? I can't come straight at it. Okay? So, that's what I'm trying to create is a warm collision. Okay? And lots of ways to do that. You you may share uh LinkedIn connections, first second connection, maybe you can get a warm introduction that way. Um you run open source intelligence, etc. And we did this on you, right? We had some fun with you. You were saying, listen, if imagine I don't know Andrew Warner, I want to ask him to be a customer or I want to ask him to be a guest on his podcast. Find me. Let me see the the the uh prompt. Actually,

[9:13] the prompt is like one line at the top. I want a warm collision with Andrew Warner on Mixer. Can you run OSENT open source intelligence and help me find a pathway to connect? Okay. And it says finding best warm entry is via Andrew's current focus AI builders and agents, not a generic guest pitch. And all three models agree on that evidence. Andrew actively frames himself around AI building now the next new thing AI agent content etc. And it's got the sources. Okay. So it's telling you that. What else? Let's keep scrolling. Tell me what stands out for

[9:44] you on this. It's going so it's going where the models agree. So great. Where the models disagree. I'm almost never looking at this and I go back. I'm like okay great. You've you've had your arguments. Like having three experts debate what they agree about and what they don't. But really what I want is the conclusion. So I get down to the comprehensive found findings and it goes look I think a strong a second strong consensus is around AI for sales and relationship workflows which is ironically exactly where we landed as we walked through the pre of this show right um and look we didn't set this up it

[10:16] wasn't listening to us areas of divergence etc and then it goes okay here's how I would get a hold and it and it picked up that we had done a podcast I think in 2018 or 2019 pre- pandemic And it said like look do you if you know this guy reach out but reach out in this way. And it's saying uh first quickly verify that this is the actual page. Then next build a tiny testable artifact maybe a one pager or a threeminute loom that riffs on Andrew's uh mutual intro AI agent point. Wow. Interesting. Then comment substantively on one of his

[10:49] LinkedIn posts. Uh I totally get that. That's been working a lot on me. Then DM email with a short ask. Want me to send the tear down? Tear it apart if you want. paste and then got it. All right. It's giving you the step by step of how based on what it knows about me and you, you should connect. That's your warm

[11:05] collision. Exactly. And and if you look at it like this is not look, it's looked at 194 sources to collate this. You know this the interesting thing still for me is that you prefer perplexity computer to Claude or to any other system. I I do because I value AI as a thinking partner, not as an answer machine. And so, as a thinking partner, what I'm able to do with Model Counsel, and remember, full transparency here, this is the $200 a month version, but it's $200 a month. Like, one sale for most people will more than justify the 24,00 bucks a month. So, I prefer it because I can hop between models. It is

[11:49] much more rigorous around citation and taking you to the links. It is also I think much stronger between just the LLM and then the agentic version with perplexity computer where I can get it to really dig and do some big work. Okay. All right. So now you've got the warm collision. You're not going to just start a conversation with me. You then get the dossier. In fact, you get it on everyone for every meeting. Can I see what that looks like? what your version

[12:16] of it is and why that's so helpful. So I get it to run OSENT doseier on Andrew Warner here and it'll go okay and it knows what OSENT is. I never heard of OSENT before. Yeah. So we it it is it is a military term for open source intelligence. Another way that my academic background became useful. I studied military history. I never thought my 10 years in academia would help. Turns out it helps a lot with AI and with military terms because Silicon Valley seems to be borrowing from the military a lot. Human in the loop is nothing more than who gets to pull the trigger. That's another

[12:47] military term that they've borrowed. But OSENT is open source intelligence. So how do you look at all the open- source dots on Andrew Warner and how do I pull together a doseier and overlap it with me so that we can find points of trust. Look, I I know that you're not big on the stats, but I'm going to give you the stats anyway. That it's taken about 29 touch points on average for a B2B sale to happen. That's a lot of interaction

[13:13] before you get to a sale. We're getting it down to about three to seven touch points before we get a conversion. Most of that is because we know how to show up to the first meeting to create trust so that we can get the permission to transact. Now, let me be really clear here. You can't cheat this system. If I don't actually like humans sniff BS a mile away. So, if you're trying to manipulate a situation, it's it's not going to work for you. It's not going to work for you. You have to find natural overlap. So, we're back to LinkedIn here. Executive Reed,

[13:45] right? So, it says who you are. Oo, it even has my birth name in there, which I guess it got from LinkedIn. Okay, but that's an old post on LinkedIn. All right. What else does it have? I'm in Austin career arc. Yes, it got that. Um, okay. What are you looking for as you're going through this? I'm reading for all of the potential overlaps with me. Now, what I would do after this is I'd put I'd say, "Hey, can you give me like natural overlaps?" And it will at the bottom of this. It'll say, "Hey, if you're trying to connect, here's what you want to here's what you

[14:17] want to focus on, right?" And you know, it's it's giving me an understanding of you. The reason why I want to know that is is sales is not about showing up as quote unquote your authentic self. I don't know about you, but I'm like five different people depending on who I'm talking to. So, it's about showing up in

[14:32] a way where does 1 plus 1 equal three. I'm looking for all the professional and personal overlaps so that we can have a really good meaningful conversation quickly. And let's say so in here you see that I run marathons. There it is. Solo marathon running. That's what I was going to get to. Do you then start bringing it up at because you're a marathon or two. I if it naturally goes there, I'll have between three and 20 overlaps personally and professionally and I am ready to go there depending on where the

[15:02] conversation goes. Okay. Right. So I I was on with one who was an ultramarathoner. I'm a marathoner and I knew that if we needed to go there, we could go there and create some common ground. Turns out he was also an engineer. He came on guns ablazing. He had already done his due diligence. I didn't need to do that. But I was able to close up the conversation to say, "Look, I was ready for this. This has been a great conversation, but I'm going to close this up and say you run ultramarathons. Can you help me understand what why you do that? Cuz I

[15:27] run marathons and those are hard enough. And we were able to have a good laugh and another notch in the relationship building. That was a three-etouch close in that one. I totally get that. And you know, the other interesting thing is you don't have this automated. You don't have it automatically go into your calendar every day and just create one of these for everybody and put a link up on your

[15:46] calendar or something. I do have that. Um, so if you go in here and you go into spaces, here's my sales meeting prep over here. So I created because, by the way, not to make this into a commercial for perplexity, but I've got to tell you the thing that I like about Perplexity Computer is you don't have to keep your laptop open the whole time the way you do with with Cloud Co-work. With Cloud Co-work, it has to be up and it could do this, but I just want Well, I guess they now have a cloud version. It's not It's not like

[16:09] this. This is all super cloud. It And if you look in here, I have this. So right ro context common ground step one the triage rules site everything tell me where it is look at the last 18 months tell me whether it's verified so this is a fairly long prompt in my spaces okay and then what it will do is it run a schedule tasks every morning at 5 a.m. takes a look at my Google calendar and it goes through and it gives me the outside meetings I have for that day and then it prepares the longer

[16:41] doseier for me for the next 48 hours. Now, if I want to go deeper, I can go deeper and say, "Okay, I really want to isolate this one. Can you go a little bit deeper?" But this is a a one click from my email in the morning. Okay, here's who I'm meeting. Here's my prep. Got every single person with the prep there. Okay, I got it. All right, let's close it out with the final thing is you get on a call with them and I ask you, do you use a meeting notetaker and you pulled something out of your wallet or out of your phone? What is that thing? I

[17:07] didn't get to to fully understand it. Yeah. So, on the back of my phone is something that looks like a wallet, but it's not. It is plaude. Um, and flawed. What is it? P L A U D. It's a I, you know, short of turning it into a commercial for them, too. I mean, it seems like perplexity and plaud is the way today on this on this show. Plaude is a AI transcription device. Now, in Canada, single-sided

[17:29] recording is legal. Um, in parts of the US it's not, parts of the US it is. So, make sure you understand whether it is, but I I have a best practice of just letting people know I'm recording. Now, Plaude uh is it's a couple hundred bucks and then it's uh $9.99 a month. It is Sock 2 compliant. Um, and it has multiple different prompts. So, if I'm in with my EA, if I'm with a sales meeting, whatever it is, I can grab that and then I can quickly with a couple of prompts get a proposal or a follow-up email or whatever it happens to be out

[18:01] of that. Got it. It's recording for you on the device then into the cloud and it gets everything. And why do you use that? Why did you tell me about that versus like a Zoom type meeting noteaker? Is that where you take most of your meetings? Well, so I I have some that are on phone. I have some that are in person. I have some that are online. I can keep it all in plaude, all three of them. It also records both sides of a phone call, which is really quite How does it do that? It's like I don't actually know, Andrew. I just

[18:30] know that it does very well. Yeah. Even if you're using earphones, it will record on the device. Yes. Okay. And so then now you've got it and you have all the uh all the things that you need in order to figure out how do I close the sale? The thing that you were telling me that you're looking for is like where's the breakdown in the business? And once you understand what the breakdown is, you can start to create a financial model for them of how you can help them over the next 12 months. You now have told me you're thinking about what's the big strategy that I could help them with. How do we

[18:58] help them with workflows? Because just telling people over and over, you're wrong, you're not getting enough, not going to get it done. Once you have workflows, the AI supports them. That's the whole thing. I love seeing your system. Can we give people in the des in the description I guess of this um the two prompts that you just showed us? The one for the the prep for the day and the other for here's who I am and what uh what o is or what what what I Yes, absolutely. I'll send those over to you. very happy to the the the one warning here is that I want when

[19:27] I'm looking for leads, I want to use whatever LLM I have, but I use Perplexity Computer and Model Counsel to create the prompt and it in creating the prompt, it will tell you when you're too general. And that's the single biggest mistake on lead generation is they go too wide, right? It's not look, you have infinity options to search here. You can go deep in multiple, but make sure you're going

[19:51] deep enough to get precision results. Now that you watch this, the next step is to see a founder who uses OpenClaw to run his computer. Maybe that's your jam. Click the link below and I'll see you in that

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt