A (0:00)

Okay, today we're going to do something that I've been meaning to do for a while. So I want to test the differences between OpenAI chat GPT's deep research and Perplexity's Deep Research. I mean, the simplest way to think about it is chatgpt or Perplexity. When you prompt it, it's going to give you a quick answer. But Deep Research, it's almost like a junior analyst, a McKinsey analyst, that they would probably pay hundreds of thousand dollars a year to give you deep research into the question that you have. And today we're going to use it to actually build a business. So we're going to prompt it, we're going to give the same prompt to ChatGPT. This is a $200 a month pro plan and per Paxity's deep research, I think that's a $20 a month. And we're going to see which is better. And we're going to go through the startup ideas to actually see, hey, maybe there's actually a grid startup idea that one of you can steal. We're going to get chatgpt and Perplexity to give us the tactics of actually how to build this business. So this is what we're going to prompt it. We're going to give the same prompt to chatgpt as to Perplexity. So we're going to say, I want to create a startup. I want it to be an agent, say an AI agent that focuses on a high value niche. I want it to be defensible, not just a flash in the pan. I want you to create a zero to one playbook of how you'd launch it to get customers too. Give me five ideas with five playbooks on how to grow it. I'm looking for $1 million air $3 million year one, $3 million year two, $5 million year three. So we're gonna, we're gonna see if ChatGPT Pro or Perplexity could give us a $5 million AR business in three years with the exact tactics, how to grow. And we're going to be real with it. We're going to see, is this actually going to give us good particular advice. So if you're going to do this, you got to click the deep research button here. So it says that sounds like an exciting venture to tailor the best ideas and playbooks for you. Could you clarify industries of interest? Sure. So we're going to say, are there any specific industries? No, we are looking for the startup ideas with the highest, like the hood of success 2. Would you rather target enterprises, SMB or individuals, professionals? The truth is I don't really care if it works, it works. So, so open to B2B or B2C and then we're going to say AI capabilities. Do you have a preference for AI applications, automation, data analysis? Not really. Again, we're looking for the highest probability of success with a moat. Okay, technical resources. Do you have a technical team or plan to build one or do you prefer a low code, no code approach? And we can say I prefer an approach that's going to cost $5,000 or less for an MVP. I'm not super technical. What else can we say? We can say I might want to just cash flow this business but but I want to potentially raise VC money too. If I really think this could go beyond $100 million a year business go to market. Would you lean towards direct sales partnerships or viral self serve growth? I think I lean towards viral self serve growth but I'm open to what you suggest. Boom. Now we are going to wait until to see how long it's going to take. Usually it takes like eight minutes. So let's see how long. So it's starting research. So while it's doing that we're actually going to head to Perplexity and give it the same prompt and see what happens. And we're going to compare, we are going to compare all these together. So let's go open up Perplexity. Quick break in the pod to tell you a little bit about Startup Empire. So Startup Empire is my private membership where it's a bunch of people like me, like you, who want to build out their startup ideas. Now they're looking for content to help accelerate that. They're looking for potential co founders, they're looking for tutorials from people like me to come in and tell them how do you do email marketing, how do you build an audience, how do you go viral on Twitter, all these different things. That's exactly what Startup Empire is. And it's for people who want to start a startup but are looking for ideas or it's for people who have a startup but just they're not seeing the traction that they need. So you can check out the link to StartupEmpire Co in the description. Now Deep Research is a brand new product in order to enable it. You'll see it in this little dropdown here. Let's go add our prompt. And the interesting thing is going to be to see if is it going to ask us for follow up questions like ChatGPT did not sure. We will find out soon. Deep research. One of the cool things about Deep Research, both chatgpt and Perplexity, I think, is they go through. Instead of going through two sources or three sources, they're going through way more sources. So you can see already this is pulling for six sources. By the way. I'm like, pretty new to this. I'm using this for the first time really. I've used ChatGPT a little bit, but we're kind of learning this together and we're going to find out by the end of the episode if this is super Value or not. I'm not sure. So we'll find out. So this is, it feels like a reasoning model. So you can see, you can kind of see how it's thinking about things. So it says I need to search the web to find information on zero to one playbooks specific to AI. Found some insight into zero to one startup Playbooks. Cool. Great. From the previous search, I found insights. Great. I found insights on defensible AI strategies. Okay, this is cool. I don't necessarily need to know exactly how it's doing it, but it's, you know, it's, it's, it's, it's good to know that it's crossing. It's. What's the expression crossing? Its T's, dotting its I's. Someone in the comment section, tell me what the expression is. But, you know, it's basically that, you know, it's good that Perplexity is checking out all the stuff and you can see it's going through 29 sources. Right. So if you, if we were to do this manually, if we went to Google, I mean, if you went to Google, I don't, you couldn't do this. But if we actually went and did this manually, this is going to take us many, many, many hours. So the hope here is that this is basically AI and intelligence that could help us. One thing interesting is that Perplexity is not asking me a qualification question like we saw in ChatGPT. ChatGPT asked that qualification question. So we'll see the difference in terms of what's a higher quality end result. Okay, cool. So going back to Perplexity, it says it's now gathered sufficient information to answer your query. Answering the query 30. Now we're up to 34 sources. So this is exactly, I mean, if this was a real person, this is exactly how they would answer it. Okay, so first things first is that Perplexity is faster than ChatGPT. You can see that Chat GPT is about halfway done. Perplexity is done. So that's really cool. So let's go through the ideas. Let's go see. Funny enough, actually one of the sources is me. One of the sources is went through one of my YouTube videos. Three $1 million AI startup ideas to build in 2025 links to it and that's hilarious. That's actually, it's kind of ridiculous but it's cool that it goes through YouTube. You can see and it obviously downloads that transcript and goes through it. So let's go through these ideas and you be the judge, you tell me, are these good ideas? Are they bad ideas? Let's go see. So number one legal contract. This is on the Perplexity deep research answer Legal contract lifecycle automation agent problem space Corporate legal teams waste 23% of their time manually reviewing boy the prey contracts while struggling with version control errors that cost enterprises 2.4 billion annually compliance penalties. Existing CLM solutions require extensive customization and fail to adapt to company specific negotiation patterns. Okay, so an AI agent. Here's a defensible solution that integrates number one integrates directly with email, SharePoint and DocuSign via API. Two learns organizational risk thresholds through historical contract analysis. Three. Auto redlines third party paper using firm specific fallback positions and four maintains version lineage with blockchain style audits trails. Now by the way, I think this is actually cool but like for an MVP minimal viable product, do we need blockchailstein audit trails? Not really. But I guess we can further prompt Perplexity to, you know, ask it what would be the MVP for this? Right? And unlike. Well in ChatGPT we said it was going to be $5,000 or less. So that's an interesting thing that if you are going to prom Perplexity you might want to include some more constraints because you probably won't get that follow up question why is it defensible? Develops institutional memory through continuous exposure to negotiation outcomes. Becomes embedded in legal department workflows. Average 14.3 logins per week per user. This is data. This is great. This is a great insight. This is something that if you're building a business, you want to know the end. You want to know that it's really helpful. Requires retraining equivalent to six to nine months of human experience. Replace it. And by the way, you can always click on one of these numbers to get to the source and go deeper. So here we have the growth playbook by by Perplexity. And let's go see if this is actually Good. This, I will say this agent, I mean seems super valuable and I totally agree why it would be defensible. So I like what I'm seeing. I like what I'm seeing from perplexity. So the growth playbook phase one niche validation 90 days product build a Chrome extension recording legal contact contract review sessions with permissions to train initial model go to market Offer free redline audits for 100AMLA 200 firms. I guess that's American law targeting mid market companies 50 to 500 million revenue. So this actually idea, free redline audits is interesting as a Chrome extension like I would never have thought of that again I'm not, well I'm not a lawyer, I don't play to be on the Internet so but just, just as like a person who builds products that you know, a lot of people use. I will say that just looking at that, that makes a lot of sense to me. Monetization $2,500 a month subscription locked to specific document types. NDA's first. Okay, that's it. Again really interesting. Wouldn't have thought about locking a subscription to specific document types. So that's really cool. Phase 2 Workflow Embedding Months 4 to 9 Integrate with legal Ops platforms like Lexion and Entre Launch clause library builder using clients historical favorite language. Implement usage based pricing at $0.25 per review clause. Really interesting, interesting business model. I like that, I like the phases. You know phase one to phase two, it totally makes sense to me. I like what I'm seeing. Phase 3 Enterprise Scaling Months 10 to 12 Add SoC2 Type 2 compliance for regulated industries. Develop custom connectors for ServiceNow and Salesforce CPQ. Upsell to $8,000 a month Enterprise tier with custom Playbooks Financial roadmap. Okay, so let's see how they, how they break this out. So on year one you're going to get 40 clients at two and a half thousand dollars MRR. That's 1.2 million ARR. Not crazy. 150 clients at 5000 MRR plus 250k usage fee. Year two you get 3.1 million and then year three platform fees from 300 plus embedded instance 5.4 million. So I look at this idea and I'm like I could see this idea being so and so YC company gets funded to go build this. This is a really good idea. The Growth Playbook. Although it's not super long, I have a feeling that the ChatGPT deep research is going to give me a longer answer. But I think that overall this Is like, you know, we say on this channel, do you, do you spit not like an idea or do you sip an idea? Do you like the idea? This is a sip. I sip. This idea makes a lot of sense. This is a pretty good one. Pager. On the negative side, I think you do need to go deeper into this meaning I think you need to prompt it more. But overall, how can I complain? It took me a couple minutes and I got an incredible idea and that's just the beginning of it. Let's go back to ChatGPT deep research. Let's see if it's done. Okay, so it's done. So let's compare what we saw with perplexity. Deep research with ChatGPT. It would be hilarious if it was the same idea, but it is not. So AI sales outreach assistant. I mean I'm kind of skeptical. I will say I feel like I've seen so many of these, but I'm open minded. All right, let's go see. First of all, let's just quickly scroll. Look how long this is like. This is a full on. It's like a business mind, literally. Okay, so I'm going to read this whole thing. This is, you know, it could be a really good idea. I don't know. We're going to find out together. So it's an AI powered sales outreach assistant that acts as a virtual SDR which is a sales development rep, automating prospect research and personalized cold email LinkedIn messages. The target market is B2B companies which are startups to midsize firms that need more pipeline without hiring extra SDRs. The AI agent enriches lead data and crafts tailored outreach at scale. That events ability comes from a data network effect. It learns from every interaction. By the way, data network effect, good term, really good term. I feel like Deep research knows what it's talking about. Like that is an incredible term. It learns from every interaction which messaging yields replies. Building a unique data set, what works in each industry. Again, this is really important if you're building an AI startup. Building a unique data set of what works is going to be your moat. So really cool to see that in there. Over time it develops superior personalization and lead scoring tuned to each customer's context, which is hard for competitors to replicate. Integration into CRMs and email systems further deepens the moat by embedding the AI in the sales team's workflow. B2B sales teams are increasingly adopting such AI prospecting tools and 66% of sales reps say AI helps them better Understand customers and personalized outreach underscoring demand for an agent like this. The agent can leverage massive contact databases. Okay, we get the idea. So how would they actually build this? What does ChatGPT say? So focus on a narrow but high value use case to keep it simple. For example, start with automating follow up emails to warm leads. Okay, cool. I'm really happy it's saying this because this is an idea that I've heard before. It's actually become almost a meme on X. Of course you're building an AI virtual sdr, but this idea of automating follow up emails to warm leads using a templated but AI personalized approach. Now this is something that is unique I really haven't seen anywhere. And it's a cool idea. The MVP could be a web app or a Chrome extension where a salesperson inputs a list of prospects or connects their CRM and the AI generates and sends personalized emails on a schedule to build this cheaply. Okay, now we're talking. They talk in our language. Use existing AI APIs. Leverage a service like OpenAI GPT4. Hilarious that like ChatGPT is being like you. Use a service like ChatGPT via API to generate outreach text given key inputs, prospect info, value prop. No expensive model training needed initially. No frills ui. Create a simple interface using low code tools or templates for the user to review. Edit AI suggested messages. Skip complex analytics at first single channel focus. Start with just email. Cheaper to implement before tackling LinkedIn or phone. True manual fallback. If AI is unsure, let it flag for human to handle. Ensuring quality while gathered data improve prompts. That's an important. It's important that it has that and I'm happy that they put that in there. This MVP could be built by a freelance developer in a few weeks. The main cost, less than $5,000 would be a developer time and API calls. By using off the shelf AI and focusing on one workflow, you can launch quickly and start onboarding a few beta users to prove its value. Okay, overall helpful. You're definitely going to need to double click into this and prompt it deeper to really understand exactly what the Playbook is. I mean it asks for a Playbook. Is this a playbook? You know, maybe it's kind of like an outline to the Playbook, but it's definitely some really good ideas on how to get started. Customer acquisition and growth so product led viral adoption among sales teams and low friction signups. So let's see if this makes sense. So a self serve free tier offer a limited free plan 50 AI generated emails per month so individuals, so individual sales reps or small startups can try out without approval. This gets foot in the door at companies viral loop via results. Whenever the AI assistant books a meeting or gets a reply prompt the user to share the success internally. This meeting was set by AI salespeople Love sharing hacks at work. This word of mouth can drive wide adoption. That's true. That is really. That is good referral incentives. Yeah as more reps use it the AI can perform better with shared data creating a network effect. Content marketing published case studies how X startup tripled using AI and short demos I mean that's good right? And look there's the link to start a story how a 27 year old grew Jenny AI to 5 million ARR. I mean what's cool about this is like this is so relevant. This is such a relevant story and it's cool that it's bringing that in there and it's knowing that this is high quality. I like that. That's definitely points for deep research list In CRM marketplaces like HubSpot being present in these channels gives you distribution. Yeah so the goal would be to reach on the order of a few hundred paying teams within year one. If say 200 companies convert to paid plans at $5,000 a year on average that's a million bucks ARR growth to 3 million in year two might come from a land and expand. I like that they're including that Free users converting to paid and small team paid users expanding usage or upgrading tiers as they see roi. The bottoms up adoption means minimal spend on sales. The product essentially sells itself by proving value Similar to how another AI SaaS product cursor scaled purely by developers, word of mouth with no sales team needed and individuals pulling into their teams again I really am impressed that they're using Jenny AI cursor. These are the examples I want to see in here. So that's really cool. So we're going to go through monetization and pricing then we're going to know VC readiness and then that's it and then we're going to compare perplexity to chat GPT. So let's go back to monetization. So it's a per seat pricing or usage based tier. For example charge per sales rep user $150 per month for unlimited outreach or by volume Typical structure could be basic plan for $49 a month, solo rep, limited emails, team plan for $199 a month five reps more emails and CRM and Enterprise for advanced or higher volumes. That makes sense. I'm more interested in usage based tiers. I wish it would have said like you should do usage based tiers because that's the future of SaaS. But I mean I'm happy that it even included usage rate tiers. Again, if you're doing, if you're going through, if you're going to be using deep research, I think you gotta on both Perplexity or chatgpt you definitely need a. If you have question marks, it seems like you gotta ask it again, which is totally fair at least with this. And you might have to prompt it five more times or seven more times. But if you're just using ChatGPT Regular or Perplexity Regular, you might go through 50 or 100 prompts to get to where you're getting how high definition and the clarity that this is. So it's, you know, it gives you some ideas on how to optimize revenue. You know, freemium upsell annual deals, premium features, performance based options. These are all good ideas. You're not really there yet. Honestly like when you're, it's good to like think about what this could be but like if you're building an mvp, zero to one like you don't necessarily need this but it is good that they included it. The VC readiness and scaling to $100 million ARR. So I think I included that in the, in the second prompt where I was like this needs to be able to get to $100 million arrow. It's cool that it says that it's possible. It's cool that this is a hundred million dollar a year idea that it's like yeah, it's possible and it's cool that it's like to get from $200 million ARR, you're probably going to need a raise VC it's pretty awesome. It's pretty honest with it. So that's the ChatGPT deep research now perplexity. And now if we can compare them, you can see Perplexity is a lot smaller, right? This is a lot more in depth. But this is really like there's no fluff with this. I gotta say this is, you know, the question is, is ChatGPT Pro worth 10 times the price of Perplexity? I'll let you decide. You know, it depends on how you know, do you like in this you're going to have to prompt a little bit more this, you're going to one shot it more often. ChatGPT you're going to one shot it more often. So it depends on your style. But I actually don't pay for perplexity on this account. So yeah, you can see I don't have pro on this account. So you can even use this for free. I think you can only use a few times, maybe a month for free and then you have to pay $20 a month. But that's incredible. Um, it's incredible that. Let's see here. Let's just click. Yeah. Okay. So I'm not sure exactly how many free deep research you get. Probably not that many. But the good news is both of these are really really good. These are like insanely good, scary good. And let's just do for fun, for idea number one, let's just say I feel like I've seen this idea before on X. How can you take this idea and make it more non obvious? See what happens. So it says to make the AI startup idea more non obvious and unique. We can find underserved niches, combine AI with another trend left proprietary data, introduce a new business model or enhance with the viral element. Do you have any preferences for how we make it non obvious? For example, should we explore a new niche? Let's say you're the pro. I pay you $200 a month for this. How about you decide on the idea that has the most legs and let's see, let's let Chachi V2 do the work and let's see what happens. I've actually never done a follow up prompt to to something to once it gives you like the full answer. So it's unfortunate that it takes so long. You know, I wish it just like was faster. That's what I like about perplexity. You know you can get in the zone. Let's see, let's do a follow up question to this one and let's say aren't there a bunch of highly funded competitors in the AI legal space? I feel like I'm going to get smoked. How do we take this idea and make it more non obvious? Also can you give me the step by step plan on how to implement the the MVP features etc. And make sure that this startup MVP will cost $5,000 max. Let's see how fast or slow Perplexity Deep Research will take. It says here deep research takes a couple minutes. Let's go back to ChatGPT. So while it's thinking both of these are thinking, I think the thing to think about is for whatever your workflow is, if you're building a zero to one startup, you want to build a zero to one startup, you want to build zero to one features. I think that having a companion like Deep Research, putting it into your workflow is just, it's a no brainer. It's a no brainer because it's going to make you smarter at what you do and it's going to help you ref, it's going to help you refine the types of products that you build so that you have the highest probability of success. It's like why not? Right? So let's see what's happening here. Perplexity is saying that it's found highly funded competitors like Harvey and Case Tax. So you do need to like I know this, right? I know that there's highly funded competitors in the AI legal space. I think that you do you still, the onus is still on the prompter. On the prompt because you do need to know the right questions to ask. I mean this isn't rocket science in terms of what I said, but you do need to be somewhat knowledgeable. Okay, so we just got the answer from Perplexity Deep Research and it's given me an executive summary. Wow. An executive summary of what's going on in the legal AI startup space. So our analysis of 37 funded competitors reveal three underdefended niches and it just goes through it gives me a competitive analysis. It writes this amazing kind of cool opportunity matrix. Wow, this is cool. This is cool. Just like writing what the differentiator is for example clause genealogy. What is the technical approach blockchain inspired version lineage and what is the defensibility source 22,000 annotated NDA corpus and then it gives you a full on MVP implementation plan with a $4,800 budget. Less than 5,000. They're going to give you $200 to go for dinner. Thank you Perplexity. They're going to give you $200 not to spend it at Chat GPT Pro. Maybe that's why they chose $4800. And they say here's a Chrome extension prototype records and anonymized contract review sessions, tracks cursor movements, edit frequencies, external research sources and the tools are manifest v3 and tensorflow js which is zero dev cost legal team interviews and then it gives you phase two core feature build with the different components like the clause library and how to build it with the cost is this is pretty darn good. This is like this is really good. And then the go to market sequencing which to me like is among the most important Parts like yes, knowing what product to create is obviously really helpful and what the rollout of the MVP should be. But distribution is the new moat. So how to figure out go to marketing is really important. Offer fee free clause vulnerability audit, generating historical enforcement risk scores, alternative language success rates and then 15 cents a clause auto red line. I mean this is, these are great ideas. It's giving you the monthly searches what the CPC is. So it says go to organic search. You know, it's telling you how to do webinars. I mean it's not as deep as the ChatGPT Pro, but I love these tables. I think this is awesome. And this is like this is what you need to know and you have to just re prompt it to go deeper into these sections. So let's go see if ChatGPT is finally done. It is done. So it gave me new ideas and it says AI platform for Long Tail Podcasts and sponsors an AI agent that not only streamlines podcast production but also bundles small podcasts for advertisers. Unlocking the Long Tail podcast advertising. This is pretty cool. Notice how it's like a literal research document. If you're watching this on YouTube or if you're watching the. If you're listening to the startup ideas podcast on Spotify and Apple, check it out on YouTube because you can, you can just see how in depth this is. Wow. So this is a pretty good idea. So it talks about how basically what the competitive mode is the proprietary data advantage. So for example, as the platform onboards podcasts and runs sponsorship campaigns, it accumulates unique data on what content content resonates and which ad placements work for each niche, which is really cool. Over time, the AI learns the style, tone and listener behavior of each show and which sponsors messages yield results. This creates a continuous learning loop that improves targeting. So it's got some good ideas. In the MVP development for 5000. Start with the pain point of time consuming content creation. Implement an AI transcript and summary generator for podcast episodes. Tells you exactly how to do it, what technical how to do it, Whisper or assembly AI. I do feel a bit weird that it recommends its own product, but I get it. It's kind of biased, but here we are. And then it tells you make a basic web interface manual wizard of Oz sponsor matching. It just goes a little deeper. It goes a little deeper than perplexity. Yeah, I'm not going to go through all of this. Obviously this is like super long, but I'm just scanning it and I'm like this is really darn Good. This is really good. So I didn't have time to go through all five of the ChatGPT deep research startup ideas and the five from Perplexity deep Research, but I've included it in a doc. It's completely free to download. I'll include the link in the show notes for anyone who wants to steal those ideas, go implement them and build a big business. There you have it, folks. If you aren't using one of these platforms, you probably should. The good news is there's something for free. Perplexity Deep Research. It's really good. It's really fast. You might have to do a little more prompting to get to the level of depth that you want, but some people might even prefer it. It's like driving manual instead of driving automatic on your car. ChatGPT Pro Deep Research, you have to wait longer, you get more in depth reports. But the quality, you know, it's high quality stuff. I don't think you can go wrong with, you know, either product. They're both awesome and it's really, you know, what floats your boat. I think you just need to, you know, if you can play with both and then you can always cancel one or the other. Well, Perplexity is free, so you won't even need to cancel it. But I think the important thing is use it, play with it, come up with ideas. Because this is an unfair advantage right now. And that's why I wanted to show it to you all. I mean, frankly, I wanted to learn it myself because I want to implement it into my day to day. If you enjoyed this video, please let me know in the comment section. Just say something. I read every single comment you can like and subscribe for more of this in your feed and I'll see you on the next one. Happy building. And I hope this got your creative juices flowing. Thanks for listening to the Startup Ideas podcast later guys.