18 | Breaking Business Boundaries: Learn How Dror Gill Harnessed ChatGPT to Build and Sell a Company in Under 60 Minutes!

0:00Hello, and welcome to Leveraging AI. This is Isar Metis, your host, and I have a very special episode today. Our guest today is Dror Gil. Dror is an Emmy Award winning technology and business strategist that is also an expert on generative AI. And shortly after ChatGPT came out, Dror gave himself an interesting task. He created everything needed for a startup, from idea to product through marketing plan, business plan, and even an exit strategy using ChatGPT in just one hour. When he published that on LinkedIn, it caught like wildfire. And I was as excited as everybody else and invited him to be a guest of the show.

0:40And he obliged in our conversation, he's going to share all the different steps that he took and everything he learned for every different aspect of the business going through this amazing and fascinating exercise. So it doesn't really matter which part you are in the organization of what you do in business, you will find something highly relevant to you on how to leverage AI in your particular role. At the end of the episode, I'm going to share some fascinating, exciting, and really big news that happened in the AI world in this past week. So stay tuned for that as well. And now to the amazing interview with Dror Gill.

1:18In the next few years, AI technology will change our world dramatically. Whether you are a business executive trying to catapult your business forward, or just somebody who refuses to be left behind and want to advance your career, this is the show for you. I'm your host, Isar Matis, a serial entrepreneur and an AI enthusiast. You'll hear invaluable practical tips from innovative business leaders, AI practitioners, and some of the brightest AI minds in our world today on how you can leverage AI in ethical ways to advance your career and grow your business.

1:59Hello and welcome to Leveraging AI, the podcast that shares practical, ethical ways to leverage artificial intelligence to improve efficiency and drive business growth. This is Isar Maitis, your host, and I am truly excited today. You know, I'm a startup guy. I've been a startup guy for many years. I've been in several different startups. I'm mentoring startups, and I'm investing in startups. And launching a startup is not easy. It's a very long process. There's a lot of things you can do, and all these things require generating different kinds of documents. Even from the very first step of research, you've got to start with researching what business you want and coming up with an idea and then picking the right idea and then coming up with a name and a slogan for the startup and a business model and a product plan and a marketing plan and an org chart and job descriptions to hire people for this org chart and investor presentations and content for your website and social media posts and marketing.

2:56materials. And then you start to develop the product. So you need to start writing code and managing the product. And when the time comes, you also need to start looking for people to purchase the business. So your exit strategy, find people who might be the right companies to acquire you. And then once you find the right ones, it's exchanging different kinds of legal documents for the acquisition itself. So this is tedious, long, and takes a very long time. And a lot of people are experts on these different topics to generate all these documents. At least it was until ChatGPT came up. And my guest today, Dror Gil, challenged himself to do all these things.

3:37So from inception of an idea for a startup, through everything that I mentioned, all the way to exit strategy, finding the right companies and writing legal documents for the purchase in one hour using ChatGPT. And he did it successfully, which is even crazier than just coming up with the idea to do that. He created an amazing post about it on LinkedIn. This is how I found him. That post obviously exploded. So I'm not the only person who saw that post or shared it or responded to it. But the idea itself is obviously not practical. You're not really going to start a business this way, but it generates an incredible toolkit, if you want, for every aspect of the business, which is really unique, right?

4:18Because a lot of people still look at generative AI and ChatGPT specifically as a marketing tool to generate content for different things. And this really opens and widens the applications to everything and anything in the business. Djor himself is a brilliant tech entrepreneur and strategist. He's been doing startup in various positions, both inside the businesses, starting the businesses, consulting to businesses on business strategy. And in the past year, he's an AI expert as well. And so I'm truly excited to have him as a guest of the show today.

4:56Thank you.

5:37Dror, welcome to Leveraging AI. Thank you very much, Esar. It's my pleasure being here. And thank you for this wonderful introduction. Thank you. It's all you. I'm just enjoying quoting everything you did. Dror, before we dive into the details and the process that you did, what drove you to that idea? Like, how did this start? I was playing around with ChatGPT. That was a few days after it came out. And I was really amazed by the amount of things that you can do with it and how much it can help you in being productive in almost anything that you do. And I tried various things like create some tweets or write a blog post or write a website, all these types of contents.

6:24But then I said, okay, what about ideas? And I found out that you can actually request ideas and ask the tool to come up with features for a product or for new markets for you to explore or business models or things like that. So I said, hey, I think I found a brainstorming buddy that could be a partner to my startup. Now, I have a lot of experience in startups. And as you said, you need to create a lot of documents, but you also need to come up with a lot of ideas and you have to be innovative. And initially people said, this thing, this engine just guesses the next word in a sentence. It's not really creative.

7:11My experience was completely different. It was creative. And I want to tell you about the aha moment I had with ChatGPT. And this is when I visited a friend of mine who showed me an idea I had for a product. And just the same week, I got acquainted with ChatGPT. So I said, do you know ChatGPT? I said, no. I said, okay, I opened my laptop and I typed, give me 10 features for a product that does this. and I described his product in one sentence. And I got 10 features and I turned the screen to this entrepreneur and his jaw dropped. He was amazed. He was looking at this and said, wow, these five features, I know them.

7:55I have them in the presentation already. But these five features, I didn't even think of them. But these are great ideas. I need to try them out. Even if one or two of them are good, I'm going to put them in the product. And this was all invented on the spot by ChatGPT based on the immense amount of information that it gathered from the training on all the information on the internet. So that's when I realized this is not just a tool for creating essays and blog posts. It can really be a brainstorming partner and come up with ideas. And I was discussing this with a friend. And then I said, okay, what if I could do, accelerate the cycle of a startup and pack it all in one hour?

8:39This was the challenge that I took upon myself in one hour to go through the whole process from no idea to exit. Now, as you said, I didn't plan to actually build a startup and start making money and revenues and all that. Some people wrote in the comments, yeah, but you didn't really. That wasn't the idea. The idea, as you said, was to create a toolbox for entrepreneurs and for people who work in startups and in other small businesses that you can use ChatGPT as your partner, ask it anything, and it can come up with a lot of information that you need, both ideas and also documents that you need and plans for making your startup work.

9:22So that's basically the background of where I got the idea. And I was surprised. And people keep asking me, did you really do it in one hour? And then I have to say one thing that at the time there was no paid plan. I was using the free plan. And after half an hour, it said too many messages in one hour. See you later. Okay. And this was in the morning. So I said, okay, half an hour. That's enough. In the afternoon, I came back and I worked for another 30 minutes and finished the project. So it really was one hour. It wasn't one hour straight because of the limitation of the free version at the time, but I actually did it.

10:02And it was great. And I was really surprised by the amount of traction that the post got on LinkedIn. It really went viral. Yeah. So first of all, I think it's brilliant. I think the thought process, I think the amazing thing for people who are in just paying attention, the amount of things people are doing with these tools, it gets broader every single day. Like literally every single day, somebody comes up with another use case that is amazing and that is working right now. So it's not the future, it's the present. And all you have to do is embrace it and you can use it for so many different aspects, both in your personal life as well as business.

10:39And with that, I really think let's start going step-by-step on the things you did. and if you can share some prompts that you've used or what was the process that you used in order to drill deeper? Because again, for people who haven't used these tools, drilling down is a part of the process. Just like I think, Dora, you use the perfect word. It's a brainstorming buddy. You send it a question, it gives you an idea. You're like, oh, so what about this? And then you keep on going deeper and deeper into the rabbit hole and finding more and more stuff. So how did you start? Okay, so this is a really, a key point what you said about the conversation.

11:15It has to be in conversation. And this is one of the major tips I'm giving in the workshops I do about ChatGPT in prompt engineers. Utilize the conversation and ask additional questions, ask for examples and et cetera. It's a podcast, so you can't really see the prompts and the responses, but Isar, if you want, I can show you. Yeah, I would love that. And then I can, when we produce the videos from it, then we can share them as well. Sure. This is where I started. I started with a question of how do you build a startup? What are the stages of building a startup? I pretended I don't know anything about it.

11:54Of course, I know a lot about it because I built several startups. So the answer was very straightforward, but also very complete. It said you need to start up, come up with an idea, then do market research, decide on your business model, do the product development of building an MVP and testing it with customers. Then you fundraise, you get funding so you can scale your startup. Then you do the scale, you expand the business and increase your customers and your revenues. And finally, you exit, you sell the business. Very simple, seven steps. And this is what I got from ChatGPT. I said, that's great.

12:32So let's start with step number one. Step number one is the idea. So give me 10 ideas for an innovative product or service that fills a need in the market and can generate significant revenues. Okay. So in the prompt, I already directed ChatGPT, not to just throw off general ideas, but things that it's a really innovative, have a good business model. And I got various ideas, like 10 ideas, like a subscription-based meal delivery, which customizes it based on your diet or a platform to connect homeowners with service providers such as cleaners and handymen and gardeners and stuff like that. And augmented reality mobile app for language translation, various things.

13:21And then I pushed it a bit further and said, okay, can you give me ideas that are more innovative and solve larger problems? And the interesting thing is, as I said, more innovative and larger problems and can get even higher revenues, then ChatGPT responded with more and more ideas that are based on AI. I don't know if this is a bias in the system because it is an AI or because today you have to use AI if you want to build a significant business and to become competitive. And it gave me 10 more and I asked for 10 more. And finally, I took one of those ideas, which was based on AI. And I said, okay, let's go with your idea number six, a platform that utilizes AI and natural language processing to improve the accuracy and speed of legal document review and analysis.

14:15Okay, so here, it's another tip for prompt engineering. I reminded him, what is that idea? Because he sent me so many ideas, and I wanted to be in the buffer, in the context. So I said, okay, this idea, give me a catchy name. So he came up with great names for a law base, for a startup that does legal document analysis with AI, like Case Cracker and Legal Ease. That's a good one. That is a good one. Yeah, Judgify. And eventually I chose Contract IQ. By the way, he wrote here as a note, please note it's important to check the name if it's taken. Now, I had one hour. I didn't have time to check.

14:56Contract IQ indeed is taken, but in real life, of course, you would have to check that. And by the way, just to add one thing to you, in the next iteration, and Bing already does it, you can actually check it in real time, right? So ChatGPT is offline. It's based on 2021 data, or the latest data, I should say it's from 2021. But the newer platforms are actually online, and it will not give you stuff. You can basically, again, prompt engineering, say, only give me names that are not currently taken and it will do that. So we're definitely going in a direction. It will try to do it. It will try to do it.

15:31That's fair. Because I've seen Bing give you references from the internet for things that it wrote. And I go to that reference, there's no connection between what it wrote and the reference. So you also need to check the reference that are given. Just go to that website online and see if it makes sense. I want to open parentheses for a second because what you said is very important. And I think a lot of people don't understand it. When you go to Bing and it shows you references, Bing doesn't know references. Bing is a huge, so chat GPT engine behind it, so GPT-3.4, well, now GPT-4 behind it, consumed a huge amount of data.

16:10And based on that data, it gives you answers. And the answers are literally created in real time, word by word, because that's how this engine works. It's not based on one or two or three articles. It's based on all the data it's ever consumed. So the references it's giving you, it's actually an extra step where it goes back after it thought about the answer. It's like, ooh, where did I see references that are somewhat related to what I'm saying? And then it quotes these references. It's not really based on these one or three or two links that it gives you. And hence what Drory is saying is very important.

16:42Go and check the links because sometimes they're not really related because that's not what the answer is based on. But I saw one great tip about articles. Sometimes it gives you references to articles, to papers, academic papers that don't exist. So somebody added to the prompt, give me only citations to papers that appear in Google Scholar and have at least 500 citations that somebody cited those documents. And it turns out that this really helps. Yeah. So going back to prompts, sorry for throwing us off the path. So now you have a name. Now I have the name. Yeah. So that's good. That's a good start.

17:22And then I said, okay, I have a name. Now give me a slogan. So the first slogans weren't very good. It was like revolutionizing legal document review with AI, streamlining legal workflows with contract IQ, smart legal document analysis with contract IQ. Pretty boring. So I said, okay, again, continuing the conversation, can you come up with some more witty or whimsical slogans? He said, sure. He's always there to help. Leave the legalese to us. Contract IQ has got you covered. Like this one, reading between the lines, contract IQ style. And this one, law and order, contract IQ way. Some very good ones here.

18:07And then the next step is the business model. So what are the potential business model if you have service that can analyze legal documents using, so I got nice answers. It can be a subscription model. It can be pay-per-use. You can use value-added resellers who you could partner with, and they would resell it as a service. It can be a licensing model that you license the technology to someone. You can get commission and, of course, freemium. Freemium is part of the subscription. So that was pretty good. And then I said, okay, let's dive in. Again, continuing the conversation, diving into value-added retailer, reseller, who could be possible companies.

18:49So he came up with a list. You can sell it through law firms, through accounting firms, through BPOs, business processing, outsourcing, and various other types of companies. Now, BPOs was something that was new to me. So I asked, what is a BPO and who are the major BPOs in the USA? say. So he said, okay, you can go with Accenture, Capgemini, Cognizant, Genpak, all these names. And then I said, okay, this can be really useful because instead of searching on Google and getting a lot of results and some of them are sponsored and some of them are biased and all of that, you just get a list of companies that you can approach one by one and use as your channel.

19:32So this was really great. And the next step, before starting with actual development, I said, I want to fund this. I asked it for an investor presentation. It gave me a very nice outline. And then I digged into each one of the slides and said, okay, give me the bullets of the slide. And you can also ask for an image to accompany the slide. and if you want, you can go outside of ChatGPT. Today, there are tools that give you complete presentations based on a few terms that you put in. So you can use those tools or you can build it with ChatGPT. And then came the step that I have the presentation.

20:11I need to show it to investors. So who can I ask for money? And then I got a list of investor types. You can go to angels, you can go to VCs, to incubators, family offices. The funny thing about family offices, when I said, okay, who are the family offices? Then he gave me a list of the largest billionaire family offices like Hoch and Rockefeller and Pritzker and all of those that you would never approach with a small startup. And that's another tip. Always, when you look at the answer, it's not always the answer you expect. And it's not always the right answer. You should basically know what you're doing when you're asking it to generate lists or information or ideas, and you need to verify.

20:55And this is an important step is verifying the output before you go. I wouldn't use it blindly because there are mistakes. Yeah. I want to pause you for a second on two aspects of what you're saying, which are very important for people just starting to use Chachi PT. First of all, the first answer you're going to get is really generic, right? It's like vanilla. You got angel investors, you got VCs, you got private equity. You got to say, I knew that. That's not really helpful. And you got to keep on drilling deeper and peeling the onion until you find gold nuggets in there. And it's true for literally everything that you do with ChatGPT.

21:30The first answer is going to really disappoint you. Unless you're really good with prompt engineering, if you know how to craft the question very well. And I'll give an example related to what you just said. I was doing real market research for something I'm considering investing in. And the first answers he gave me is, again, all the giants in the market, which obviously are not relevant customers for me because I have nothing. I'm just starting a new company. They're not going to buy from me. And then you start crafting and saying, okay, who are less known? Who are smaller companies? Who are companies that are not public?

22:00Companies who make less than X number of dollars. Companies who've been around less than X number of years that only do this little thing, but don't have other bigger services. So in the question, you start customizing it, telling it what you want, but also what you don't want. And then I got a list. And the other thing that I got to mention that people may or may not know, most of the answers you get from ChatGPT, if you don't tell it otherwise, is text, right? It's just bullet points. But you can ask it to create a table of a comparison between the list of these companies that you just have and say, okay, show me on the column one, all the companies, column two, links to the website, column three, a description of them, column four, how much money they make per year.

22:41If you have access to that, column five, the link to their CEO's LinkedIn data, and you'll get a table with all that information. It's not just the data that you can get. You can also get it in a variety of formats. If you define it that way, that will be helpful for you and for you to share with other people on your team. Yes, I agree. And one thing about the tables, I did a presentation to a nonprofit that helps people get jobs in the technology sector. And one of them said, can it improve my CV? And I said, of course. So they sent me the CV through the chat and I uploaded to copied and pasted the text into ChatGPT.

23:21And I said, improved the CV, what's wrong with it and why you chat. This actually showed them that not only getting the result is important, but also understanding the process. Because if you just send your CV and it improves it and you paste it into some job website, you didn't learn anything. But if you look at the table and you look and you see what was wrong with the CV and how it changed it, the next time you will know to do it better and you will improve. So that's another important tip. use chat gpt not just as an end-to-end tool but also ask it to show you what changes it made and how it improved and then you will learn and you will improve brilliant so taking us back to the startup story so now you had a name you had investors you had a slogan you had a product What was next?

24:16Yeah. So next is building the company. I need people. So I asked it for a typical org chart. So I said, okay, in your startup, you probably need a CEO, a CTO, a CFO, sales, marketing, operations, engineering, legal, and HR. Great. So I started with the technical and I asked it, how can I recruit someone for the startup? and I got a lot of recruitment sources for the candidates. I can go to job boards and career websites. I can ask other employees for referrals. There are recruitment agencies. Of course, you can post on social media, et cetera, et cetera. And then I asked it to create a job description for the CTO.

25:00And after that, I said, okay, I have the job description for the CTO. So then I went to LinkedIn and just took a random CTO who has some machine learning experience and pasted their experience into ChatGPT and asked it, okay, based on the job description of the CTO, here is an actual experience of a CTO from LinkedIn. Is he suitable for this? So then I got a very nice answer that the CTO knows about machine learnings and he knows the cloud platforms and all of that, but he has more experience in finance and less in legal. So you should invite him for an interview and see how well this fits. So another important tip here, ChatGPT not only generates information and documents and questions, it can analyze.

25:44So you can paste a CV and see if it's good or not. it will tell you according to a job description that you upload or it invents on its own. This is incredible. It's absolutely brilliant. Yeah. Then I said, okay, I'm inviting the CTO for an interview. What should I ask? So I got 10 questions to ask at the CTO. Now, if I record my interview with the CTO and transcribe it, I can go in to chat GPT said, okay, these are the answers to the interview questions that you generate it. Are these good answers or bad answers? And should I hire the CTO or not? So I can use ChatGPT in various stages in the process of recruiting, not just for generation, but also for analysis.

26:30And then I asked it what would be the typical salary of the CTO. And he said, okay, based on data from Glassdoor, this is the average, this is the range, et cetera. And also asked about benefits and things like that. The next thing is to build a product. So I got from ChatGPT a crash course by asking, how do I reach product market fit for ContractIQ? So I said, okay, this is what you need to do. You define your target market. You do research. You develop an MVP. You start testing. You gather the feedback, measure and evaluate, rinse and repeat. That's the basic steps. And it's true for ContractIQ, which is a software as a service for legal documents.

27:12So that was very nice. and then I started digging into the product, the specification of the product. And this is where I was really amazed because I didn't tell it anything about this startup. JGPT invented the startup. So I said, okay, give me a spec. So here is the spec. You need a user interface. You need a module that will do the document analysis. You need collaboration features so people can work together on those documents. you need reporting and analytics, you must handle the security, and you can do integration with other platforms that handle those legal documents. So that was very nice.