ChatGPT is my co-founder

Common barriers hold entrepreneurs back, AI can help founders cross them

The world is full of entrepreneurs-in-waiting. And, thanks to AI, many of them can stop waiting.

Most entrepreneurial journeys end before they begin. This comprehensive study shows around 1/3 of Americans have had a startup idea in the last 5 years but few act on it — less than half even do any web research! This matches my own experience an entrepreneurship professor (and former entrepreneur). The number one question I get asked is “what do I do now?”

I have been trying to teach people how to answer that question for a long time (you can even play my short free game, or use my longer game, that teaches you entrepreneurship), but I think ChatGPT might be able to provide some incredibly useful help along the way. It is no substitute for education¹, instead it provides a tool for helping founders in a very practical way. And this is particularly valuable for founders who may otherwise have not started a company because they are held back by a perceived weakness, whether that be poor writing ability or a lack of experience with topics like market research.

This graph shows how few potential entrepreneurs take action (the link to the Bennet and Chatterji paper is above)

Interestingly, ChatGPT’s own weaknesses (like lying often and not being connected to the internet) are not major problems for this narrow use in entrepreneurship. We are not going to use it to research facts, or take its legal advice (more on that later), or substitute for human judgement. Instead we are going to use it as a tool to lower the friction of the parts of the entrepreneurial journey that short circuit so many people. Let’s start with…

Idea Generation

I have a whole guide on how to come with ideas using ChatGPT, but here I want to focus on its really impressive ability to generate plausible, customized business ideas for many types of people.

This is remarkably easy. I simply typed in Generate 20 business ideas for me… and then a bunch of different identities. As someone who looks at startup ideas for a living, a decent number of the results had potential, even if they aren’t breakthroughs. Just some examples:

I am a truck driver who speaks two languages but has no college degree… Idea # 2. A transportation company that specializes in transporting goods across international borders, leveraging your language skills and knowledge of the trucking industry.

I am a college student specializing in graphic design with an interest in cars… Idea #7. Vehicle wrap and vinyl graphics design and installation.

I am a beekeeper who doesn’t want to keep bees…. #6 Bee-friendly gardening and landscaping services.

I am a doctor who no longer wants to practice medicine, but wants to start a company… #10. Medical tourism planning and coordination services.

These ideas aren’t bad, but they work even better to spark personal exploration of startup concepts. And, if you like to have the AI help, you can always tell ChatGPT to revise the results as well. Once we have an idea we like, say the Medical Tourism idea² (don’t know what that is? Ask: Can you explain idea 10 on the list?), we can consider other steps.

Researching your idea

Now, we run into a weak spot of ChatGPT, asking it for facts is usually a mistake. It is often wrong, or out-of-date, or will refuse to answer the question. ChatGPT is not Google.

But what can actually help an entrepreneur succeed is not a market sizing estimate (which, even if it was accurate, is not particularly helpful for really assessing an idea; market sizing is usually something you use to persuade others), but rather how we go about doing that. Remember, potential founders often get lost, and need just a little guidance on the. And ChatGPT is very good at that…

None of these are bad approaches, but they may not be intuitive, either. Lets get some help on what it means to learn about customer needs, for example.

Again, all good advice, and not that different from what I suggest in my book. But we can go even further.

Give me a list of questions I should ask potential customers about this idea. Assume they don’t know what medical tourism is.

Again, ChatGPT gives you a customized list, that, while not quite what I would suggest in class (too many leading questions, not enough focus on surfacing the root cause of their needs), is not a bad starting place for someone who doesn’t have a clear next step. And we can go further. You can paste in interview transcripts, and it can help you explore them.

Since I don’t have a medical tourism transcript, I used an interview with Steve Wozniak from Jessica Livingston’s book Founders at Work. I put in pages of interviews (broken into parts), and the summaries it produced were pretty neat.

Summarize the interview.

In this interview, Steve Wozniak discusses his experiences building the first Apple computer, the Apple I. He talks about how he designed the terminal for the computer and built the first language for the microprocessor, even though he had no formal training or experience in writing computer languages. He also discusses how he and Steve Jobs started Apple Computer and how they sold the Apple I to a computer store in Palo Alto, California. He talks about how the Apple II was a more efficient and refined computer than the Apple I, and how he and Steve Jobs built the company into a successful business.

How did he create a language without having any training?

Steve Wozniak created the first language for the microprocessor in the Apple I by studying manuals and books on compiler design and coming up with a syntax table that showed the grammar of the language. He then used this table to scan the user's input and output the appropriate code. This allowed him to create a language without having any formal training or experience.

You can see how this tool can help you begin to analyze and explore your results. Again, it is no substitute for a careful human, but it can be useful companion for exploring the complex data associated with an interview, or a market research report.

Building a plan

It should not be surprising at this point that you can ask ChatGPT to outline a business plan for this idea. But, again, the value is in going deeper. ChatGPT suggests that I should have a sales strategy. Give me some examples of sales strategies for this business idea. be specific.

Again, these are fine results, if not particularly inspired. The fact that some might be bad, or inaccurate, doesn’t matter because no entrepreneur will be able to simply follow the AI, or be able to get enough useful advice to run a business. Instead, the lowering of barriers is what is important. And these answers are more than good enough to help point to next steps.

To be clear, I would be careful with taking specific advice from the AI too seriously (when I asked it how to split equity, it gave good advice until I pushed it for a specific formula, and it suggested an even equity split - always a potential problem), but as a motivator for next steps, and a way of outlining potential plans, it can be very helpful.

Helping with Execution

Let us return to the sales techniques listed above. The third idea the AI provided was to create a blog post. That seems tedious, lets ask can you create a blog post as suggested in point 3? Here is an excerpt:

Again, serviceable, and it needs to be fact-checked, but it serves as a useful first step for an overworked founder.

It also suggested collaborating with a travel agent. I don’t know how to do that. Can you create an email for collaborating with a travel agent, like in Point 5?

It can also write invoices, letters to banks, applications to local grant competitions, and so much more. While a human will need to go through each, having a base to work with dramatically lowers the effort required to do something new. I think you can see how this might save an inexperienced founder incredible amounts of time and worry.

AI as co-founder

AI is not a substitute for advice from professionals, or even for classes on entrepreneurship (and yes, controlled studies show entrepreneurship can be, and should be, taught), instead, it fills a different role: it lowers barriers and multiplies the often over-stretched time of the busy founder.

Many people stop too quickly on the road to entrepreneurship, held back by real or perceived gaps in their knowledge, talents, and time. AI will not help with all of these problems, but it can certainly overcome many. And since entrepreneurship is a great path to both personal and community success, this is an exciting development.

1. And I am not just saying that as a professor. If you don’t have some idea of what you are doing, you can be led astray by the AI or make bad decisions based on its advice. Plus, it doesn’t really have a grasp of the most advanced techniques in entrepreneurship, or the reason why founders may want to use some particular approaches over others.

2. I picked this idea because a very similar one won an idea contest in a MBA class a few years back.