From Zero to “Top 10 Gen AI Startups” in 289 Days: 11 Lessons From the Frontline
289 days ago, 3 of us created a no-code custom chatbot building platform in 4 days — and launched it on ProductHunt.
Fast forward to this month: Our startup, CustomGPT, was named to “Top 10 Emerging Leaders in Generative AI” along with OpenAI, Microsoft, Google, Amazon, Anthropic, Cohere, Databricks, A121 Labs and Aleph Alpha — by GAI Insights, an analyst firm focussed on Generative AI. That is some serious company there!
Now, this is the first time in nine months that I’ve had the chance to come up for air, and I thought I’d share some of my lessons from the front lines.
These are raw lessons from the School of Hard Knocks.
So without further ado, let’s get started.
Lesson 1: BC — Before ChatGPT
Any form of thinking and thought process from B.C. — throw it out the window.
This is true for your own line of thinking or the line of thinking of people you are working with — partners, stakeholders, employees, contractors, developers, etc.
If you sense somebody’s thought process is stuck in 2022 and before, run away as fast as you can.
For example: If you get a quote from a vendor and you sense that he is NOT using ChatGPT (or AI) and the quote is based on a 2022 way of doing things, stop right there! (And either run away — or force the vendor to adapt!)
Now, this might be tough because our natural instinct is to always rely on our experience. But the world has changed so much in 2023 that traditional experience is now a liability.
Especially for old people like me. I had to make a conscious effort in my old age (LOL!) to throw away some of the old forms of thinking and experiences I would normally rely on.
Lesson 2: Once In A Century Event
What is happening today with generative AI is a once-in-a-century event. I’ve seen some quotes saying that the last time something like this happened was with the invention of the steam engine.
AI will do to human intelligence what the steam engine did to human muscle.
This means that a lot of the traditional rules change. It would have been impossible for young startups to disrupt billion-dollar companies. But now, with the democratization of generative AI, you could take your pick of billion-dollar companies to disrupt.
For example: Look at Midjourney, which apparently hit a billion-dollar valuation with 11 employees.
Lesson 3: Launch, Launch, Launch
It took us 4 days to go from idea to launch. We launched our MVP product on a single machine, duct-taped together with 4–5 scripts and launched it on ProductHunt.
Not only was it a bare-bones MVP, but it ran in a completely unscalable way on a single machine.
But the launch worked. Less than four hours after the launch on ProductHunt, it was evident that the market wanted an easy, no-code way to build ChatGPT with their own content.
And the launch exceeded our wildest expectations. I expected we would have 2–3 paid subscribers on Day 1; instead, around 50 paid subscriptions rolled in on Day 1.
Lesson 4: What Matters Most Is Your Daily Pace Of Iteration
Yes, launching an MVP is cool. But what matters even more is:
On a day-to-day basis, what is your pace of iteration?
Because it really does not matter where you start. What matters is your growth rate on a daily basis. Good things will happen if you can listen to customers and do 2–3 releases a day.
It will smooth over mistakes that you make. It will allow you to iterate features that customers want quickly. It will make things exciting for your engineers, developers, and every stakeholder who comes in touch with you.
The pace of iteration and daily growth rate is the soothing balm for many problems.
Lesson 5: Concentrate On A Good Product
Let the product sell itself. And let your product be your marketing. At Least initially!
The natural instinct of entrepreneurs is to always get into marketing and sales quickly. And all sorts of analytics and dashboards on Day 1.
Startup founders like to see nice, beautiful analytics dashboards.
But what matters most is the quality of your product and do customers love it.
If there’s only one thing you need to do, it’s:
Keep an eye out for the number of customers that really love your product. Not just like, but LOVE.
In the first 100 days, our website had zero SEO, zero SEM and almost no social media presence. But we had started accumulating tens of happy customers. I could sense this in the “Thank You” notes I was getting each day in the Helpdesk.
Lesson 6: Global Workflows
There is no luxury of waiting for the right guy to come along when you are a startup.
So, by creating artificial constraints on yourself, like requiring a full-time local employee, you are severely hampering your startup’s pace of iteration and the growth that you can achieve on a daily basis.
It’s okay. Hire that expert from an unknown village somewhere in a remote country.
For example: you can find diamonds in the rough using global marketplaces like Upwork and Fiverr.
Ninja Tip: Give the same small test project to 3 freelancers and see which works best for you. Most likely, 1 in 3 will work well. Once you get rolling with that freelancer, scale it up.
And if they need to work part-time, so be it. If they need to work as a contractor, so be it. If they need to work in their own time zone, so be it. Adapt a little. Hire people for their expertise. The rewards will be huge. Consider all 8 combinations:
- Local vs Global
- Full-time vs Part-time
- Employee vs Contractor
So that gives you 2³ = 8 combinations to work with. Use whatever combination you need to get the job done.
Lesson 7: It’s Fine If It’s Not Scalable — Initially
Not everything can be scalable from Day 1. It’s okay to learn. It’s okay to struggle a little through a non-scalable process.
And then later make it scalable.
For example: For the 1st 100 days, I answered each email personally with no helpdesk or customer support help. I went from semi-retired to working 18-hour days. Some customers thought I was a bot — LOL. Answering 300 tickets manually each day while still building the platform was painful.
Eventually, we put scalable processes in place — including using our own technology to improve CX and “ticket deflection”.
Tip: Always eat your own dog food!
Lesson 8. The Democratization Of Generative AI
It’s okay to throw out some old technologies, even though they might sound fancy, and see if the solution can be done just with generative AI.
This gives you a little edge in development because, with $50 and 1 API call, you could technically achieve something that would have cost millions to do before ChatGPT.
For example: See this case study that explored a cure for a rare disease with $50. Doing something similar in 2022 would have cost millions.
English is the new coding language.
Creating AI systems was a very developer-heavy, machine-learning-intensive, data-science-oriented exercise.
I learned this the painful way, having suffered through it in previous companies where simple AI projects would take an army of data scientists and months to complete.
Now, extremely complex AI systems can be done with prompt engineering and $50 with no-code systems or the OpenAI API.
Lesson 9: Give Before You Take
As a young entrepreneur, my mind always revolved around how I could get someone to help me with my startup.
And so, I would walk around looking at each person as someone I could fleece to help me with whatever I wanted. My mind always focused on “Hey — what can I learn from this person?” or “Oh wow — maybe I can sell my wares to this person.”
Every person I interacted with was a prospective customer, advisor, or employee. I learned catchy slogans like “Always Be Selling,” “Always Be Closing,” and “Always Be Hiring.”
This time around, I tried a different approach:
Give before you take.
Try to understand what the other person is going through or wants to achieve, and good things will happen after that. Ask yourself: How can you help the other person?
Thanks to this approach, the startup community around me put me on their shoulders and helped in all ways possible.
Lesson 10: ChatGPT Everywhere
In your interactions with your developers, contractors, and stakeholders, the expectation should be around hyper-optimization.
If something took 10X last year, it should take X this year.
And it should start with yourself.
Don’t just tell a developer what you want. First, have a chain-of-thought (COT) conversation with ChatGPT and lay out a clear spec.
Have clear conversations and specs in ChatGPT and educate yourself. Refine the spec. And do all of this in a ChatGPT conversation and Share that chat with the developer. Let them understand through the ChatGPT conversation what is happening in your mind.
You don’t have to create full specifications, like you did before in 2022. Have a 5-minute conversation with ChatGPT and then send that to the developer.
The conversation will help them get into your mind and explain what is needed. You could even generate 90% of the code so they have to finish up the last 10%.
And this is just with software development. The same sort of process can work for business strategy, marketing, generating blog posts, generating content, web pages, landing pages, etc.
For example: I just used ChatGPT to lay out my thought process around a negotiation I am currently involved in. In 2022, I would have called 3 different people to get similar results — today, it was done much more efficiently in 5 minutes.
Lesson 11: AI + HI Is The Magic Formula
Many people are trying to get AI to do 100 percent of the work. This will never happen. Instead:
Get AI to do 90 percent of the work and have smart human intelligence to finish the remaining 10%.
That is the magic formula.
For example: Don’t try to get AI to do 100% of a blog post. Instead, let it do 90% and let a smart, intelligent copywriter do the remaining 10%.
This is also true for programming, blog post writing, or data analysis.
This is true in literally every aspect of your business.
For example, we used to get 300 tickets a day to our customer support.
Then, we created our own custom GPT chatbot with ALL of our knowledge, and now almost 90% of customer issues are resolved by the bot itself. Only about 10% of real account-specific issues requiring human intervention come to the support desk.
Similarly, every aspect of the business, from analytics to marketing, should be hyper-optimized by AI.
I hope you enjoyed these lessons from the field. A lot of my line of thinking when it comes to startups is derived from the book “On Entrepreneurship and Impact” by Desh Deshpande.
Wishing you the best of luck in your entrepreneurial journey.
