Expertly crafted and automated GPT-3 prompts in a familiar, form-based UI.
A case study that highlights my ability to quickly validate ideas in an unprecedented and rapidly evolving AI industry.
After hitting 300k+ views on a TikTok video, I began to wonder if I was on to something. The original concept was simple — I showed people how to connect their Notion account to GPT-3 for automatic analysis of startup ideas. With just a one-line description of their idea, users could receive a full analysis report within seconds.
However, even after creating a follow-up tutorial video, it was clear that while some could recreate the tool on their own, many found the setup to be too technical and complicated.
I quickly realized the need for a user-friendly and publicly available version of these types of tools. This way, anyone can easily automate their work with AI, regardless of their technical ability.
Adoption
Need for technical knowledge made it difficult to build and use AI automations.
Reusability
Copy pasting the same set of prompts with slight tweaks in ChatGPT was tedious.
Quality
Lack of awareness of the benefits of high-quality ChatGPT prompting techniques to avoid poor responses from the AI.
Optimize with AI offers access to high-quality, curated AI-powered tools that allow you to ask the right questions and get the right data. By using forms, you can receive better AI responses in a familiar, reusable, and guided format.
The beta testing waitlist approach was crucial in capturing potential user interest from a business perspective. The site was quickly released as an MVP following the viral success of a previous TikTok, with the main goal of gauging demand for AI products and using it as an experiment.
First, some context about the TikTok...
I started my TikTok account with the intention of sharing useful content and in return gathering feedback from my audience to understand their pain points. It wasn't until my Startup Analyzer video went viral that I truly realized the value of this approach.
While the high number of views was a clear indication that the video resonated with viewers, the comments were even more valuable in providing insights into their specific needs.
I categorized and prioritized the comments into key areas. I found that people wanted to leverage and automate with AI, but didn’t know how. Additionally, the setup process was too technical.
How can I make this more accessible?
To make the tool more accessible and user-friendly, I decided to replace my personal Notion with an actual website as the front-end for the startup tool.
I also opted to use forms as the main input format. I felt most people are familiar with forms, compared to Notion and ChatGPT, which may not be as intuitive for non-technical people. I was curious to see if this change in UX would make a difference as well.
Beyond the startup analyzer
To test whether people wanted these types of tools, I added another sample tool for writing content scripts. I also included a "coming soon" section with descriptions of future tools. The goal was to create a directory of tools that can automate various aspects of both personal and professional life, as well as entire businesses.
Once I settled on the concept, I got to work. My goal was to use the momentum from previous videos to build up the launch of my site. This meant I had to work quickly.
Planning out the design and branding
To accomplish this, I designed the site in Figma to get a rough idea of the necessary structure and design elements.
No code? No problem!
I turned the Figma mockup into a functional website using Webflow (for the front-end) and Zapier (for the back-end), two no-code tools:
- Leveraging no-code helped me achieve much faster time-to-market, a main goal of mine
- I had used both tools already, so there was no learning curve
- Webflow has authentication built-in, which saved a lot of time
- However, one of the tradeoffs of using no-code is that I couldn't display the AI responses on the website itself, so I chose to send them via email
- Zapier handled the connection with GPT-3 — sending a new request with each form submission
- Additionally, Zapier helped me automate the process of sending emails and capturing form submissions
- The key was to tie the input fields of the form to the variables inside the prompt, making it reusable and template-driven
It's amazing that from idea to deployment, it took only 3 days...a truly scrappy MVP!
Not quite what I expected
To prepare for the launch, I created a TikTok follow-up video showcasing the site, set up Google Analytics, and organized my monitors to track all the stats and submissions that would pour in. I was ready. However, I was met with disappointing results. Specifically, people would sign up but not use the form. And if they did use the form, it was typically only once.
But wait!
Luckily, I had thought ahead and decided to use a waitlist sign-up form to manage users in case I went viral again. This ended up being the most valuable part of my entire experiment.
I received 170 responses from people interested in my site and AI, which they had to tell me about to gain access. From this, I gathered a completely new set of rich data to form new hypotheses, and I am looking forward to conducting new experiments based on all this information.
Template driven — Users can fill out a form and receive an AI response in seconds.
Template driven — Users can fill out a form and receive an AI response in seconds.
Quality outputs — Well-crafted prompts leading to higher quality results.
Quality outputs — Well-crafted prompts leading to higher quality results.
Waitlist — Easily capture feedback and insights from potential users.
Waitlist — Easily capture feedback and insights from potential users.
