Please Read Before Building AI Tools on OpenAI's API
Avoid building a 'Wrapper Product.' To make money with AI tools, you need to offer more than just a better prompt engine.
Nothing has transformed the world of technology as much as artificial intelligence, except perhaps the internet itself. Every day, hundreds of new AI tools emerge, and the use cases become increasingly creative.
It's a bit like the time when Apple introduced the App Store, and suddenly, there was an app for everything. What was technically possible was built, regardless of whether anyone needed it.
Similar to that time, the current AI trend wave has created a massive gold rush. However, one difference is that the barrier to building AI-powered products has significantly lowered.
Naturally, many entrepreneurs are now diving into this field, desperately seeking ways to get their share of the pie.
And it works!
In broad terms, there are two main business models for making money with AI:
AI-related services (e.g., consulting, courses, events)
AI Tools (SaaS)
In this article, I primarily want to focus on what should be considered when developing AI SaaS Tools.
Most Product Builders use OpenAI’s API
AI technology is evolving so rapidly that the challenge is not the technical possibilities but the speed at which solutions become obsolete.
If you're developing a SaaS solution today and want to leverage AI, you're likely using OpenAI's API, the company behind ChatGPT. This isn't surprising because OpenAI not only offers one of the most advanced AI language models but is also very developer-friendly.
The excitement for developers was evident in the advancements OpenAI introduced on November 6, 2023, which are truly mind-blowing and make every developer's heart beat faster.
Now, through the API, you can request not only simple text prompts but also images (Dall-E), voice (Whisper), and document processing. Thanks to the new Assistants API, it's even easier, even for non-developers.
Even if you're not a developer, you can use no-code/low-code UI builders like Bubble, WeWeb, Webflow, and others to create complex applications in simple interfaces.
These tools either offer their own integration with OpenAI's AI capabilities, or you can easily connect them using Integration Platforms as a Service (IPaaS) like Make or Zapier.
The Problem with "Wrapper Products"
Imagine you've come up with a great use case for your AI tool and have built a solution that fills gaps in the existing use of ChatGPT and similar models.
For example, you've developed an interface that helps users operate AI optimally "as if by magic" without having to learn and create prompts themselves. TechCrunch calls these "Wrapper Startups" or "Wrapper Products."
The aim is to make the experience as easy as possible for users with specific use cases, without them needing to get too technical.
A good example is creating SEO blog posts with AI tools. For a long time, ChatGPT couldn't generate long texts. The AI's “attention span” was too short, and the output per prompt was too limited.
With that in mind, AI tool developers have created solutions that take on the task of breaking down a large content pieces (such as a 1500-word SEO text) into smaller packages. The data may be cached, processed with AI, and then reassembled into a final text.
AI tools for text generation, for instance, create an outline first, which is gradually filled with paragraphs. It's feasible from a technical standpoint, even though I have my doubts about this type of content creation.
The recent OpenAI Developer Conference demonstrated how quickly solutions like these can become obsolete. With the new GPT-4 Turbo model with 128K context, ChatGPT can now process texts in novel lengths.
Another example is the ability for anyone to build their own GPTs. The many AI chatbots created in recent months may now face serious challenges.
What are Sustainable AI Business Models?
It's challenging. Not only do AI courses, lectures, blog posts, or videos on the subject have a very short shelf-life, but AI tools can be overtaken by development and quickly become redundant.
For a model to be sustainable, the core of the product should not rely solely on the performance of OpenAI or another third-party tool.
There's nothing wrong with using AI, but OpenAI should provide excellent opportunities to automate tasks and, most importantly, reduce human effort.
This particularly applies to aspects of customer support, marketing, and more. AI can be used to simplify customer communication, but should not be the only reason customers are there. They shouldn't be there just because of the "AI Magic."
Because as soon as the task can be done better and more affordably elsewhere, the model will collapse.
In Conclusion: It Takes More Than Just a Better Prompt Engine
For "Wrapper Startups," things will become challenging in the future.
What is perceived as an innovation from a customer's perspective today can be unspectacular tomorrow. Therefore, the focus should be on a tangible overall solution for customers that does more than what could have been done with a few extra steps using ChatGPT alone.
A good example of this is Notion*. The tool stands out because it excels at organizing knowledge. AI (through the OpenAI API) is present in Notion but needs to be actively integrated by the user. This allows content to be improved, for instance, through AI with a right-click (e.g., translation, summarization, etc.).
Another example is Canva*. While AI can create images, Canva can place them in the right context and process them, such as for social media (including sharing), in documents, and more. If we were to evaluate the AI functions in Canva alone, it wouldn't be particularly exciting.
If you know of any more positive or negative examples, please feel free to leave a comment.
*=Affiliate Links