Strategies for building in the AI space
If you use Twitter or are in the tech space, you’ve probably seen countless AI demos and prototypes using generative AI technologies like GPT-4. Many of them are super impressive and are being built at hackathons, weekend projects, or colleges. You’ve probably also seen an onslaught of AI announcements practically daily that are shaking up the space. With all these products and so much change, the tricky question right now is, what kind of AI product will have any amount of staying power and be valuable businesses and startups?
There is an argument surfacing that building on top of companies like OpenAI’s models is a bad idea because there is very little defensibility or a “moat”. While true, the truth of the matter is that very little software these days have any defensibility at all. I also thought that Guillermo Rauch said it well when he tweeted the following:
Assuming there will be many valuable companies in the AI space — a rather safe assumption, here are some thoughts on how to think about building in the AI space.
Think bootstrap, not startup
Because everything in the technology world is changing so fast, it makes sense to think from a bootstrapper (or traditional business) perspective over a startup perspective right now. OpenAI, in particular, is launching new and innovative products regularly, and many of the products, like GPT-4 or ChatGPT plugins, are transformational changes. Because things are changing so fast, committing to a startup idea for the long term and raising money is risky.
If you think from a bootstrapper’s perspective, you don’t need to commit to a certain idea for the long haul, and you don’t need to plan to raise money — you just need to commit to launching something, charging for it, and finding out quickly if it works.
If you start bringing in revenue and there is clearly something valuable there, you can always raise money later on.
Think of an Idea you can charge for
This is much harder than it seems. For someone to take out their credit card and agree to give you money, you really need to provide value. Every purchasing decision fits one of Maslow’s hierarchy of needs:
It’s essential to come up with an idea that satisfies one of these needs and is closest to the bottom of the pyramid as possible (these are the goods and services most needed by humans).
Try to focus on one core feature that solves a problem for the user that fits one of these needs.
Build a fully-fledged product with payments
After picking an idea that you can charge for, it’s essential to go out and build a first version with payments built in. Doing this will allow you to validate your idea quickly. Pieter Levels does this very well. Every product he launches is fully functioning, solves at least one problem, and has payments built in.
Prioritize sales and marketing
It’s not essential to have a technology moat if you have strong distribution and marketing. Before committing to an idea and building it, it’s important to have a well-thought-out sales and marketing plan that you can stick with.
Have a detailed plan and strategy for getting your first 100 paying users.
The plan should include things like Twitter automation, Google/Facebook ad spend, posting in public forums, and how you will execute various public launches on sites like Product hunt
Prioritize UI/UX
Having a great user experience, great product design, and a strong brand makes for really solid differentiators in the sea of AI products.
From a UX perspective, one recent example is the AI company FirstVoices. They built technology that allows you to have a voice conversation with famous celebrities. Rather than building a site, they decided to launch it as a telegram bot. It’s really easy to share a link to the bot and get started. Their team thought outside the box and came up with an easy and fast user experience. The same thing goes for Midjourney, whose product lives exclusively in Discord.