Thinking of building an AI startup? Read this first.
Building an AI startup is the hot new thing right now. Most will fizzle out once the hype starts to die down, but yours doesn't have to. If you understand these strategies you can position your product to win long term.
First, it's important to understand where most people are building: on top of OpenAI's model without model differentiation. It's amazing how easily accessible they made it to entrepreneurs. The downside is all of these products are easily replicated and don't have a moat.
No moat? No problem. Time to start digging. How can you establish a moat for your product while not building your own AI model? There are four primary strategies:
- Build a superior brand that users love and trust.
- Have superior distribution. The ability to reach more customers you can also help you bolster your brand.
- Build proprietary tech around the AI. UX matters. To differentiate, you can also build logic around the AI rather than relying on it exclusively
- Establish strategic partnerships and integrations. These are difficult for competitors to replicate and can help you on the distribution front as well.
I go deeper on this topic on my latest newsletter article post which you may find helpful:
https://www.bentherebuildingthat.com/p/how-to-build-a-winning-business-in
Love this approach, and it speaks to how SaaS was already changing prior to the mass adoption of AI. So much of the actual functionality of most apps has been commoditized (and has now basically been completely commoditized by AI). What sets products apart are experience and community.
One thing I'd add, if you can help users imagine what's possible, it's an opportunity to distinguish yourself. We've had customers say "I already use ChatGPT for this" until we showed them an actual workflow difference. Copy/paste vs. seamless integration. Better context from their previous sent messages we have that these off the shelf models don't.
Excellent post! For me, with https://evoke-app.com/, we're luckily not building on top of any API, but rather the ones providing the API ourselves. Moat is much easier to build, but there are some competitors popping up.
Is going to take some time, and we're focused more on shipping right now, but we do have quite a large discord community: https://discord.gg/dXJtarPsCm
We've been working on. Really trying to get our API in more courses for building AI apps for distribution
We're focused now on mainly hosting existing open source AI models, but we do make some optimizations for them to make them run cheaper and faster.
Same with 2. Really trying to establish partnerships with course makers, API aggregation programs, or other AI app builder programs.
Nice! Seems like you're well positioned with Evoke :)
Great post! i've been building an OpenAi/Google competitor believe it or not, https://text-generator.io its 8x cheaper than Google at speech to text, and competes well technically and on cost with OpenAI for text generation too, but getting the message out can be really tough.
#1 is hard for me and it's not a better brand honestly me being not a designer by trade... Startups can definitely care more and people appreciate people/founders who talk directly to them which helps.
#2 is really hard for the small guys, i just post on HN/reddit ml/here/producthunt etc to launch
#3 I first made it compatible so switching was easy, now added things OpenAI doesn't do like automatically download and analyse links/analyse images to generate better text.
#4 has been where actual money comes in via prompt engineering for enterprise customers so far, but those relationships can be delicate so i need a better way to cultivate and help grow those than just having an API/marketplace
Sounds like #4 is how you'll financially float it until #3 is at a point where it gives you an exclusive advantage. The exclusive advantage can then play into building your brand and distribution.
Good luck, Lee!
Thank you for sharing, Ben. I checked out the newsletter article as well, and it was really interesting - especially the part about layers and why it's hard to win in layer 3. As you said, there is only so much fine-tuning a model can do for you. I found the additional strategies interesting and thought provking.
I'm glad you found it helpful, Maya! I appreciate you taking the time to read it :)
That's an excellent and insightful observation of the current state of AI startups. It's true that many companies are leveraging OpenAI's models without differentiation, making it easy for competitors to replicate their products. Your four strategies for establishing a moat and positioning your product to win in the long run are spot on. Building a strong brand, having superior distribution, adding proprietary technology around the AI, and establishing strategic partnerships can certainly give your startup a competitive edge. I'll definitely check out your latest newsletter article post, thanks for sharing!
Thanks for reading, Jun! Appreciate you checking out the newsletter, too!
I spent a little more time thinking about how I'd want to position myself if I were to launch another AI product. I wrote a super quick snippet and drew a diagram on how i'd approach it, which you might find interesting: https://www.indiehackers.com/post/want-to-make-money-with-ai-try-this-9c2a9d271c