Generative AI seems to be a hot topic, but I’m having a hard time finding good examples of how to build a SaaS off of it

jeniforjesus

New member
Hey all,

When I look through forums and videos of generative AI and its business applications, I see a lot of low-hanging content meant for clicks.

e.g. How to build an app in 48 hours with AI and no code!
e.g. I asked AI to market my business for me!

I’m an engineer/entrepreneur, so I am constantly on the look out for interesting business opportunities using generative AI models that are available off the shelf and combining that with software. AWS Bedrock has piqued my interest, given you can perform prompt engineering and consume the responses via API.

I’m having a hard time seeing how to provide value to consumers that can’t be provided by the free tier of ChatGPT. It often feels like you’re a middle man of prompt engineering, taking something and structuring it in a way that can be better consumed by AI and returning the output. This might be useful for things like chat bots or search queries, but again, I’m struggling to see anything deeper than that.

Does anyone have any resources they’ve found that show examples of really interesting SaaS businesses that leverage AI?
 
@jeniforjesus
It often feels like you’re a middle man of prompt engineering, taking something and structuring it in a way that can be better consumed by AI and returning the output.

That's because that's about all the business application is for publicly available generative AI models. You don't control the model, or access to it, so the best you can do is create a helpful tool to accomplish a specific task using one of them.
 
@matt55 yea if you're just using a public pre-trained model you're just designing a prompt that works for the business case. Theres lots of examples these days that are basically doing this. The ones that do more than just 'wrap up' a model and actually add real value will survive long term. I think we'll see more curated models coming into play though, it will become a key way to differentiate. Like 'is my model better than yours.'
 
@jeniforjesus Disclaimer: I’m the founder of treescale.com

The best thing you can do is to experiment with no-code tools to make product samples and see what works. You can even ask Google Bard to generate Prompt ideas 🌝
The SaaS model is tricky for things you have to pay “per request basis,” but often, if the product has value, OpenAI will charge far less than your users will pay. One of our customers who built the Job Description generator micro SaaS charges monthly, but of course, he yields to OpenAI per generated token.
 
@jeniforjesus There are so many ways of using generative AI to build a SaaS. The issue is that many people gravitate mostly to LLM because of the availability of already trained models and building models from scratch is intensive in terms of data, capital, compute etc.

If you want to build, you can check out those in the computer vision space.
 
@jeniforjesus I think the best options will be ones that combine multiple sources of data with a prompt to an LLM. This data can be freely acquired online or be data that the user systematically gives you without thinking and you package nicely for the LLM.
 
Back
Top