Running an entire SaaS Company on AI

Running an entire SaaS Company on AI

Most people think about us as a SaaS company. Technically we're not, we're actually a PaaS company. However, since 95% of our customers just wants "an AI chatbot", SaaS is probably how most people think about us. In case you've got no idea what the difference is, a PaaS company delivers platforms as a service, while a SaaS company delivers software as a service. Software is something you use while platforms are something you build on top of.

Creating Personal Assistants

Today we released a custom GPT. It's an internal GPT, and we won't give you access to it. It's the beginning of porting our entire back office infrastructure, making it possible for us to manage and administrate our entire company using ChatGPT.

This custom GPT allows us to do the following.

  • Create payment links and send to clients
  • Create cloudlets and deploy into our Kubernetes cluster
  • Manage cloudlets and plans
  • Etc, etc, etc.

We used Magic of course while building it, and we spent roughly 10 hours of coding the things we're demonstrating in the video below.

One software developer built the above in 10 hours using Low-Code, No-Code, and Magic Workflows. Pretty cool if you ask me. 2 years ago the above would have been considered science fiction. Today we released it after 10 hours of development. From a resource perspective the above will optimise the way we're working internally at AINIRO.IO 2x, possibly more over time.

Having something like the above is literally like having a personal assistant, always waiting for your commands, for then to execute your commands, whatever they are. I don't know what personal assistants are charging these days, but with AINIRO.IO you can get something like the above starting out at $198 per month, if you're willing to do the work yourself. In addition it's a $25 per month fee at OpenAI to create a team account.

Got some body to spare mate?

The above is obviously a bit of a joke. However, ChatGPT is amazing, but it's almost useless without the ability to do stuff. ChatGPT without GPTs allows you to create amazing support chatbots and produce content. It can create amazingly intelligent answers to questions. It solves real world problems. But it will always in be dependent upon somebody else actually doing whatever it tells you to do.

ChatGPT without GPTs is like a brain without a body

If you connect it to a GPT though, it can all of a sudden do stuff. When I ask it to create a cloudlet in the above video, it runs through the following sequence of events.

  • Creates a volume claim inside our Kubernetes cluster
  • Create a service, an ingress, and a deployment in K8S
  • Waits for the deployment to be ready, before it logs in
  • Configures the cloudlet, a process which implies creating a database and several scheduled tasks
  • Inserts a database record into our internal back office database
  • And depending upon whether or not it's a demo cloudlet, creates a scheduled task that deletes the cloudlet after 7 days
  • Finally, it sends an email to the user we're creating the cloudlet for

Basically, ChatGPT "creates a server" for us, 100% automatically, and sends an email to the user we're creating the server on behalf of. Without the ability to connect ChatGPT to a GPT, it would at best give us the recipe for how to do it ourselves, and there's no way it would be able to actually perform the task itself.

Creating a cloudlet with Magic and a custom GPT

AI-based workflows

The above is what we refer to as "an AI-based workflow". The AI has access to an API, allowing it to reach out into other servers, and actually do stuff. What stuff it can do is up to you to decide, and depends upon your use cases. However, you could create an assistant that sends emails, creates and publishes blogs, shares links on Twitter, etc.

Your fantasy is the only restriction here

One advantage with such AI-based workflows is that we can completely drop our entire frontend codebase. Frontend is roughly 70% of technical debt in any organisation. Drop the frontend and you've instantly paid 70% of your technical debt. Frontend developers are also typically more expensive to hire, and it's a more complex domain, requiring skills related to design, estethics, etc.

If you can drop frontend development, you've easily saved 70% of your software development costs

With an AI-based workflow you can drop everything related to frontend development and save 70% of your costs. Due to the conversational interface ChatGPT provides you with, you no longer need buttons, checkboxes, drop down lists, or radio buttons. If you've got multiple choices, the AI will simply ask you to pick one - And if you don't know how to use it, you can ask it to explain itself.

Asking our Custom GPT what we can do here

Connecting your GPTs to the world

If an AI without the ability to do anything is the equivalent of a brain without a body, then a custom GPT without an API is like a brain with a body in prison. Custom GPTs are basically a useless toy without an API.

With Magic Cloud you can create an API in seconds. This allows you to query your database, invoke actions in other servers, and execute tasks any ways you want for it to execute said tasks. Below is an example of something that reaches out into our servers and returns the different cloudlet plans we've got in AINIRO.IO.

Custom GPT listing cloudlet plans

How many times did somebody ask you to "send us your different product plans, and we'll discuss it internally"? If I wanted to I could tell our custom GPT to send the above plans to some email address, coupled with purchasing links, and it would instantly save me 5 minutes of work. 5 minutes isn't a lot, but if each person saves 5 minutes of work on 10 different tasks every single day, and there are 50 employees in your company, this becomes a dramatic number over time.

With ChatGPT you've arguably been "its assistant" for a year now, since it could only give you instructions, and it was dependent upon you actually executing these instructions. With custom GPTs, it effectively becomes "your assistant", and you can tell it what you want it to do.

The changes in regards to usefulness is dramatic!

If you're interested in hearing more about custom GPTs and Magic, you can send us an email below. At least for now, it will be answered by a human being. A couple of months down the road though I'm not so sure, because this changes the game entirely!

Thomas Hansen

Thomas Hansen I am the CTO of AINIRO.IO AS. I am a software developer with more than 25 years of experience. I write about Machine Learning, AI, and how to help organizations adopt said technologies. You can follow me on LinkedIn if you want to read more of what I write.

Published 23. Jan 2024