Can I Download and Deploy a Custom AI Model Trained in Power Automate AI Builder?

Question

I'm currently working on a project where I need to extract specific information from a large number of PDF documents—such as locations, addresses, machine types, etc. I’m using Power Automate AI Builder, which lets me label hundreds of PDFs and create collections for different document layouts. It’s a great tool for training a model to handle these tasks, but it seems like I can't download the trained model or deploy it outside of the Microsoft ecosystem (e.g., using it with Python or other platforms).

The problem is, I need a model that can handle processing thousands of PDF files with varying layouts. The goal is to have a model that I can integrate with Python, and potentially, in the future, embed into a website or other applications. I'm aware of other tools like Google Document AI and Azure, but I’m wondering if these or other platforms allow the same kind of training process while also giving me the ability to download and use the model externally. Unfortunately, I haven’t been able to try out these tools or alternatives because most of them require payment before you can experiment with them.

Does anyone know if Google Document AI, Azure, or any other platforms offer this capability? Or any suggestions for other tools that might meet these needs? Any recommendations or insights would be greatly appreciated!

tried = Power automate labeling and training process /// expecting = being able to download the model

Answer 1

If I was trying to build this real quick for a website & many document formats I’d set up a flow with a When an HTTP request is received trigger & have the response be the output JSON object. And I’d switch from AI Builder models to OCR & GPT like this template: https://community.powerplatform.com/galleries/gallery-posts/?postid=31e67eea-3f73-47b4-95b7-fe4a7b646389

Answer 2

There are many ways to solve this! The functionality of Power Automate is quite strong. While there are some companies offering the same service, I think this is a great starting point to get some work done.

Azure, as a Microsoft platform, offers the functionality you describe. It's different ofcourse, but provides the functionality you need. Link You can use this to connect to Python, and scale for a potential website.

You can also start from scratch using OCR (optical character recognition) and build a script in Python along with an OpenAI API connection.

You can check these resources for inspiration: code camp, PyMupdf, medium and docsumo