Here’s a list of use cases for using ChatGTP with Power Platform. I’m updating this list whenever I find something interesting/useful. Feel free to share your ideas in the comments box below.
Power BI & ChatGTP
- Get help on how to write DAX
- Get help on how to write M queries
- Get help on how to use the Power BI Embedded
Model-Driven Apps / Dataverse & ChatGPT
- Generate sample data for Dataverse
- Use ChatGPT to teach you how to create plugins
- Use it to help with JavaScript
Power Automate & ChatGPT
- Guide on how to connect to the ChatGPT API with Power Automate
This shows the basic steps that you need to do to start building your own solutions: https://www.c-sharpcorner.com/blogs/implementation-of-chatgpt-in-power-automate
Or you can use the Independent Connector available:
- Use it to generate Power Automate Expressions:
Power Pages / Portals and ChatGPT
- Get guidance on how to start extending the portal functionality with a custom code
Canvas Apps and ChatGPT
- Get help writing PowerFX:
- ChatGPT helps you to get started with creating a new Canvas App
Power Automate Desktop & ChatGPT
- Generate PowerShell scripts for your automations
I’m yet to come up with a good idea for ChatGPT to be used with Power Automate Desktop, but hey, at least you can use it to generate PowerShell scripts that can be then executed through PAD:
- Link to a guide that explains how to technically connect to ChatGPT API with PAD https://powerusers.microsoft.com/t5/Power-Automate-Cookbook/Integrating-ChatGPT-with-Power-Automate-Desktop/m-p/1980676
And last but not least…
Here’s what ChatGPT says about the subject:
This is interesting. What I found interesting was that the questions I asked in the exact words that you had asked in gave me a different response. Wonder if that is because of history where you asked it to generate PowerFx and then a Power Automate and then time logging and whether it was because of that?
I’ve seen this happening too and it does sometimes take a bit of time to get the prompt “right” to give the expected result. It is aware of the context of the conversation, so that might also affect it. For example, generating the sample rows seem to work quite randomly, quite often it just prompts the first 3 rows, but it can be “forced” to generate more by asking 😀 Brilliant stuff nevertheless, I need to start playing more with the API to see if that can stay more consistent with the prompts & responses.