What is the easiest way forward to interact with structured organizational data through Copilot?
Is the anwser the same for Microsoft 365 Copilot and autonomous (small c) copilots built w/Copilot Studio?
Copilot Studio converts everything to text since it is leveraging Large Language Model to formulate its answers, it is not meant to reason over lots of data and numbers to create prediction / do math. You can sometimes use an LLM to do reasoning over numbers but it is not meant to be.
One option is to extract the data using power automate and manipulate the data to make it more likely compatible with LLM. Here is an example I built to search through a database.
There is an update coming in AI plugin to be able to inject Dataverse dataset directly in Copilot Studio (just like in Power App) but it is not ready yet - no ETA for now.
follow
But even if copilot can access excel files on sharepoint, it doesn't always provide a good answer because copilot convert the Excel into a text file. Can someone provide advice about how to load structured data into copilot? I read somewhere that source data need to be loaded to Power App before being consumed by copilot. Any help on this will be appreciated.
Hi @CharlesS
Depends where is your data - and what kind of data.
M365 Copilot has access to Microsoft data: emails, Sharepoint, Onedrive, etc.
If you want to connect a Copilot to other structured data that is sitting outside of Microsoft ecosystem you will need Copilot Studio (ie: you have orders in SAP that you want to retrieve), and in that scenario (calling an API) you can leverage connectors and Power Automate flows to extract structured data and use it in your Copilot.
Note that you can extend M365 Copilot with Copilot Studio if you need to access those external structured data (so the answer is the same for both Copilot: use connector and flows to retrieve structured data)
Romain The Low-Code...
76
Pablo Roldan
61
stampcoin
10