Event details
- But CoPilot will have access to your Sharepoint and OneDrive files, correct?
This is what confuses me. Copilot supposedly learns from all enterprise data in M365 and give you intelligent responses if you have access to the relevant data. So, if I want Copilot trained to provide answers based on my Service Desk's knowledge base articles in SharePoint, I guess we don't have to go out of our way to configure that (?). I'm also curious if we can specifically train Copilot to provide particular answers to particular questions for instances where we want to be very specific with responses with a very low chance of hallucination.
- Ben_Summers
Microsoft
Cody - I'll add that if you want to bring additional datasets - say from your CRM tools - into Microsoft 365 Copilot you can build (or get) Microsoft Graph Connectors that will index that data in Microsoft Graph & Semantic Index. Your IT Admin will need to configure the connector, once built (or gotten). But this just increases the surface area in which our copilot is grounded.
However, if you want to keep things separate - and actually train AI on a specific dataset (sensitive or otherwise), the answer remains unchanged - you'll want to build it yourself (ideally using Azure OpenAI services).
- Ben_SummersCopilot uses access to SharePoint, OneDrive, etc. for grounding - that is, to provide information/answers relevant to the user's organization - but not for training. To Pat's previous point, you'd use Azure services to build your own model that is trained on your private datasets.
- So, if we want to train AI on a PRIVATE dataset that we don't want everyone to have access to (e.g., sensitive R&D data), we would use Azure AI services, but if the dataset is open to everyone in the company (e.g., Service Desk Knowledge Base articles), Copilot will automatically learn from it without IT admins needing to do anything - accurate?
