Forum Discussion
Making CoPilot Work for our Organisation
- Oct 21, 2025
The structure for data access, referred to as "the brain," is defined within the orchestration engine. Currently, the engine in Copilot Studio does not perform optimally when retrieving data from sources such as SharePoint. Interestingly, the orchestration engine in Copilot Studio Lite occasionally delivers better results in this regard.
Regarding the notion that Copilot reviews the first 20 relevant documents, it is important to clarify: AI is not designed to return a fixed number of documents that may be loosely related to a query. The purpose of using an agent with document integration is to ensure that high-quality backend data enables the AI to extract precise and relevant information. If the expectation is to retrieve 20 documents via AI, it must be stated clearly that this is not a suitable use case and will not yield effective results.
Additionally, generating branded documents is currently beyond the capabilities of AI. However, modern templates may offer a viable alternative in such scenarios.
Modern Templates on SharePoint
Appreciate you commenting on here Peter, really useful comments, understand points 1-3. Just a few follow up questions based on what you have said for point 4:
- Regarding point 4, so I did actually work on making a dedicated bid assistant bot which was accessible via a Teams chat. The one stumbling block I have, is the way it's accessing the information. I'm not 100% certain on CoPilot's brain, in how it actually finds the information based off what you ask it. Is it just looking at the first 20 docs it thinks are relevant or does it already know what is relevant/not relevant and searches everything.
- Also, in terms of producing documents in branded formats with the right font, colours etc. I've done some research on this and I can't find any examples of this actually working and giving you back a document just based off a prompt.
The structure for data access, referred to as "the brain," is defined within the orchestration engine. Currently, the engine in Copilot Studio does not perform optimally when retrieving data from sources such as SharePoint. Interestingly, the orchestration engine in Copilot Studio Lite occasionally delivers better results in this regard.
Regarding the notion that Copilot reviews the first 20 relevant documents, it is important to clarify: AI is not designed to return a fixed number of documents that may be loosely related to a query. The purpose of using an agent with document integration is to ensure that high-quality backend data enables the AI to extract precise and relevant information. If the expectation is to retrieve 20 documents via AI, it must be stated clearly that this is not a suitable use case and will not yield effective results.
Additionally, generating branded documents is currently beyond the capabilities of AI. However, modern templates may offer a viable alternative in such scenarios.
Modern Templates on SharePoint