Understanding Copilot Studio Knowledge and the Importance of Indexing

[@vasilismirtsekis6148]

This is a beautiful explanation as always, thanks! Through trial and error, it seems our best responses come when we use pdf files that use correct markdown, headings, table of contents, etc, uploaded directly, so this makes a lot of sense!

[@martindreyer7213]

Great explanation. I have comment on this: “Its why you will see when you upload a PDF file to Copilot Studio directly as a file and it stores it in Dataverse, and it indexes it, vectorises it, and chunks it, you get a much higher quality response than if you take a PDF file that's in Sharepoint right now because currently SharePoint is using the Graph, which does not vectorize or chunk the file, it just indexes the file.” I believe that if you have an M365 Copilot license, Generative Answers uses the M365 vector index but otherwise falls back to the standard SharePoint text index, hence the poor results..

[@dibbers]

Was recommended to sub to your channel from Matt R (from Microsoft), I've watched several of your videos and plan on watching the rest this weekend. Great stuff, keep it coming!!!

[@FarhanAhmed-em4jf]

Awesome.This was some helpful information.

[@minermike1394]

Great video, thanks. Some things I have wondered about with copilot studio is if using company specific knowledge located on SharePoint to generate answers, how can the data be chunked to help the AI return accurate answers more often? Also, does the AI model learn with more usage of the company specific data,and if yes, then do certain actions cause the learning? For example clicking a thumbs up button, not abandoning the session, clicking yes if a chat bot ask "did I answer your question"? Lastly, if the AI model does learn over time , based on a doc located on SharePoint, does the learning start all over if the doc is updated and replaced with a new version? If you are able to cover any of this in a future video it would be appreciated. Thank you.

[@brandonvelasquez3530]

Make sure if you're going to put the bot on teams you follow the documentation and set the authentication to manual.... it's so stupid that manual authentication js needed for sharepoint as a knowledge source even when bot is hosted on Teams but it's a big problem i ran into.

[@kumarswamyp4568]

Hi Sir, I'm facing a challenge with using Adaptive Cards in custom topics within Copilot Studio. I'm designing an Adaptive Card where users need to enter organizational email addresses. I want to implement a feature that provides email suggestions as users start typing. However, I’ve been stuck on this for the past two weeks. Could you please assist me with this?

[@efrauniatlantico]

Good examples and easy to learn

[@nerdintech]

Hello! Great video, thanks! I am having some issues with Microsoft Graph connector inside copilot, witch the status always stay in "unknow", do you have any idea how to solve this? Thank you!

[@vvvvv432]

Hello Dewain! Nice Video! Question: Do you know where the info from the knowledge that we add in copilot studio is stored? And how could we check that the information from eg. sharepoint has indeed been included?

[@ACatCalledSnow]

Is there a way for knowledge sources to update automatically? So, if it's an Excel file that gets updated by users does it need to be manually updated in Copilot Studio each time to get the latest data?

[@cubaseproducer]

Hi Dewain. Thanks for your content because is amazing. Many useful topics. I have the copilot license in the tenant and also assigned to me as maker with these security roles in powerplatform admin center (System adminsitrator, environment maker and basic user) but I am unable to see the enterprise connectors when clicking on add knowledge. Is empty. Only see Public web site, SharePoint and dataverse. DO you know what can be happening? Many thanks

[@isaacsmith3861]

Great video, can you tell me where to get the configuration ID for bing custom search? and how to get to the bing custom search configuration page?

[@microsoftbizappsfico]

That's helpful! Many thanks

[@keithschaub7863]

Hi Dewain, same question about 'knowledge sources'. Very simple question...Can there be graphs, tables, equations embedded in the PDF that gets uploaded as 'knowledge'. If not, why not, or how to address it. If so, then please create a video that focuses just on that aspect. Companies have tons of material to use as data sources, but do we need to clean it up, or can we simply load the PDFs as is. How to understand the indexing quality of PDF's with additional materials like graphs, tables, etc.

[@trevoranderson1604]

I'm trying to create a custom copilot that reads an Excel form on sharepoint but it's not going well. It's not reading the date I'm putting in the file. Any suggestions for this?

[@kannankarmegam-bne]

Very useful insights

[@lbh3d]

How much does Bing Custom Search cost to use in this way? In terms of pricing what is considered a transaction?

[@islamshaban216]

Really helpful

[@keithschaub7863]

if I have 100 PDF files, each is 10MB, can I upload them all? What is the limitation? Their website does a horrible job of explaining the limitations