Forum Discussion
Copilot hallucinations when in smaller containers?
I read a lot about the general disappointment in Microsoft for not providing Copilot to smaller organizations and having a minimum seat requirement. I fully understand these arguments and reactions, since Copilot is something that we have anticipated for quite some time (or at least a year).
I'm just thinking - will smaller businesses benefit as greatly from Copilot as bigger ones? We know that LLMs are only as good as our data input and that models which are not trained on enough data will start to hallucinate. Do we see a "recommended" minimum amount of exisiting data in an organizations M365-tenant to secure that Copilot does not start to hallucinate? Will smaller businesses ever reach those quantities of data so that this will not occur?
Hi MichaelaBostrand,
I agree that the quality of Copilot's output is directly proportional to the amount of data available to it. This holds to the size of the businesses. One key aspect here is the quality and value of the data involved.
MichaelaBostrand and to all concerned,
Please, let me know if it makes sense and I encourage open discussion (this helps me to learn.)
Following the establishment of the Center of Excellence (CoE), I propose a potential deployment strategy for Copilot, divided into two phases:
Phase 1: Initial deployment of Copilot for Microsoft 365, with restricted access to data sources such as SharePoint or OneDrive. This phase allows users to receive assistance in creating Word, Excel, or PowerPoint documents, and in summarizing and composing emails.
Phase 2: After data preparation (which includes thorough cleaning), expanded access should be granted.
This two phases approach, although not my preferred one, seems practical considering the time required for data preparation.
Note: Data preparation might be an ongoing process, especially with the integration of additional sources like ERP data through plugins.
Your thoughts and insights on this plan would be greatly appreciated.
Kind regards,
Hi MichaelaBostrand,
I agree that the quality of Copilot's output is directly proportional to the amount of data available to it. This holds to the size of the businesses. One key aspect here is the quality and value of the data involved.
MichaelaBostrand and to all concerned,
Please, let me know if it makes sense and I encourage open discussion (this helps me to learn.)
Following the establishment of the Center of Excellence (CoE), I propose a potential deployment strategy for Copilot, divided into two phases:
Phase 1: Initial deployment of Copilot for Microsoft 365, with restricted access to data sources such as SharePoint or OneDrive. This phase allows users to receive assistance in creating Word, Excel, or PowerPoint documents, and in summarizing and composing emails.
Phase 2: After data preparation (which includes thorough cleaning), expanded access should be granted.
This two phases approach, although not my preferred one, seems practical considering the time required for data preparation.
Note: Data preparation might be an ongoing process, especially with the integration of additional sources like ERP data through plugins.
Your thoughts and insights on this plan would be greatly appreciated.
Kind regards,
- I agree that data quality is of essence here!
I like your approach, it seems reasonable to get access to Copilot whilst getting on with the clean up of existing data! It will be almost impossible to make sure the entire environment is cleaned and ready before launching Copilot and that will most likely create frustration in the organisation.
I guess this time is as good as any for smaller businesses in the wait for Microsoft to decrease the limits for Copilot, to already now get started on managing the existing data in their tenants, reviewing data access and data quality, so that when to time comes, the step will be a lot smaller and easier!- Maybe it will be possible to use Copilot to assist the administrators and the super users in data clean up and security check !
