Need Some Guidance on Knowledge Sources and Conversation Orchestration

I am creating an agent for our HR department to address the most common questions they receive from employees (how to take time off, what's our insurance coverage, etc). We operate in 3 countries and each country has it's own dedicated knowledge sources (example: Insurance Benefits Booklet for Canada, Insurance Benefit Booklet for US).

Here's what I have done:

On Conversation Start, I get the agent to grab the Country/Region entity from the user.
I have explicitly named each knowledge source as being "for Canada Employees" or "for US Employees".
Each knowledge source is clear and well explained in the description (I used copilot to generate a synopsis of the knowledge sources which I then used as the descriptions).
I have enabled Generative AI

Doing this has worked, for example, if I say I am from the US and then ask for the statutory holidays, the agent gives me a list of the US holidays. However, sometimes it gives me responses from knowledge sources from different countries even though there is relevant information in the knowledge source for all countries.

What is the best approach to take to ensure that information returned is coming from appropriate knowledge sources linked to the user's country selection?

My first thought was to use Topics, however, I feel like I would have to create a node for each knowledge source / each country. Am I incorrect in that assumption?

I read somewhere that it might be possible to create an Agent for each country and then hand off the conversation to the appropriate agent after getting the country from the user.

Lastly, I read that it may be possible to create additional "classic" bots and set them as skills within Copilot studio, however, this in not available under our current licensing.

Any help or guidance is greatly appreciated!

Categories:

General

Quick Links

Leaderboard > Copilot Studio - General