Genesys Cloud - Main

Hello everyone, how are you?

I hope you're well!

I would like to know, if you're open to sharing, the way you use Copilot's answer generation function. I'd appreciate some tips and suggestions on efficient ways to generate a response that can be used in customer support, as well as methods to structure a knowledge article to improve answer generation.

I have already tried to fragment the articles in little ones to try to get a ready-to-use answer for the agents, but it didn't work the way I was expecting.

#Copilot
#AIConfiguration
#AICopilot(Agent,SupervisorAdmin)
#AI

Hi Lucas, 

I can give you some tips on this. 

For article structure best practices, you should look at paragraph length and structure. Try and aim for semantically coherent paragraphs that are around 200 words long. This should allow for a more precise retrieval and accurate generation. I think your approach of fragmenting articles into smaller pieces may not have worked because its actually the opposite of the recommendation, instead of creating many tiny articles you should create a comprehensive articles with well structured, meaningful paragraphs. This includes writing full sentences as opposed to bullet points.

For knowledge base configuration, specifically content search knowledge I suggest removing all phrasings make articles only title and body contents. If the title isnt long enough to summarize the content appropriately try adding 3-5 phrases to fill the gaps. But use these phrases sparingly and evenly distributed across knowledge articles. For the AI-generated answers try setting the confidence threshold to  0.65 or below for all language. If a knowledge article supersedes an intent-based rule, try lowering the NLU confidence threshold. 

I hope these tips help. 

Hello everyone, how are you?

I hope you're well!

I would like to know, if you're open to sharing, the way you use Copilot's answer generation function. I'd appreciate some tips and suggestions on efficient ways to generate a response that can be used in customer support, as well as methods to structure a knowledge article to improve answer generation.

I have already tried to fragment the articles in little ones to try to get a ready-to-use answer for the agents, but it didn't work the way I was expecting.

#Copilot
#AIConfiguration
#AICopilot(Agent,SupervisorAdmin)
#AI

Hey Cameron,

Can you elaborate on your suggestion to eliminate all phrasings?  Seems counterintuitive

Hi Lucas, 

I can give you some tips on this. 

For article structure best practices, you should look at paragraph length and structure. Try and aim for semantically coherent paragraphs that are around 200 words long. This should allow for a more precise retrieval and accurate generation. I think your approach of fragmenting articles into smaller pieces may not have worked because its actually the opposite of the recommendation, instead of creating many tiny articles you should create a comprehensive articles with well structured, meaningful paragraphs. This includes writing full sentences as opposed to bullet points.

For knowledge base configuration, specifically content search knowledge I suggest removing all phrasings make articles only title and body contents. If the title isnt long enough to summarize the content appropriately try adding 3-5 phrases to fill the gaps. But use these phrases sparingly and evenly distributed across knowledge articles. For the AI-generated answers try setting the confidence threshold to  0.65 or below for all language. If a knowledge article supersedes an intent-based rule, try lowering the NLU confidence threshold. 

I hope these tips help. 

Hello everyone, how are you?

I hope you're well!

I would like to know, if you're open to sharing, the way you use Copilot's answer generation function. I'd appreciate some tips and suggestions on efficient ways to generate a response that can be used in customer support, as well as methods to structure a knowledge article to improve answer generation.

I have already tried to fragment the articles in little ones to try to get a ready-to-use answer for the agents, but it didn't work the way I was expecting.

#Copilot
#AIConfiguration
#AICopilot(Agent,SupervisorAdmin)
#AI

Hi Lucas, 

I can give you some tips on this. 

For article structure best practices, you should look at paragraph length and structure. Try and aim for semantically coherent paragraphs that are around 200 words long. This should allow for a more precise retrieval and accurate generation. I think your approach of fragmenting articles into smaller pieces may not have worked because its actually the opposite of the recommendation, instead of creating many tiny articles you should create a comprehensive articles with well structured, meaningful paragraphs. This includes writing full sentences as opposed to bullet points.

For knowledge base configuration, specifically content search knowledge I suggest removing all phrasings make articles only title and body contents. If the title isnt long enough to summarize the content appropriately try adding 3-5 phrases to fill the gaps. But use these phrases sparingly and evenly distributed across knowledge articles. For the AI-generated answers try setting the confidence threshold to  0.65 or below for all language. If a knowledge article supersedes an intent-based rule, try lowering the NLU confidence threshold. 

I hope these tips help. 

Hello everyone, how are you?

I hope you're well!

I would like to know, if you're open to sharing, the way you use Copilot's answer generation function. I'd appreciate some tips and suggestions on efficient ways to generate a response that can be used in customer support, as well as methods to structure a knowledge article to improve answer generation.

I have already tried to fragment the articles in little ones to try to get a ready-to-use answer for the agents, but it didn't work the way I was expecting.

#Copilot
#AIConfiguration
#AICopilot(Agent,SupervisorAdmin)
#AI

Hi Cameron,

Thanks for getting back to me.

Is there a way to structure articles that is effective for agents while also being optimized for Copilot's answer generation? I'd love to find a middle ground.
In my current scenario, I need these articles to be structured so that agents can visualize information better, while also ensuring that answer generation remains effective.

Hi Lucas, 

I can give you some tips on this. 

For article structure best practices, you should look at paragraph length and structure. Try and aim for semantically coherent paragraphs that are around 200 words long. This should allow for a more precise retrieval and accurate generation. I think your approach of fragmenting articles into smaller pieces may not have worked because its actually the opposite of the recommendation, instead of creating many tiny articles you should create a comprehensive articles with well structured, meaningful paragraphs. This includes writing full sentences as opposed to bullet points.

For knowledge base configuration, specifically content search knowledge I suggest removing all phrasings make articles only title and body contents. If the title isnt long enough to summarize the content appropriately try adding 3-5 phrases to fill the gaps. But use these phrases sparingly and evenly distributed across knowledge articles. For the AI-generated answers try setting the confidence threshold to  0.65 or below for all language. If a knowledge article supersedes an intent-based rule, try lowering the NLU confidence threshold. 

I hope these tips help. 

Hello everyone, how are you?

I hope you're well!

I would like to know, if you're open to sharing, the way you use Copilot's answer generation function. I'd appreciate some tips and suggestions on efficient ways to generate a response that can be used in customer support, as well as methods to structure a knowledge article to improve answer generation.

I have already tried to fragment the articles in little ones to try to get a ready-to-use answer for the agents, but it didn't work the way I was expecting.

#Copilot
#AIConfiguration
#AICopilot(Agent,SupervisorAdmin)
#AI