Genesys Cloud - Main

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

At first though it seems to be a lot of wrapup codes, that may be better suited for topic miner tracking,  

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

The wrap up code prediction does not work off the AI Insights as that is a totally different AI producing reason and outcome of the call.  Wrap up is based on the prompts you create in the wrap up code listing.  Make sure you have clear and distinct prompts for the AI to compare with the transcript.  40 is not unreasonable and a perfect use case for Wrap up Prediction 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Hi Robert,

Could you please advise where we can add prompts for wrap up codes? The only place we know of, where we can give any guidance to Copilot as to what the wrap up code should be used for, is the description for the wrap up code.

Thanks,

  Lee.

The wrap up code prediction does not work off the AI Insights as that is a totally different AI producing reason and outcome of the call.  Wrap up is based on the prompts you create in the wrap up code listing.  Make sure you have clear and distinct prompts for the AI to compare with the transcript.  40 is not unreasonable and a perfect use case for Wrap up Prediction 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Easy enough.  Get into Wrapup Codes under Contact Center or Orchestration -> Flow Assets in the new UI and add a prompt to the description of each wrapup code.  Be generous and specific with the prompt.  IE: for General Questions, don't just write "general questions" but "caller is calling with questions about their account that does not fit other wrapup codes"

Hi Robert,

Could you please advise where we can add prompts for wrap up codes? The only place we know of, where we can give any guidance to Copilot as to what the wrap up code should be used for, is the description for the wrap up code.

Thanks,

  Lee.

The wrap up code prediction does not work off the AI Insights as that is a totally different AI producing reason and outcome of the call.  Wrap up is based on the prompts you create in the wrap up code listing.  Make sure you have clear and distinct prompts for the AI to compare with the transcript.  40 is not unreasonable and a perfect use case for Wrap up Prediction 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Thanks Robert, we are looking to focus effort on improving our descriptions that sit against each wrap code. 

What would you say is the biggest use case for wrap codes? Is there overlap with STA topic miner? I am trying to wrap my head around the different use cases for each of these features.

Easy enough.  Get into Wrapup Codes under Contact Center or Orchestration -> Flow Assets in the new UI and add a prompt to the description of each wrapup code.  Be generous and specific with the prompt.  IE: for General Questions, don't just write "general questions" but "caller is calling with questions about their account that does not fit other wrapup codes"

Hi Robert,

Could you please advise where we can add prompts for wrap up codes? The only place we know of, where we can give any guidance to Copilot as to what the wrap up code should be used for, is the description for the wrap up code.

Thanks,

  Lee.

The wrap up code prediction does not work off the AI Insights as that is a totally different AI producing reason and outcome of the call.  Wrap up is based on the prompts you create in the wrap up code listing.  Make sure you have clear and distinct prompts for the AI to compare with the transcript.  40 is not unreasonable and a perfect use case for Wrap up Prediction 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Easy enough.  Get into Wrapup Codes under Contact Center or Orchestration -> Flow Assets in the new UI and add a prompt to the description of each wrapup code.  Be generous and specific with the prompt.  IE: for General Questions, don't just write "general questions" but "caller is calling with questions about their account that does not fit other wrapup codes"

Hi Robert,

Could you please advise where we can add prompts for wrap up codes? The only place we know of, where we can give any guidance to Copilot as to what the wrap up code should be used for, is the description for the wrap up code.

Thanks,

  Lee.

The wrap up code prediction does not work off the AI Insights as that is a totally different AI producing reason and outcome of the call.  Wrap up is based on the prompts you create in the wrap up code listing.  Make sure you have clear and distinct prompts for the AI to compare with the transcript.  40 is not unreasonable and a perfect use case for Wrap up Prediction 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Hi Robert, 

Sorry but I can't see anything under either the old UI or the new UI where you can add a prompt to the description of each wrap up code - unless you mean the description itself is effectively a prompt?
I wonder though if I have the right permissions because, in the new UI, I only have Data Tables / Flow Outcomes / Flow Milestones as options under Flow Assets - nothing about wrap up codes?

Regards,

Easy enough.  Get into Wrapup Codes under Contact Center or Orchestration -> Flow Assets in the new UI and add a prompt to the description of each wrapup code.  Be generous and specific with the prompt.  IE: for General Questions, don't just write "general questions" but "caller is calling with questions about their account that does not fit other wrapup codes"

Hi Robert,

Could you please advise where we can add prompts for wrap up codes? The only place we know of, where we can give any guidance to Copilot as to what the wrap up code should be used for, is the description for the wrap up code.

Thanks,

  Lee.

The wrap up code prediction does not work off the AI Insights as that is a totally different AI producing reason and outcome of the call.  Wrap up is based on the prompts you create in the wrap up code listing.  Make sure you have clear and distinct prompts for the AI to compare with the transcript.  40 is not unreasonable and a perfect use case for Wrap up Prediction 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Descriptions are Prompts for the AI.

Hi Robert, 

Sorry but I can't see anything under either the old UI or the new UI where you can add a prompt to the description of each wrap up code - unless you mean the description itself is effectively a prompt?
I wonder though if I have the right permissions because, in the new UI, I only have Data Tables / Flow Outcomes / Flow Milestones as options under Flow Assets - nothing about wrap up codes?

Regards,

Easy enough.  Get into Wrapup Codes under Contact Center or Orchestration -> Flow Assets in the new UI and add a prompt to the description of each wrapup code.  Be generous and specific with the prompt.  IE: for General Questions, don't just write "general questions" but "caller is calling with questions about their account that does not fit other wrapup codes"

Hi Robert,

Could you please advise where we can add prompts for wrap up codes? The only place we know of, where we can give any guidance to Copilot as to what the wrap up code should be used for, is the description for the wrap up code.

Thanks,

  Lee.

The wrap up code prediction does not work off the AI Insights as that is a totally different AI producing reason and outcome of the call.  Wrap up is based on the prompts you create in the wrap up code listing.  Make sure you have clear and distinct prompts for the AI to compare with the transcript.  40 is not unreasonable and a perfect use case for Wrap up Prediction 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Hi all, 

How are you tracking the accuracy of your wrap code suggestions? Is it just through feedback from agents and call reviews? 

I'm also experiencing difficulty getting the wrap code descriptions to prompt a few of our wrap codes, so any advice from others who have mastered it would be great. 

Matt 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Hi Matt,

We are tracking accuracy through the thumbs up / thumbs down inbuilt feedback mechanism. It has been very manual and time consuming to review.

I would be interested to see if there was another way to track this more efficiently. Robert and Alex both mentioned STA topic miner. I wonder if there could be a smarter way to validate wrap codes from topic miner.

Keen to hear from anyone else on how to do this efficiently? 

Hi all, 

How are you tracking the accuracy of your wrap code suggestions? Is it just through feedback from agents and call reviews? 

I'm also experiencing difficulty getting the wrap code descriptions to prompt a few of our wrap codes, so any advice from others who have mastered it would be great. 

Matt 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

In your original example, Super is not a great term for the AI.  I would use supervisor or manager in the description.  Topic Miner is really used to populate new Topics, but could guide you in what people are talking about to create wrap-up codes as well.  Which ones are not being recognized correctly?

Hi Matt,

We are tracking accuracy through the thumbs up / thumbs down inbuilt feedback mechanism. It has been very manual and time consuming to review.

I would be interested to see if there was another way to track this more efficiently. Robert and Alex both mentioned STA topic miner. I wonder if there could be a smarter way to validate wrap codes from topic miner.

Keen to hear from anyone else on how to do this efficiently? 

Hi all, 

How are you tracking the accuracy of your wrap code suggestions? Is it just through feedback from agents and call reviews? 

I'm also experiencing difficulty getting the wrap code descriptions to prompt a few of our wrap codes, so any advice from others who have mastered it would be great. 

Matt 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Good point. In this instance super refers to superannuation. I haven't tried topic miner yet but interested in seeing what it comes out with and how close the results are to the 40 odd wrap codes we have in place already.

In your original example, Super is not a great term for the AI.  I would use supervisor or manager in the description.  Topic Miner is really used to populate new Topics, but could guide you in what people are talking about to create wrap-up codes as well.  Which ones are not being recognized correctly?

Hi Matt,

We are tracking accuracy through the thumbs up / thumbs down inbuilt feedback mechanism. It has been very manual and time consuming to review.

I would be interested to see if there was another way to track this more efficiently. Robert and Alex both mentioned STA topic miner. I wonder if there could be a smarter way to validate wrap codes from topic miner.

Keen to hear from anyone else on how to do this efficiently? 

Hi all, 

How are you tracking the accuracy of your wrap code suggestions? Is it just through feedback from agents and call reviews? 

I'm also experiencing difficulty getting the wrap code descriptions to prompt a few of our wrap codes, so any advice from others who have mastered it would be great. 

Matt 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

There is some good news!  Able to get the Agent Copilot Summaries via API Explorer!!

Good point. In this instance super refers to superannuation. I haven't tried topic miner yet but interested in seeing what it comes out with and how close the results are to the 40 odd wrap codes we have in place already.

In your original example, Super is not a great term for the AI.  I would use supervisor or manager in the description.  Topic Miner is really used to populate new Topics, but could guide you in what people are talking about to create wrap-up codes as well.  Which ones are not being recognized correctly?

Hi Matt,

We are tracking accuracy through the thumbs up / thumbs down inbuilt feedback mechanism. It has been very manual and time consuming to review.

I would be interested to see if there was another way to track this more efficiently. Robert and Alex both mentioned STA topic miner. I wonder if there could be a smarter way to validate wrap codes from topic miner.

Keen to hear from anyone else on how to do this efficiently? 

Hi all, 

How are you tracking the accuracy of your wrap code suggestions? Is it just through feedback from agents and call reviews? 

I'm also experiencing difficulty getting the wrap code descriptions to prompt a few of our wrap codes, so any advice from others who have mastered it would be great. 

Matt 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Good point. In this instance super refers to superannuation. I haven't tried topic miner yet but interested in seeing what it comes out with and how close the results are to the 40 odd wrap codes we have in place already.

In your original example, Super is not a great term for the AI.  I would use supervisor or manager in the description.  Topic Miner is really used to populate new Topics, but could guide you in what people are talking about to create wrap-up codes as well.  Which ones are not being recognized correctly?

Hi Matt,

We are tracking accuracy through the thumbs up / thumbs down inbuilt feedback mechanism. It has been very manual and time consuming to review.

I would be interested to see if there was another way to track this more efficiently. Robert and Alex both mentioned STA topic miner. I wonder if there could be a smarter way to validate wrap codes from topic miner.

Keen to hear from anyone else on how to do this efficiently? 

Hi all, 

How are you tracking the accuracy of your wrap code suggestions? Is it just through feedback from agents and call reviews? 

I'm also experiencing difficulty getting the wrap code descriptions to prompt a few of our wrap codes, so any advice from others who have mastered it would be great. 

Matt 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,

Hi Deepa,

It might be that the description is too verbose. Check out some tips here: https://help.mypurecloud.com/faqs/what-are-the-best-practices-to-increase-the-accuracy-and-surfacing-of-wrap-up-codes-with-genesys-agent-copilot/

Amanda

Good point. In this instance super refers to superannuation. I haven't tried topic miner yet but interested in seeing what it comes out with and how close the results are to the 40 odd wrap codes we have in place already.

In your original example, Super is not a great term for the AI.  I would use supervisor or manager in the description.  Topic Miner is really used to populate new Topics, but could guide you in what people are talking about to create wrap-up codes as well.  Which ones are not being recognized correctly?

Hi Matt,

We are tracking accuracy through the thumbs up / thumbs down inbuilt feedback mechanism. It has been very manual and time consuming to review.

I would be interested to see if there was another way to track this more efficiently. Robert and Alex both mentioned STA topic miner. I wonder if there could be a smarter way to validate wrap codes from topic miner.

Keen to hear from anyone else on how to do this efficiently? 

Hi all, 

How are you tracking the accuracy of your wrap code suggestions? Is it just through feedback from agents and call reviews? 

I'm also experiencing difficulty getting the wrap code descriptions to prompt a few of our wrap codes, so any advice from others who have mastered it would be great. 

Matt 

Hi everyone,

We have recently rolled out CoPilot Predictive Wrap Codes and are seeing that our accuracy rates are sitting around ~50%. We are keen to improve this to at least 75%+ to gain efficiency.

My understanding is that it works off the 'resolution' and 'reason for call' field and matches that with the wrap code description.

We have approximately 40 different wrap codes. We have noticed that predictive wrap codes seem to struggle to differentiate between wrap codes that are similar such as:

• 'Access their super - make a withdrawal'  - Customer requests a standard super withdrawal for example reached preservation age or retirement
• 'Access their super - withdrawal progress / receipt of forms' - The customer called regarding a withdrawal request that was submitted but not yet processed.

 Does anyone have any advice on how to improve accuracy with predictive wrap codes?

Is there an ideal structure? 

Thanks heaps,