Genesys Cloud - Developer Community!

Hello, @Hann Chiad Hooy.

Just adding, I would avoid using only one-word utterances like "Outlook", "Word", "Teams", or "Visio" as the main training examples. They can work in some cases, but they can also create ambiguity and false positives, especially with words like "Teams" and "Word", which can appear in normal sentences with different meanings.

In my opinion, the design depends on what you want to do after identifying the product.

If all Microsoft Office products follow the same path, I would keep one intent, something like "Microsoft Office Products", and use slots to capture the specific product, such as Outlook, Word, Teams, Visio, Excel, etc. Then you can route or personalize the response based on the slot value.

But if each product has a very different process, different troubleshooting steps, different routing, or different teams responsible for support, then separate intents may be cleaner, like Microsoft Outlook Support, Microsoft Teams Support, Microsoft Word Support, and so on.

Either way, I would train the bot with full user phrases, not only product names. For example: "I need help with Outlook", "Teams is not opening", "I cannot access Word", "I have an issue with Visio". That gives the model more context and usually improves intent detection.

Hi @Arthur Pereira Reinoldes / @Phaneendra Avatapalli 

Thanks for your reply. Appreciate your time in providing your thoughts and feedbacks.

Yes, all the Microsoft products follow the same process is in going to the same queue. My concern is if all are grouped under 1 single intent, then the utterances would be huge. Just imagine we have more than 10 sub-products under Microsoft and each of them has their own utterances. So I guess managing under slots should be the best practice in this case?

Hello, @Hann Chiad Hooy.

Just adding, I would avoid using only one-word utterances like "Outlook", "Word", "Teams", or "Visio" as the main training examples. They can work in some cases, but they can also create ambiguity and false positives, especially with words like "Teams" and "Word", which can appear in normal sentences with different meanings.

In my opinion, the design depends on what you want to do after identifying the product.

If all Microsoft Office products follow the same path, I would keep one intent, something like "Microsoft Office Products", and use slots to capture the specific product, such as Outlook, Word, Teams, Visio, Excel, etc. Then you can route or personalize the response based on the slot value.

But if each product has a very different process, different troubleshooting steps, different routing, or different teams responsible for support, then separate intents may be cleaner, like Microsoft Outlook Support, Microsoft Teams Support, Microsoft Word Support, and so on.

Either way, I would train the bot with full user phrases, not only product names. For example: "I need help with Outlook", "Teams is not opening", "I cannot access Word", "I have an issue with Visio". That gives the model more context and usually improves intent detection.

Hey, Hann! 

If this approach fits your business rule, yes, it may help you.

Since all Microsoft products follow the same process and go to the same queue, I would use one intent and a slot to capture the specific product, such as Outlook, Word, Teams, Visio, or Excel.

This keeps the intent focused on the customer goal, while the slot manages the product variation.

Hi @Arthur Pereira Reinoldes / @Phaneendra Avatapalli 

Thanks for your reply. Appreciate your time in providing your thoughts and feedbacks.

Yes, all the Microsoft products follow the same process is in going to the same queue. My concern is if all are grouped under 1 single intent, then the utterances would be huge. Just imagine we have more than 10 sub-products under Microsoft and each of them has their own utterances. So I guess managing under slots should be the best practice in this case?

Hello, @Hann Chiad Hooy.

Just adding, I would avoid using only one-word utterances like "Outlook", "Word", "Teams", or "Visio" as the main training examples. They can work in some cases, but they can also create ambiguity and false positives, especially with words like "Teams" and "Word", which can appear in normal sentences with different meanings.

In my opinion, the design depends on what you want to do after identifying the product.

If all Microsoft Office products follow the same path, I would keep one intent, something like "Microsoft Office Products", and use slots to capture the specific product, such as Outlook, Word, Teams, Visio, Excel, etc. Then you can route or personalize the response based on the slot value.

But if each product has a very different process, different troubleshooting steps, different routing, or different teams responsible for support, then separate intents may be cleaner, like Microsoft Outlook Support, Microsoft Teams Support, Microsoft Word Support, and so on.

Either way, I would train the bot with full user phrases, not only product names. For example: "I need help with Outlook", "Teams is not opening", "I cannot access Word", "I have an issue with Visio". That gives the model more context and usually improves intent detection.