Genesys Cloud - Developer Community!

Hi all,

Hoping someone here has attempted this. Our use case is this: At the main menu of a voice bot we have an Ask for Intent. Sometimes the caller speaks an utterance that takes the No Match route, which is absolutely fine. What we're attempting to do is present what the caller spoke to an answering agent in the agent script. We cannot locate a way to store this utterance as a variable when it's a No Match. 

When I debug the flow (and Execution Data is set to All for the level of debug data), I simply don't see an utterance being written anywhere when it's a No Match. I can however see what I or the caller said to reach the No Match branch in the Communications section of. It just isn't showing up in the Data section.

Has anyone ran into this? If so, did you determine a way for storing that utterance as a variable or a way to access that utterance within Architect?

We think the answering agent would get a benefit to seeing what the caller spoke to aid them in starting the conversation.

This is a native voice bot within Genesys Cloud and not a third party voice bot.

As always we appreciate any insight or help. 

Regards,

#Architect

------------------------------
Shane Jenkins
------------------------------

Hi Shane, how are you?

Have you checked all the built-in variables? (Architect built-in variables View summary - Genesys Cloud Resource Center)

I think one of those stores that information, but I'm not sure if in every context (like your "no match" scenario), but if you haven't yet, try to check this variable:

Session.LastCollectionUtterance

------------------------------
Marcello Jabur
------------------------------

Hi Marcello,

Thanks for replying. Yes, we have checked the built-in variables and also double-checked the Session.LastCollectionUtterance, etc. Unless we overlooked something, just doesn't appear to be storing that utterance should the bot determine it's a No Match. We'll keep searching. It isn't a show stopper, thankfully.

Thanks again!

------------------------------
Shane Jenkins
IT Sys Admin Mgr
------------------------------

Original Message:
Sent: 04-16-2026 16:29
From: Marcello Jabur
Subject: Is there a way to store a No Match utterance within a voice bot flow?

Hi Shane, how are you?

Have you checked all the built-in variables? (Architect built-in variables View summary - Genesys Cloud Resource Center)

I think one of those stores that information, but I'm not sure if in every context (like your "no match" scenario), but if you haven't yet, try to check this variable:

Session.LastCollectionUtterance

------------------------------
Marcello Jabur
------------------------------

Hi Shane,

I went through a very similar scenario and ended up implementing a pattern that might help in your case.

From what I could validate in the Genesys Cloud behavior, there isn't a native way to directly capture the raw utterance from an Ask for Intent when it falls into a No Match and store it as a variable in Architect. The utterance is visible in Utterance History, but not exposed directly as a reusable variable in the flow.

What worked well for us was combining Ask for Slot with event handling:

• In the bot flow, we use an Ask for Slot to capture what the caller says (even if loosely structured).
• We avoid handling No Match directly inside the action (no reprompt or minimal retry), so we don't "trap" the flow locally.
• Instead, we let the bot fall through to the Recognition Failure event handling.
• In the Recognition Failure, we route to a Reusable Task.
• In that task, we play a more natural prompt (e.g., "Sorry, I didn't quite get that, could you repeat?") and attempt to collect the input again.
• The slot used for capture is configured as an output of the bot flow.
• Back in the inbound flow, we bind that output to a variable and pass it to the agent script (via participant data).

This approach gives you two benefits:

1. You centralize your "didn't understand" experience using Reusable Tasks, which keeps the bot consistent and easier to maintain.
2. You still capture what the customer said and make it available to the agent, even in scenarios where intent recognition fails.

One important detail: if you configure a Max No Matches path inside the Ask for Slot, that path will typically take precedence, and the flow may not reach the global Recognition Failure handler. So it's best to choose a single strategy-either handle it locally in the action or centrally via event handling.

In our case, letting Recognition Failure handle it gave us more control and a cleaner design.

Hope this helps!

------------------------------
Fernando Sotto dos Santos
Consultor de Atendimento Senior Grupo Casas Bahia
------------------------------

Thank you Fernando! Appreciate the detailed response. We will test this out to see how it behaves. I'm glad that we aren't the only ones that ran into this issue. 

Thanks again!

Take care

------------------------------
Shane Jenkins
IT Sys Admin Mgr
------------------------------

Original Message:
Sent: 04-17-2026 15:58
From: Fernando Sotto dos Santos
Subject: Is there a way to store a No Match utterance within a voice bot flow?

Hi Shane,

I went through a very similar scenario and ended up implementing a pattern that might help in your case.

From what I could validate in the Genesys Cloud behavior, there isn't a native way to directly capture the raw utterance from an Ask for Intent when it falls into a No Match and store it as a variable in Architect. The utterance is visible in Utterance History, but not exposed directly as a reusable variable in the flow.

What worked well for us was combining Ask for Slot with event handling:

• In the bot flow, we use an Ask for Slot to capture what the caller says (even if loosely structured).
• We avoid handling No Match directly inside the action (no reprompt or minimal retry), so we don't "trap" the flow locally.
• Instead, we let the bot fall through to the Recognition Failure event handling.
• In the Recognition Failure, we route to a Reusable Task.
• In that task, we play a more natural prompt (e.g., "Sorry, I didn't quite get that, could you repeat?") and attempt to collect the input again.
• The slot used for capture is configured as an output of the bot flow.
• Back in the inbound flow, we bind that output to a variable and pass it to the agent script (via participant data).

This approach gives you two benefits:

1. You centralize your "didn't understand" experience using Reusable Tasks, which keeps the bot consistent and easier to maintain.
2. You still capture what the customer said and make it available to the agent, even in scenarios where intent recognition fails.

One important detail: if you configure a Max No Matches path inside the Ask for Slot, that path will typically take precedence, and the flow may not reach the global Recognition Failure handler. So it's best to choose a single strategy-either handle it locally in the action or centrally via event handling.

In our case, letting Recognition Failure handle it gave us more control and a cleaner design.

Hope this helps!

------------------------------
Fernando Sotto dos Santos
Consultor de Atendimento Senior Grupo Casas Bahia
------------------------------

Hi Shane,

Just following up on this-were you able to test it out and see how it behaves? I'd be interested to hear if it worked as expected or if you ran into any issues.

Thanks!

------------------------------
Fernando Sotto dos Santos
Consultor de Atendimento Senior Grupo Casas Bahia
------------------------------

Original Message:
Sent: 04-20-2026 11:00
From: Shane Jenkins
Subject: Is there a way to store a No Match utterance within a voice bot flow?

Thank you Fernando! Appreciate the detailed response. We will test this out to see how it behaves. I'm glad that we aren't the only ones that ran into this issue. 

Thanks again!

Take care

------------------------------
Shane Jenkins
IT Sys Admin Mgr
------------------------------


Original Message:
Sent: 04-17-2026 15:58
From: Fernando Sotto dos Santos
Subject: Is there a way to store a No Match utterance within a voice bot flow?

Hi Shane,

I went through a very similar scenario and ended up implementing a pattern that might help in your case.

From what I could validate in the Genesys Cloud behavior, there isn't a native way to directly capture the raw utterance from an Ask for Intent when it falls into a No Match and store it as a variable in Architect. The utterance is visible in Utterance History, but not exposed directly as a reusable variable in the flow.

What worked well for us was combining Ask for Slot with event handling:

• In the bot flow, we use an Ask for Slot to capture what the caller says (even if loosely structured).
• We avoid handling No Match directly inside the action (no reprompt or minimal retry), so we don't "trap" the flow locally.
• Instead, we let the bot fall through to the Recognition Failure event handling.
• In the Recognition Failure, we route to a Reusable Task.
• In that task, we play a more natural prompt (e.g., "Sorry, I didn't quite get that, could you repeat?") and attempt to collect the input again.
• The slot used for capture is configured as an output of the bot flow.
• Back in the inbound flow, we bind that output to a variable and pass it to the agent script (via participant data).

This approach gives you two benefits:

1. You centralize your "didn't understand" experience using Reusable Tasks, which keeps the bot consistent and easier to maintain.
2. You still capture what the customer said and make it available to the agent, even in scenarios where intent recognition fails.

One important detail: if you configure a Max No Matches path inside the Ask for Slot, that path will typically take precedence, and the flow may not reach the global Recognition Failure handler. So it's best to choose a single strategy-either handle it locally in the action or centrally via event handling.

In our case, letting Recognition Failure handle it gave us more control and a cleaner design.

Hope this helps!

------------------------------
Fernando Sotto dos Santos
Consultor de Atendimento Senior Grupo Casas Bahia

By the way, if this helped resolve your question, please consider marking it as the best answer so it can help others in the community as well.

------------------------------
Fernando Sotto dos Santos
Consultor de Atendimento Senior Grupo Casas Bahia
------------------------------

No‑match utterance is only available in the interaction transcript / communications view and isn't exposed as a flow variable. (a platform limitation that can be solved in Architect.)

& yes, please do mark the Best Answer once you're through this.

Thanks

------------------------------
Ashiesh Sharma
GCX- GCP, ARC, SCR, QM
Producer | Conductor | Composer
BT plc
------------------------------

Hi,

We've implemented it in a way similar to what Fernando comments on, we don't save the utterance from the ask intent, we save it at the end in the participant's data to also show it to the agent in the script when he receives the call.

If you have intentions that ask more than one question or have confirmation, you can put a decision that consults these intentions so as not to overwrite if your intention is to save the first thing the client said.

------------------------------
Marek Pietrzik
.
------------------------------

So, it worked. Great!

------------------------------
Fernando Sotto dos Santos
Consultor de Atendimento Senior Grupo Casas Bahia
------------------------------

Original Message:
Sent: 04-27-2026 10:58
From: Marek Pietrzik
Subject: Is there a way to store a No Match utterance within a voice bot flow?

Hi,

We've implemented it in a way similar to what Fernando comments on, we don't save the utterance from the ask intent, we save it at the end in the participant's data to also show it to the agent in the script when he receives the call.

If you have intentions that ask more than one question or have confirmation, you can put a decision that consults these intentions so as not to overwrite if your intention is to save the first thing the client said.

------------------------------
Marek Pietrzik
.
------------------------------