# Connecting Google Dialogflow CX to Webex Contact Center

Systems: dialogflow-cx to webex-contact-center

For Webex Contact Center administrators and partner engineers configuring Google Dialogflow CX voice integration.

Canonical: https://warmtransfer.net/guides/wxcc-dialogflow-cx-setup

Last verified: 2026-10-01

Before configuring the Virtual Agent V2 activity with Dialogflow CX, the Google CCAI connector and a CCAI feature configuration must exist, along with an entry point and routing flow[^55]. Webex Contact Center system limits allow 6000 active entry points[^62].

## Before you start

- Cisco lists CCAI feature regional availability as US, UK, Frankfurt, Australia, Japan, Singapore and Canada[^11].
- The Google CCAI connector prerequisites include a Cisco subscription Flex SKU and a Contact Center order placed through Cisco Commerce Workspace[^17].
- The virtual agent integration requires an active GCP billing account because the Dialogflow, Speech-to-Text, Natural Language and Text-to-Speech APIs require an active billing account[^41][^63].
- The Virtual Agent-Voice article's Limitations section states that en-US is currently the only supported language[^64].
- Virtual Agent-Voice supports only the u-law codec[^61][^65].

## What changes by situation

- Which virtual agent platform will handle the conversational voice interaction? Google Dialogflow CX through the Google CCAI connector; Webex AI Agent built in Webex AI Agent Studio.
- Who owns the billable Google Cloud Platform project behind the connector? Cisco-provisioned billable project and service account; Customer-owned GCP project for both CCAI and billing.
- Which Google Cloud location hosts the Dialogflow CX agent? Global serving with data at rest in the United States; A specific regional location keeping data at rest in that region.
- Do inbound voice callers connect through a single PSTN region or across multiple regions? Single PSTN region; Multiple PSTN regions requiring regional media.

## Step 1: Validate the conversational platform selection

**Google Dialogflow CX through the Google CCAI connector**

### Do
Confirm that your tenant is deployed in one of the supported CCAI regions: US, UK, Frankfurt, Australia, Japan, Singapore or Canada[^11]. The Virtual Agent-Voice article's Limitations section states that en-US is currently the only supported language[^64].

### Verify
Suggested check: verify that your tenant region matches the supported list and that caller requirements align with the en-US language constraint.

### Rollback
Suggested rollback: no changes have been applied to either platform yet.

**Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3][^66]. Cisco states that autonomous AI agents on the voice channel are currently limited to specific customers[^1][^67]. Since 8 June 2026 Webex AI Agent supports a selection of non-English languages for voice and digital channels as generally available[^2]. If you select Webex AI Agent, construct the bot in Webex AI Agent Studio rather than following the Google Dialogflow CX configuration steps below[^3].

### Verify
Suggested check: confirm access to AI Agent Studio and verify engine language compatibility.

### Rollback
Suggested rollback: no changes have been applied yet.

## Step 2: Prepare the Google Cloud project

**Google Dialogflow CX through the Google CCAI connector**

### Do
Billing must be enabled on the Google Cloud project because the Dialogflow, Speech-to-Text, Natural Language and Text-to-Speech APIs used by the virtual agent require an active billing account[^41][^63]. The Dialogflow API must be enabled on the Google Cloud project, which requires the `serviceusage.services.enable` permission typically held through the Owner role[^25]. Assign appropriate Google access control roles to required principals, such as Dialogflow API Client and Dialogflow API Admin[^34].

### Verify
Suggested check: confirm in the Google Cloud console that the Dialogflow API is enabled and that a valid billing account is linked.

### Rollback
Suggested rollback: disable the Dialogflow API in the Google Cloud console. Do not delete the Google Cloud project, as agents linked to a deleted project cannot be recovered[^33].

**Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3]. Manage agent configuration directly within Webex AI Agent Studio[^3].

### Verify
Suggested check: confirm that you are managing configurations within Webex AI Agent Studio.

### Rollback
Suggested rollback: no Google Cloud settings were configured.

## Step 3: Complete Cisco onboarding prerequisites

**Google Dialogflow CX through the Google CCAI connector**

### Do
Before configuring the Google CCAI connector, a GCP project ID and a Dialogflow agent application must already exist in Google Cloud Platform[^18]. Complete the Assessment to Quality (A2Q) process for CCAI eligibility[^15][^68]. The connector prerequisites include submitting Cisco's CCAI provisioning form to map the customer GCP project ID to Cisco's service account and sharing the project ID and email with a Cisco representative so a primary service account is set up for billing[^19][^68]. Cisco sends updated GCP credentials in an email titled CCAI Onboarding Provision Status Update, and the billable GCP project name used in the connector comes from that notification[^16].

### Verify
Suggested check: confirm receipt of the onboarding status update email containing the provisioned billable project information.

### Rollback
Suggested rollback: contact the account team to modify or withdraw the onboarding provisioning submission.

**Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: confirm that administrative access to AI Agent Studio is present.

### Rollback
Suggested rollback: no external onboarding request was submitted.

## Step 4: Create the Dialogflow CX agent in the target location

**Global serving with data at rest in the United States + Google Dialogflow CX through the Google CCAI connector**

### Do
The tech note creates the Dialogflow CX agent at dialogflow.cloud.google.com/cx with Create Agent > Build your own and selects the location Global serving, data-at-rest in US[^40]. The global location stores data at rest in the US[^26].

### Verify
Suggested check: inspect the agent settings in Dialogflow CX console to verify that the location is set to `global`.

### Rollback
A Dialogflow CX agent's location cannot change after the agent is created; changing region requires exporting the agent and restoring it into a new agent[^28].

**A specific regional location keeping data at rest in that region + Google Dialogflow CX through the Google CCAI connector**

### Do
WarmTransfer's reading of the sources is that because agent location is fixed at creation and Cisco uses one conversation profile per PSTN region, the Google location decision should be made before building the agent and repeated per region for multi-region tenants[^36]. Dialogflow CX agent locations include us-central1, northamerica-northeast1, us-east1, us-west1, us, europe-west1, europe-west2, europe-west3, europe-west4, europe-west6, australia-southeast1, asia-northeast1, asia-south1, asia-southeast1, asia-southeast2 and global[^29]. Regional Dialogflow CX locations keep data at rest in that region[^26].

### Verify
Suggested check: confirm in Dialogflow CX agent settings that the selected regional identifier matches your data residency requirements.

### Rollback
A Dialogflow CX agent's location cannot change after the agent is created; changing region requires exporting the agent and restoring it into a new agent[^28].

**Global serving with data at rest in the United States + Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: confirm agent setup directly in Webex AI Agent Studio.

### Rollback
Suggested rollback: no Dialogflow CX agent was provisioned.

**A specific regional location keeping data at rest in that region + Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: confirm agent setup directly in Webex AI Agent Studio.

### Rollback
Suggested rollback: no Dialogflow CX agent was provisioned.

## Step 5: Configure conversation flows, hand-back payloads, and escalations

**Google Dialogflow CX through the Google CCAI connector**

### Do
The tech note assigns the published flow to an entry point and tests by calling the number mapped to that entry point, checking greeting, data capture and that live agent handoff reaches the right queue; the Dialogflow CX Test Agent simulator is used to test pages before that[^48]. The virtual agent transcript appears in Agent Desktop only if Enable Conversation Transcript is on and an Agent Says fulfillment response is configured in Dialogflow CX[^60]. To return data on the Handled path, the tech note adds a Custom payload of the form Execute_Request.Data.Params on a route that transitions to End Flow, and the flow reads it with a Parse activity on VirtualAgentV2.MetaData using the expression $.Params.<name>[^45]. To escalate, the tech note adds a Live agent handoff dialog option on the route with a payload under the key dialogflow.ccai.live-agent-escalation and transitions to End Flow, which takes the call out of the Escalated output of Virtual Agent V2[^46][^69]. Dialogflow CX custom payloads must follow the JSON format defined by the integration, may reference parameters inside double quotes, and are limited to 24 levels of JSON depth[^31]. Google documents the live agent handoff response as a signal to the detect-intent caller that the conversation should go to a human, used by Dialogflow CX only for measurement and not altering session state, with a payload structure of the integrator's choosing[^27][^70].

### Verify
The Dialogflow CX Test Agent simulator is used to test pages[^48].

### Rollback
Suggested rollback: delete or reconfigure the transition routes and remove custom payloads from the affected Dialogflow CX pages.

**Webex AI Agent built in Webex AI Agent Studio**

### Do
AI Agent Studio transfer actions have a toggle: when on, the AI agent tells the customer about the transfer before executing it; when off, it performs a silent transfer[^71]. In Webex AI Agent Studio the Agent handover action is enabled by default, allowing the AI agent to escalate the conversation to a human agent[^72].

### Verify
Suggested check: verify test interactions in AI Agent Studio and confirm handover triggers.

### Rollback
Suggested rollback: modify the bot instructions or disable the handover action in AI Agent Studio.

## Step 6: (Optional) Set up custom events, parameters, and streaming responses

**Google Dialogflow CX through the Google CCAI connector**

### Do
A Dialogflow CX custom event for Webex Contact Center is a Custom payload with Execute_Request containing Event_Name and Data.Params, and Cisco instructs defining only the custom payload with no other dialogue options[^22]. To resume the conversation after a custom event, the route's Transition is set to the same page, and an event handler with the event name is created in the Dialogflow CX flow[^23]. Webex Contact Center recognises three Dialogflow CX custom parameters set under Fulfillment > Parameter Presets: wxRecognitionModelName, wxRecognitionModelVariant and wxCallCompletion[^21]. With Dialogflow CX, Webex Contact Center supports partial response, custom event, custom parameters and custom payload[^49]. Dialogflow CX can return queued responses as a partial response on the streaming APIs while a long-running webhook is processing, when Return partial response is enabled[^30].

### Verify
Suggested check: simulate the custom event in the Dialogflow CX Test Agent simulator and confirm that the event handler triggers and routes back as expected.

### Rollback
Suggested rollback: remove parameter presets, event handlers, and custom payload blocks from the Dialogflow CX flow.

**Webex AI Agent built in Webex AI Agent Studio**

### Do
Custom data passed from Flow Designer through the Virtual Agent V2 State Event settings is referenced inside an AI agent's goal, welcome message, instructions and action or slot descriptions with {{variable name}} syntax[^73]. Custom data for autonomous AI agents is currently supported only on the voice channel[^74].

### Verify
Suggested check: test variable resolution within conversational prompts in AI Agent Studio.

### Rollback
Suggested rollback: remove variable placeholders from prompt text in AI Agent Studio.

## Step 7: Create Google Cloud conversation profile(s)

**Single PSTN region + Google Dialogflow CX through the Google CCAI connector**

### Do
The tech note creates the conversation profile in the Google console under Agent Assist > Conversation profile > Create, enables the virtual agent by selecting the Dialogflow CX agent, and copies the Conversation Profile ID[^43]. A Google conversation profile has a display name and language, and is connected to a Dialogflow CX agent by choosing the profile, selecting Link with agent and picking the agent[^32]. When creating the Google conversation profile, the Enable virtual agent option under Choose to use Dialogflow must be turned on for use with Webex Contact Center virtual agent[^6]. For the Google CCAI connector, the CCAI configuration takes a Google Conversation Profile ID, obtained from the Agent Assist UI in Google Cloud Platform[^9].

### Verify
Suggested check: verify that the profile appears under Agent Assist conversation profiles and displays as linked to your Dialogflow CX agent.

### Rollback
A Google conversation profile has a display name and language, and is connected to a Dialogflow CX agent by choosing the profile, selecting Link with agent and picking the agent[^32].

**Multiple PSTN regions requiring regional media + Google Dialogflow CX through the Google CCAI connector**

### Do
To handle Virtual Agent-Voice calls from multiple PSTN regions, a Google conversation profile and a CCAI configuration are created for each region and entry point mappings are configured with a PSTN region and mapped to a single flow[^38]. The tech note creates the conversation profile in the Google console under Agent Assist > Conversation profile > Create, enables the virtual agent by selecting the Dialogflow CX agent, and copies the Conversation Profile ID[^43]. When creating the Google conversation profile, the Enable virtual agent option under Choose to use Dialogflow must be turned on for use with Webex Contact Center virtual agent[^6]. A Google conversation profile has a display name and language, and is connected to a Dialogflow CX agent by choosing the profile, selecting Link with agent and picking the agent[^32].

### Verify
Suggested check: verify that a conversation profile exists for every planned PSTN region and that each profile has a distinct Profile ID linked to an agent.

### Rollback
A Google conversation profile has a display name and language, and is connected to a Dialogflow CX agent by choosing the profile, selecting Link with agent and picking the agent[^32].

**Single PSTN region + Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: continue directly to flow configuration.

### Rollback
Suggested rollback: no Google conversation profiles were provisioned.

**Multiple PSTN regions requiring regional media + Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: continue directly to flow configuration.

### Rollback
Suggested rollback: no Google conversation profiles were provisioned.

## Step 8: Configure the Google CCAI connector in Control Hub

**Cisco-provisioned billable project and service account + Google Dialogflow CX through the Google CCAI connector**

### Do
The Google CCAI connector is set up in Control Hub under Services > Contact Center > Tenant Settings > Integrations > Connectors by clicking Set Up or Add More on the Google Contact Center AI card[^13]. Connector setup requires clicking Sign In with Google and authenticating with the same Google ID that was used to create the CCAI project, then granting the requested data access permissions[^12]. Cisco sends updated GCP credentials in an email titled CCAI Onboarding Provision Status Update, and the billable GCP project name used in the connector comes from that notification[^16]. In the connector form the administrator enters a connector name, selects the CCAI Project Name, the Billable Project Name and the billable service account from dropdowns, then clicks Save and Refresh to see the new connector[^20].

### Verify
In the connector form the administrator enters a connector name, selects the CCAI Project Name, the Billable Project Name and the billable service account from dropdowns, then clicks Save and Refresh to see the new connector[^20]. Cisco tech note 221526 states that the Google Contact Center AI connector is typically preconfigured for Dialogflow CX integration and that creating a new connector requires Cisco Project permissions[^42].

### Rollback
Suggested rollback: remove or disable the connector in Control Hub if permitted by tenant permissions.

**Customer-owned GCP project for both CCAI and billing + Google Dialogflow CX through the Google CCAI connector**

### Do
The Google CCAI connector is set up in Control Hub under Services > Contact Center > Tenant Settings > Integrations > Connectors by clicking Set Up or Add More on the Google Contact Center AI card[^13]. Connector setup requires clicking Sign In with Google and authenticating with the same Google ID that was used to create the CCAI project, then granting the requested data access permissions[^12]. To use the customer's own GCP project, the CCAI Project Name and the Billable Project Name selected in the connector must be the same[^14][^75]. In the connector form the administrator enters a connector name, selects the CCAI Project Name, the Billable Project Name and the billable service account from dropdowns, then clicks Save and Refresh to see the new connector[^20].

### Verify
To use the customer's own GCP project, the CCAI Project Name and the Billable Project Name selected in the connector must be the same[^14]. In the connector form the administrator enters a connector name, selects the CCAI Project Name, the Billable Project Name and the billable service account from dropdowns, then clicks Save and Refresh to see the new connector[^20].

### Rollback
Suggested rollback: delete the connector card from Control Hub Integrations > Connectors.

**Cisco-provisioned billable project and service account + Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: continue directly to flow creation.

### Rollback
Suggested rollback: no connector was configured.

**Customer-owned GCP project for both CCAI and billing + Webex AI Agent built in Webex AI Agent Studio**

### Do
Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: continue directly to flow creation.

### Rollback
Suggested rollback: no connector was configured.

## Step 9: Create Contact Center AI configuration(s)

**Single PSTN region + Google Dialogflow CX through the Google CCAI connector**

### Do
The service-specific integration connector, for example the Google Contact Center AI connector, must exist before a Contact Center AI configuration can be created[^8][^76]. A Contact Center AI configuration is created in Control Hub under Services > Contact Center > Tenant settings > Integrations, Features tab, New > Contact Center AI Config, with a unique Config Name and optional Description[^7][^77]. The Contact Center AI Connector field offers three connector types: Webex CCAI, Integration Connector (third-party platforms such as Google) and Service App[^4]. For the Google CCAI connector, the CCAI configuration takes a Google Conversation Profile ID, obtained from the Agent Assist UI in Google Cloud Platform[^9][^78]. A CCAI configuration can be set as the default for Virtual Agent conversations; the Agent Answers default option is not applicable to Webex Contact Center[^5]. After Save, a confirmation message appears and the configured CCAI card is shown on the Features page[^10].

### Verify
After Save, a confirmation message appears and the configured CCAI card is shown on the Features page[^10]. The CCAI config ID is copied from the configured Contact Center AI card under Control Hub > Services > Contact Center > Tenant Settings > Integrations > Features[^37].

### Rollback
Suggested rollback: delete the CCAI configuration card from the Features tab.

**Multiple PSTN regions requiring regional media + Google Dialogflow CX through the Google CCAI connector**

### Do
The service-specific integration connector, for example the Google Contact Center AI connector, must exist before a Contact Center AI configuration can be created[^8]. A Contact Center AI configuration is created in Control Hub under Services > Contact Center > Tenant settings > Integrations, Features tab, New > Contact Center AI Config, with a unique Config Name and optional Description[^7][^77]. The Contact Center AI Connector field offers three connector types: Webex CCAI, Integration Connector (third-party platforms such as Google) and Service App[^4]. For the Google CCAI connector, the CCAI configuration takes a Google Conversation Profile ID, obtained from the Agent Assist UI in Google Cloud Platform[^9]. To handle Virtual Agent-Voice calls from multiple PSTN regions, a Google conversation profile and a CCAI configuration are created for each region and entry point mappings are configured with a PSTN region and mapped to a single flow[^38]. After Save, a confirmation message appears and the configured CCAI card is shown on the Features page[^10].

### Verify
After Save, a confirmation message appears and the configured CCAI card is shown on the Features page[^10]. The CCAI config ID is copied from the configured Contact Center AI card under Control Hub > Services > Contact Center > Tenant Settings > Integrations > Features[^37].

### Rollback
Suggested rollback: delete the regional CCAI configuration cards from Control Hub.

**Single PSTN region + Webex AI Agent built in Webex AI Agent Studio**

### Do
The Contact Center AI Connector field offers three connector types: Webex CCAI, Integration Connector (third-party platforms such as Google) and Service App[^4]. Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: continue directly to flow creation.

### Rollback
Suggested rollback: no CCAI feature cards were created.

**Multiple PSTN regions requiring regional media + Webex AI Agent built in Webex AI Agent Studio**

### Do
The Contact Center AI Connector field offers three connector types: Webex CCAI, Integration Connector (third-party platforms such as Google) and Service App[^4]. Webex AI Agents are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3].

### Verify
Suggested check: continue directly to flow creation.

### Rollback
Suggested rollback: no CCAI feature cards were created.

## Step 10: Place the Virtual Agent V2 activity in Flow Designer

**Single PSTN region + Google Dialogflow CX through the Google CCAI connector**

### Do
The Virtual Agent V2 activity cannot be placed after a Queue Contact activity and cannot be placed in the Event Flow canvas[^54][^79][^80]. The Virtual Agent V2 activity selects its CCAI configuration either statically from a dropdown of configured CCAI features or through a Variable Contact Center AI Config chosen at run time, for example by caller PSTN region[^57][^81]. The virtual agent transcript appears in Agent Desktop only if Enable Conversation Transcript is on and an Agent Says fulfillment response is configured in Dialogflow CX[^60][^82]. Global_Language and Global_VoiceName global variables set the input language and output voice for Virtual Agent V2, and can be overridden with Set Variable activities placed before the activity[^50][^83][^84]. Non-configurable Virtual Agent V2 defaults include infinite retries on invalid or no input, barge-in enabled, # as DTMF termination, a 5-second no-input timeout and a 3-second inter-digit timeout[^58][^85][^86].

### Verify
Flow validation checks expressions for structural errors only and cannot evaluate a function or check that a variable resolves to the expected value[^87]. The Virtual Agent V2 activity has three outcome paths: Handled when the virtual agent execution completes, Escalated when the call must go to a human agent, and Errored for any error[^52].

### Rollback
Flow Designer undo and redo covers the last 10 canvas actions[^88]. Suggested rollback: remove the activity from the flow canvas or restore a previously saved canvas state.

**Multiple PSTN regions requiring regional media + Google Dialogflow CX through the Google CCAI connector**

### Do
To handle Virtual Agent-Voice calls from multiple PSTN regions, a Google conversation profile and a CCAI configuration are created for each region and entry point mappings are configured with a PSTN region and mapped to a single flow[^38]. The multi-region flow uses the NewPhoneContact.PSTNRegion event output variable in a Case activity, a Set Variable per region that assigns that region's CCAI config ID to a flow variable, and Virtual Agent V2 set to Variable Contact Center AI Config using that variable[^39]. The Virtual Agent V2 activity cannot be placed after a Queue Contact activity and cannot be placed in the Event Flow canvas[^54][^79][^80]. In the Virtual Agent V2 activity Conversational Experience settings, a Static Contact Center AI Config uses the default PSTN region while a Variable Contact Center AI Config selects the CCAI config dynamically, for example based on the caller's region[^81].

### Verify
The multi-region flow uses the NewPhoneContact.PSTNRegion event output variable in a Case activity, a Set Variable per region that assigns that region's CCAI config ID to a flow variable, and Virtual Agent V2 set to Variable Contact Center AI Config using that variable[^39]. Flow validation checks expressions for structural errors only and cannot evaluate a function or check that a variable resolves to the expected value[^87].

### Rollback
Flow Designer undo and redo covers the last 10 canvas actions[^88]. Suggested rollback: revert the flow to static configuration or use undo.

**Single PSTN region + Webex AI Agent built in Webex AI Agent Studio**

### Do
The Virtual Agent V2 activity cannot be placed after a Queue Contact activity and cannot be placed in the Event Flow canvas[^54][^79]. Webex AI Agents, scripted or autonomous, are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity[^3][^66]. For Webex AI agents, the voice set through Global_VoiceName must be supported by the engine that the AI agent uses[^89].

### Verify
Suggested check: validate the flow and ensure the Webex AI Agent binds cleanly to Virtual Agent V2.

### Rollback
Suggested rollback: remove the activity or revert to an earlier flow version.

**Multiple PSTN regions requiring regional media + Webex AI Agent built in Webex AI Agent Studio**

### Do
The Virtual Agent V2 activity cannot be placed after a Queue Contact activity and cannot be placed in the Event Flow canvas[^54][^79]. The Virtual Agent V2 activity selects its CCAI configuration either statically from a dropdown of configured CCAI features or through a Variable Contact Center AI Config chosen at run time, for example by caller PSTN region[^57]. For Webex AI agents, the voice set through Global_VoiceName must be supported by the engine that the AI agent uses[^89].

### Verify
Suggested check: validate the flow and verify that all regional branch paths connect cleanly to the activity.

### Rollback
Suggested rollback: delete or disconnect the activity on the canvas.

## Step 11: Route the Handled, Escalated, and Errored paths

**Google Dialogflow CX through the Google CCAI connector**

### Do
The Virtual Agent V2 activity has three outcome paths: Handled when the virtual agent execution completes, Escalated when the call must go to a human agent, and Errored for any error[^52][^90][^91][^92].
To return data on the Handled path, the tech note adds a Custom payload of the form Execute_Request.Data.Params on a route that transitions to End Flow, and the flow reads it with a Parse activity on VirtualAgentV2.MetaData using the expression $.Params.<name>[^45][^93].
Two unrelated expression syntaxes are used in the same activity configuration: the Pebble double brace form for flow variables and a JSON path form for parsing a response body[^94].
WarmTransfer's reading of the sources is that because the Escalated outcome only exits the activity and the Dialogflow handoff signal performs no transfer, an 'escalation never reaches the queue' symptom is most likely in the flow after the Escalated outcome (Parse, Case and Queue Contact wiring) or in a missing handoff or transfer action on the bot side[^95].
On the Escalated path the tech note parses VirtualAgentV2.MetaData with $.dialogflow.ccai.live-agent-escalation.<key> into a flow variable, branches with a Case activity and sends each branch to its own Queue Contact activity, with the Case Default output routed to one queue[^44][^96].
The custom event name is mapped to the State Event Name in the Virtual Agent V2 activity in Flow Designer for decision mapping[^24][^53][^97].
The Virtual Agent V2 activity's Errored output is the path the flow takes in any error scenario[^52][^92].

### Verify
Flow validation checks expressions for structural errors only and cannot evaluate a function or check that a variable resolves to the expected value[^87]. The flow version number counts how many times the flow has been published, and version labels can be applied[^98].

### Rollback
Flow Designer has no version control feature[^98]. The flow version number counts how many times the flow has been published, and version labels can be applied[^98].

**Webex AI Agent built in Webex AI Agent Studio**

### Do
The Virtual Agent V2 activity has three outcome paths: Handled when the virtual agent execution completes, Escalated when the call must go to a human agent, and Errored for any error[^52][^90][^91][^92]. In Cisco's example, the Escalated path parses the handoff payload into a flow variable and uses a Case activity to route to the matching Queue Contact activity[^96]. In European regions the Webex AI Agent transparency disclosure is enabled by default, and administrators who disable it must acknowledge the risks and add audit comments before opting out[^99]. Outside European regions, Webex AI Agent administrators choose whether to enable the AI transparency disclosure and can configure its message; Cisco recommends enabling it as a best practice[^100]. Webex AI Agent can play an AI transparency disclosure on voice channels that callers cannot interrupt and that plays before the welcome message[^101].

### Verify
Suggested check: validate and publish the flow in Flow Designer.

### Rollback
Suggested rollback: restore the prior flow version.

## Step 12: Assign entry points and test voice interactions end to end

**Google Dialogflow CX through the Google CCAI connector**

### Do
Inbound telephony entry points are created in Control Hub under Services, Contact Center, Customer Experience, Channels[^102]. An inbound voice entry point is created with channel type Inbound Telephony and a routing flow selected from a drop-down, and its timezone defaults to the tenant timezone[^103]. The tech note assigns the published flow to an entry point and tests by calling the number mapped to that entry point, checking greeting, data capture and that live agent handoff reaches the right queue; the Dialogflow CX Test Agent simulator is used to test pages before that[^48]. Virtual Agent-Voice supports only the u-law codec[^61][^65]. Calls reaching the Webex Contact Center media edge must use G.711uLaw or G.711aLaw[^104].

### Verify
The tech note assigns the published flow to an entry point and tests by calling the number mapped to that entry point, checking greeting, data capture and that live agent handoff reaches the right queue; the Dialogflow CX Test Agent simulator is used to test pages before that[^48].
WarmTransfer's reading of the sources is that documented constraints that could produce silence or a non-answering virtual agent are a missing CCAI config or connector, a tenant region outside the listed CCAI regions, a non-u-law media path, and placing the activity after Queue Contact or in the Event Flow[^105].
To return data on the Handled path, the tech note adds a Custom payload of the form Execute_Request.Data.Params on a route that transitions to End Flow, and the flow reads it with a Parse activity on VirtualAgentV2.MetaData using the expression $.Params.<name>[^45].
The output variable VirtualAgentV2.MetaData stores the JSON data that the system receives from the agent as part of fulfillment, custom-event handling or a transfer action, and can drive later flow logic[^106].
The Virtual Agent V2 activity takes its Escalated outcome when the call must be escalated to a human agent[^91].
When a call is transferred to a live agent, the virtual agent transcript appears in the Transcript gadget on Agent Desktop, which requires the Transcript gadget to be configured in the desktop layout[^107].
The virtual agent transcript appears in Agent Desktop only if Enable Conversation Transcript is on and an Agent Says fulfillment response is configured in Dialogflow CX[^60].
The Virtual Agent V2 activity's Termination Delay accepts 0 to 30 seconds with a default of 30[^108].

### Rollback
Suggested rollback: reassign the entry point to point back to its previous production routing flow.

**Webex AI Agent built in Webex AI Agent Studio**

### Do
Inbound telephony entry points are created in Control Hub under Services, Contact Center, Customer Experience, Channels[^102]. An inbound voice entry point is created with channel type Inbound Telephony and a routing flow selected from a drop-down, and its timezone defaults to the tenant timezone[^103]. Webex AI Agent can play an AI transparency disclosure on voice channels that callers cannot interrupt and that plays before the welcome message[^101]. In Webex AI Agent Studio the Agent handover action is enabled by default, allowing the AI agent to escalate the conversation to a human agent[^72].

### Verify
Suggested check: confirm that disclosure plays prior to the greeting and that agent transfer places the call into queue.

### Rollback
Suggested rollback: reassign the entry point to its previous flow.

## Applicability

Applies to: Cisco Webex AI Agent, Cisco Webex Contact Center, Google Dialogflow CX, Cisco Webex AI Agent (Webex Contact Center), Cisco Webex AI Agent Studio, and Cisco Webex Contact Center with Webex Calling. Deployments: cloud and multi-tenant. Sources checked 2026-10-01. Cisco lists CCAI feature regional availability as US, UK, Frankfurt, Australia, Japan, Singapore and Canada[^11][^109]. The Virtual Agent-Voice article's Limitations section states that en-US is currently the only supported language[^64]. Virtual Agent-Voice supports only the u-law codec[^61][^65]. Access to the autonomous Webex AI Agent for voice calls is limited to specific customers[^1][^67]. Webex Contact Center PSTN, sold by Cisco as a bundle with Webex Contact Center, is available only to customers in the contiguous United States and Canada[^110].

## What remains uncertain

The Virtual Agent V2 activity's Termination Delay accepts 0 to 30 seconds with a default of 30[^108]. Whether a Google conversation profile must reside in the exact same Google Cloud region as a regional Dialogflow CX agent is not covered by the sources below. Specific Control Hub deletion or rollback workflows for existing CCAI connectors and CCAI feature configurations are not covered by the sources below. Whether the manual CCAI onboarding provisioning form remains mandatory when a customer uses their own GCP project for billing is not covered by the sources below.

## Sources

[^1]: Access to the autonomous Webex AI Agent for voice calls is limited to specific customers. Source: [Use AI agents for customer interactions](https://help.webex.com/en-us/article/s0qro1), Prerequisites, important limitation. Checked 2026-09-30.
[^2]: Since 8 June 2026 Webex AI Agent supports a selection of non-English languages for voice and digital channels as generally available. Source: [What's new for administrators in Webex Contact Center](https://help.webex.com/en-us/article/a1gx3h), Entry dated June 8, 2026: Expanded Language Support for Webex AI Agent. Checked 2026-09-30.
[^3]: Webex AI Agents, scripted or autonomous, are built in Webex AI Agent Studio and brought into Webex Contact Center voice flows with the Virtual Agent V2 activity. Source: [Use AI agents for customer interactions](https://help.webex.com/en-us/article/s0qro1), AI agent types; Integration in contact center flows; Prerequisites. Checked 2026-09-30.
[^4]: The Contact Center AI Connector field offers three connector types: Webex CCAI, Integration Connector (third-party platforms such as Google) and Service App. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j), Procedure step 4. Checked 2026-09-30.
[^5]: A CCAI configuration can be set as the default for Virtual Agent conversations; the Agent Answers default option is not applicable to Webex Contact Center. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j), Procedure step 7. Checked 2026-09-30.
[^6]: When creating the Google conversation profile, the Enable virtual agent option under Choose to use Dialogflow must be turned on for use with Webex Contact Center virtual agent. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j), Procedure step 6, notes. Checked 2026-09-30.
[^7]: A Contact Center AI configuration is created in Control Hub under Services > Contact Center > Tenant settings > Integrations, Features tab, New > Contact Center AI Config, with a unique Config Name and optional Description. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j), Procedure steps 1-3. Checked 2026-09-30.
[^8]: The service-specific integration connector, for example the Google Contact Center AI connector, must exist before a Contact Center AI configuration can be created. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j), Prerequisites. Checked 2026-09-30.
[^9]: For the Google CCAI connector, the CCAI configuration takes a Google Conversation Profile ID, obtained from the Agent Assist UI in Google Cloud Platform. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j), Procedure step 6. Checked 2026-09-30.
[^10]: After Save, a confirmation message appears and the configured CCAI card is shown on the Features page. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j), Procedure step 8. Checked 2026-09-30.
[^11]: Cisco lists CCAI feature regional availability as US, UK, Frankfurt, Australia, Japan, Singapore and Canada. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j), Supported solutions and features, regional availability. Checked 2026-09-30.
[^12]: Connector setup requires clicking Sign In with Google and authenticating with the same Google ID that was used to create the CCAI project, then granting the requested data access permissions. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Control Hub configuration, steps 4-5. Checked 2026-09-30.
[^13]: The Google CCAI connector is set up in Control Hub under Services > Contact Center > Tenant Settings > Integrations > Connectors by clicking Set Up or Add More on the Google Contact Center AI card. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Control Hub configuration, steps 1-3. Checked 2026-09-30.
[^14]: To use the customer's own GCP project, the CCAI Project Name and the Billable Project Name selected in the connector must be the same. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Control Hub configuration, note after step 10. Checked 2026-09-30.
[^15]: Cisco lists completion of the Assessment to Quality (A2Q) process for CCAI eligibility as a prerequisite of the Google CCAI connector. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Prerequisites. Checked 2026-09-30.
[^16]: Cisco sends updated GCP credentials in an email titled CCAI Onboarding Provision Status Update, and the billable GCP project name used in the connector comes from that notification. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Prerequisites; Control Hub step 8 (billable project). Checked 2026-09-30.
[^17]: The Google CCAI connector prerequisites include a Cisco subscription Flex SKU and a Contact Center order placed through Cisco Commerce Workspace. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Prerequisites. Checked 2026-09-30.
[^18]: Before configuring the Google CCAI connector, a GCP project ID and a Dialogflow agent application must already exist in Google Cloud Platform. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Prerequisites, first bullet. Checked 2026-09-30.
[^19]: The connector prerequisites include submitting Cisco's CCAI provisioning form to map the customer GCP project ID to Cisco's service account and sharing the project ID and email with a Cisco representative so a primary service account is set up for billing. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Prerequisites. Checked 2026-09-30.
[^20]: In the connector form the administrator enters a connector name, selects the CCAI Project Name, the Billable Project Name and the billable service account from dropdowns, then clicks Save and Refresh to see the new connector. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Control Hub configuration, steps 6-10. Checked 2026-09-30.
[^21]: Webex Contact Center recognises three Dialogflow CX custom parameters set under Fulfillment > Parameter Presets: wxRecognitionModelName, wxRecognitionModelVariant and wxCallCompletion. Source: [Configure custom parameters in Dialogflow CX](https://help.webex.com/en-us/article/cq52gi/Configure-custom-parameters-in-Dialogflow-CX), Configuration location; Custom parameters supported. Checked 2026-09-30.
[^22]: A Dialogflow CX custom event for Webex Contact Center is a Custom payload with Execute_Request containing Event_Name and Data.Params, and Cisco instructs defining only the custom payload with no other dialogue options. Source: [Configure custom events in Dialogflow CX](https://help.webex.com/en-us/article/x18bao/Configure-custom-events-in-Dialogflow-CX), Custom payload configuration. Checked 2026-09-30.
[^23]: To resume the conversation after a custom event, the route's Transition is set to the same page, and an event handler with the event name is created in the Dialogflow CX flow. Source: [Configure custom events in Dialogflow CX](https://help.webex.com/en-us/article/x18bao/Configure-custom-events-in-Dialogflow-CX), Looping configuration; Event handler creation. Checked 2026-09-30.
[^24]: The custom event name is mapped to the State Event Name in the Virtual Agent V2 activity in Flow Designer for decision mapping. Source: [Configure custom events in Dialogflow CX](https://help.webex.com/en-us/article/x18bao/Configure-custom-events-in-Dialogflow-CX), Flow Designer integration. Checked 2026-09-30.
[^25]: The Dialogflow API must be enabled on the Google Cloud project, which requires the serviceusage.services.enable permission typically held through the Owner role. Source: [Setup and cleanup | Dialogflow CX](https://docs.cloud.google.com/dialogflow/cx/docs/quick/setup), Enable the API. Checked 2026-09-30.
[^26]: Regional Dialogflow CX locations keep data at rest in that region, while the global location stores data at rest in the US. Source: [Regionalization and location settings | Dialogflow CX](https://cloud.google.com/dialogflow/cx/docs/concept/region), Data residency; global vs regional. Checked 2026-09-30.
[^27]: Google documents the live agent handoff response as a signal to the detect-intent caller that the conversation should go to a human, used by Dialogflow CX only for measurement and not altering session state, with a payload structure of the integrator's choosing. Source: [Fulfillments | Dialogflow CX](https://cloud.google.com/dialogflow/cx/docs/concept/fulfillment), Response messages, Live agent handoff. Checked 2026-09-30.
[^28]: A Dialogflow CX agent's location cannot change after the agent is created; changing region requires exporting the agent and restoring it into a new agent. Source: [Regionalization and location settings | Dialogflow CX](https://cloud.google.com/dialogflow/cx/docs/concept/region), Location immutability. Checked 2026-09-30.
[^29]: Dialogflow CX agent locations include us-central1, northamerica-northeast1, us-east1, us-west1, us, europe-west1, europe-west2, europe-west3, europe-west4, europe-west6, australia-southeast1, asia-northeast1, asia-south1, asia-southeast1, asia-southeast2 and global. Source: [Regionalization and location settings | Dialogflow CX](https://cloud.google.com/dialogflow/cx/docs/concept/region), Available regions table. Checked 2026-09-30.
[^30]: Dialogflow CX can return queued responses as a partial response on the streaming APIs while a long-running webhook is processing, when Return partial response is enabled. Source: [Fulfillments | Dialogflow CX](https://cloud.google.com/dialogflow/cx/docs/concept/fulfillment), Partial response. Checked 2026-09-30.
[^31]: Dialogflow CX custom payloads must follow the JSON format defined by the integration, may reference parameters inside double quotes, and are limited to 24 levels of JSON depth. Source: [Fulfillments | Dialogflow CX](https://cloud.google.com/dialogflow/cx/docs/concept/fulfillment), Response messages, Custom payload. Checked 2026-09-30.
[^32]: A Google conversation profile has a display name and language, and is connected to a Dialogflow CX agent by choosing the profile, selecting Link with agent and picking the agent. Source: [Conversation profiles | Dialogflow CX](https://docs.cloud.google.com/dialogflow/cx/docs/concept/conversation-profiles), Create; Link with agent. Checked 2026-09-30.
[^33]: If a Google Cloud project is deleted, the Dialogflow CX agents linked to it cannot be recovered. Source: [Setup and cleanup | Dialogflow CX](https://docs.cloud.google.com/dialogflow/cx/docs/quick/setup), Create a project. Checked 2026-09-30.
[^34]: Google controls Dialogflow resource access by granting roles to principals, including Dialogflow API Client and Dialogflow API Admin, and service accounts are owned by projects. Source: [Setup and cleanup | Dialogflow CX](https://docs.cloud.google.com/dialogflow/cx/docs/quick/setup), Authentication; Access control. Checked 2026-09-30.
[^35]: A deployment that needs non-English voice self-service has, on Cisco's current documentation, a documented language path on Webex AI Agent but not on the Dialogflow CX Virtual Agent-Voice integration (inferred). Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Limitations (read with the June 8 2026 What's new entry). Checked 2026-09-30.
[^36]: Because agent location is fixed at creation and Cisco uses one conversation profile per PSTN region, the Google location decision should be made before building the agent and repeated per region for multi-region tenants (inferred). Source: [Regionalization and location settings | Dialogflow CX](https://cloud.google.com/dialogflow/cx/docs/concept/region), Location immutability (read with Cisco regional media steps 1-2). Checked 2026-09-30.
[^37]: The CCAI config ID is copied from the configured Contact Center AI card under Control Hub > Services > Contact Center > Tenant Settings > Integrations > Features. Source: [Configure regional media for Virtual Agent-Voice](https://help.webex.com/article/necavmp/Configure-regional-media-for-Virtual-Agent-Voice), CCAI config ID retrieval. Checked 2026-09-30.
[^38]: To handle Virtual Agent-Voice calls from multiple PSTN regions, a Google conversation profile and a CCAI configuration are created for each region and entry point mappings are configured with a PSTN region and mapped to a single flow. Source: [Configure regional media for Virtual Agent-Voice](https://help.webex.com/article/necavmp/Configure-regional-media-for-Virtual-Agent-Voice), Procedure steps 1-4. Checked 2026-09-30.
[^39]: The multi-region flow uses the NewPhoneContact.PSTNRegion event output variable in a Case activity, a Set Variable per region that assigns that region's CCAI config ID to a flow variable, and Virtual Agent V2 set to Variable Contact Center AI Config using that variable. Source: [Configure regional media for Virtual Agent-Voice](https://help.webex.com/article/necavmp/Configure-regional-media-for-Virtual-Agent-Voice), Procedure steps 5-10. Checked 2026-09-30.
[^40]: The tech note creates the Dialogflow CX agent at dialogflow.cloud.google.com/cx with Create Agent > Build your own and selects the location Global serving, data-at-rest in US. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Dialogflow CX agent creation, steps 1-5. Checked 2026-09-30.
[^41]: Billing must be enabled on the Google Cloud project because the Dialogflow, Speech-to-Text, Natural Language and Text-to-Speech APIs used by the virtual agent require an active billing account. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Google Cloud project setup section. Checked 2026-09-30.
[^42]: Cisco tech note 221526 states that the Google Contact Center AI connector is typically preconfigured for Dialogflow CX integration and that creating a new connector requires Cisco Project permissions. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Webex Control Hub: Google CCAI connector section. Checked 2026-09-30.
[^43]: The tech note creates the conversation profile in the Google console under Agent Assist > Conversation profile > Create, enables the virtual agent by selecting the Dialogflow CX agent, and copies the Conversation Profile ID. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Conversation Profile in Agent Assist, steps 1-7. Checked 2026-09-30.
[^44]: On the Escalated path the tech note parses VirtualAgentV2.MetaData with $.dialogflow.ccai.live-agent-escalation.<key> into a flow variable, branches with a Case activity and sends each branch to its own Queue Contact activity, with the Case Default output routed to one queue. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Configure Escalated Path, steps 1-10. Checked 2026-09-30.
[^45]: To return data on the Handled path, the tech note adds a Custom payload of the form Execute_Request.Data.Params on a route that transitions to End Flow, and the flow reads it with a Parse activity on VirtualAgentV2.MetaData using the expression $.Params.<name>. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Estimate_Details page step 4; Configure Handled Path step 2. Checked 2026-09-30.
[^46]: To escalate, the tech note adds a Live agent handoff dialog option on the route with a payload under the key dialogflow.ccai.live-agent-escalation and transitions to End Flow, which takes the call out of the Escalated output of Virtual Agent V2. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Create additional intents for routing, steps 5-7. Checked 2026-09-30.
[^47]: Cisco tech note 221526 instructs increasing Termination Delay from 3 to 5 seconds in Advanced Settings, implying a default of 3 seconds (disputed). Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Configure Virtual Agent V2 block, step 3. Checked 2026-09-30.
[^48]: The tech note assigns the published flow to an entry point and tests by calling the number mapped to that entry point, checking greeting, data capture and that live agent handoff reaches the right queue; the Dialogflow CX Test Agent simulator is used to test pages before that. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Customize initial greeting step 6; Entry Point configuration; Testing and verification. Checked 2026-09-30.
[^49]: With Dialogflow CX, Webex Contact Center supports partial response, custom event, custom parameters and custom payload. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Supported features with Dialogflow CX. Checked 2026-09-30.
[^50]: Global_Language and Global_VoiceName global variables set the input language and output voice for Virtual Agent V2, and can be overridden with Set Variable activities placed before the activity. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Conversational Experience, global variables. Checked 2026-09-30.
[^51]: Cisco's Virtual Agent-Voice article states that en-US is currently the only supported language. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Limitations. Checked 2026-09-30.
[^52]: The Virtual Agent V2 activity has three outcome paths: Handled when the virtual agent execution completes, Escalated when the call must go to a human agent, and Errored for any error. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Outcome Paths. Checked 2026-09-30.
[^53]: The Virtual Agent V2 activity outputs VirtualAgentV2.TranscriptURL, VirtualAgentV2.MetaData (JSON received from the bot during fulfillment or custom event handling) and VirtualAgentV2.StateEventName. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Output Variables. Checked 2026-09-30.
[^54]: The Virtual Agent V2 activity cannot be placed after a Queue Contact activity and cannot be placed in the Event Flow canvas. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Limitations. Checked 2026-09-30.
[^55]: Before configuring the Virtual Agent V2 activity with Dialogflow CX, the Google CCAI connector and a CCAI feature configuration (which generates the config ID used in flows) must exist, along with an entry point and routing flow. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Prerequisites. Checked 2026-09-30.
[^56]: The Virtual Agent V2 State Event settings have optional Event Name and Event Data (JSON) fields that send a custom event to the AI platform, and accept {{ variable }} expressions. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), State Event Settings. Checked 2026-09-30.
[^57]: The Virtual Agent V2 activity selects its CCAI configuration either statically from a dropdown of configured CCAI features or through a Variable Contact Center AI Config chosen at run time, for example by caller PSTN region. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Conversational Experience section. Checked 2026-09-30.
[^58]: Non-configurable Virtual Agent V2 defaults include infinite retries on invalid or no input, barge-in enabled, # as DTMF termination, a 5-second no-input timeout and a 3-second inter-digit timeout. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), System-level defaults. Checked 2026-09-30.
[^59]: The Virtual Agent-Voice article gives Termination Delay a range of 0-30 seconds with a default of 30 seconds (disputed). Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Advanced Settings, Termination Delay. Checked 2026-09-30.
[^60]: The virtual agent transcript appears in Agent Desktop only if Enable Conversation Transcript is on and an Agent Says fulfillment response is configured in Dialogflow CX. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Advanced Settings; Limitations. Checked 2026-09-30.
[^61]: Virtual Agent-Voice supports only the u-law codec. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Limitations. Checked 2026-09-30.
[^62]: Webex Contact Center system limits allow 6000 active entry points, 1000 active outdial entry points, 15000 dialed numbers configured as entry points in total and 500 dialed numbers per entry point. Source: [System limits in Webex Contact Center](https://help.webex.com/en-us/article/n7s6ed7), Table 1. Maximum system limits for configuration entities and attributes - entry point and dialed number rows. Checked 2026-09-16.
[^63]: Cisco's Dialogflow CX technote lists four Google APIs used by the integration (Dialogflow, Cloud Speech-to-Text, Cloud Natural Language and Cloud Text-to-Speech). Their usage is not free and needs a billing account. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Requirements. Checked 2026-10-01.
[^64]: The Virtual Agent-Voice article's Limitations section states that en-US is currently the only supported language. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Limitations (Dialogflow CX). Checked 2026-10-01.
[^65]: The Virtual Agent-Voice Limitations section states that only the u-law codec is supported. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Limitations (Dialogflow CX). Checked 2026-10-01.
[^66]: Webex AI agents built in AI Agent Studio reach voice calls through the Virtual Agent V2 activity in Flow Designer. Digital channels use the AI Agent node in Webex Connect flows instead. Source: [Use AI agents for customer interactions](https://help.webex.com/en-us/article/s0qro1/Use-AI-agents-for-customer-interactions), Voice channel and Digital channel sections. Checked 2026-10-01.
[^67]: Cisco states that autonomous AI agents on the voice channel are currently limited to specific customers. Source: [Use AI agents for customer interactions](https://help.webex.com/en-us/article/s0qro1/Use-AI-agents-for-customer-interactions), Voice channel > prerequisites / note. Checked 2026-10-01.
[^68]: Google CCAI connector onboarding requires a Cisco Commerce Workspace order, completion of the Assessment to Quality (A2Q) process and a CCAI provisioning form that maps the customer's GCP Project ID to Cisco's service account. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Prerequisites. Checked 2026-10-01.
[^69]: In Cisco's Dialogflow CX example, a Live agent handoff response with a custom JSON payload (for example dialogflow.ccai.live-agent-escalation with Type_Of_Agent) makes the call leave the Escalate output of the Virtual Agent V2 activity. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Configure > Dialogflow flow design > What_Kind_Of_Agent page (live agent handoff). Checked 2026-10-01.
[^70]: Dialogflow CX's Live agent handoff response only signals to the detect-intent caller that the conversation should go to a human. Dialogflow uses it for measurement and does not alter session state or perform the handoff itself. Source: [Fulfillments | Dialogflow CX](https://cloud.google.com/dialogflow/cx/docs/concept/fulfillment), Response messages > Live agent handoff. Checked 2026-10-01.
[^71]: AI Agent Studio transfer actions have a toggle: when on, the AI agent tells the customer about the transfer before executing it; when off, it performs a silent transfer. Source: [Webex AI Agent Studio Administration guide](https://help.webex.com/en-us/article/ncs9r37/Webex-AI-Agent-Studio-Administration-guide), Actions > Custom transfer action > announce transfers toggle. Checked 2026-10-01.
[^72]: In Webex AI Agent Studio the Agent handover action is enabled by default, allowing the AI agent to escalate the conversation to a human agent. Source: [Webex AI Agent Studio Administration guide](https://help.webex.com/en-us/article/ncs9r37/Webex-AI-Agent-Studio-Administration-guide), Actions > Agent handover. Checked 2026-10-01.
[^73]: Custom data passed from Flow Designer through the Virtual Agent V2 State Event settings is referenced inside an AI agent's goal, welcome message, instructions and action or slot descriptions with {{variable name}} syntax. Source: [Configure custom data and custom events for AI agents](https://help.webex.com/en-us/article/n5uo60x/Configure-custom-data-and-custom-events-for-AI-agents), Custom data. Checked 2026-10-01.
[^74]: Custom data for autonomous AI agents is currently supported only on the voice channel. Source: [Configure custom data and custom events for AI agents](https://help.webex.com/en-us/article/n5uo60x/Configure-custom-data-and-custom-events-for-AI-agents), Custom data > limitation note. Checked 2026-10-01.
[^75]: To use the customer's own GCP project with the Google CCAI connector, the CCAI Project Name and the Billable Project Name must be the same. Source: [Configure Google CCAI connector](https://help.webex.com/en-us/article/9zvadfb/Configure-Google-CCAI-connector), Set up the Google CCAI connector > procedure note. Checked 2026-10-01.
[^76]: Before using Virtual Agent-Voice the administrator must configure the provider's integration connector (for example the Google CCAI connector) and create a Contact Center AI (CCAI) feature in Control Hub, which generates the CCAI config ID used in the flow. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Prerequisites. Checked 2026-10-01.
[^77]: A Contact Center AI configuration is created in Control Hub under Services > Contact Center > Tenant settings > Integrations, on the Features tab, with New > Contact Center AI Config. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j/Create-a-Contact-Center-AI-configuration), Create a Contact Center AI configuration > procedure. Checked 2026-10-01.
[^78]: In Cisco's worked example, the Webex Contact Center AI Config for Dialogflow CX references a Dialogflow conversation profile ID that has the CX agent assigned to it. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Configure > Virtual agent creation and Webex integration steps. Checked 2026-10-01.
[^79]: Cisco instructs administrators not to place a Virtual Agent V2 activity after a Queue Contact activity in a flow. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Limitations. Checked 2026-10-01.
[^80]: Cisco instructs administrators not to use the Virtual Agent V2 activity in the Event Flow canvas. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Limitations. Checked 2026-10-01.
[^81]: In the Virtual Agent V2 activity Conversational Experience settings, a Static Contact Center AI Config uses the default PSTN region while a Variable Contact Center AI Config selects the CCAI config dynamically, for example based on the caller's region. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Virtual Agent V2 activity > Conversational Experience settings. Checked 2026-10-01.
[^82]: The Enable Conversation Transcript check box on the Virtual Agent V2 activity lets Agent Desktop display the transcript of the conversation between the virtual agent and the caller. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Virtual Agent V2 activity > Advanced Settings > Enable Conversation Transcript. Checked 2026-10-01.
[^83]: To change the virtual agent conversation language, a Set Variable activity sets the Global_Language variable to the required language code (the documented example is fr-CA) before the Virtual Agent V2 activity. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Virtual Agent V2 activity > Conversational Experience settings (custom language and voice). Checked 2026-10-01.
[^84]: To change the synthesized voice, a Set Variable activity sets the Global_VoiceName variable to the required output voice name code, for example en-US-Standard-D. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Virtual Agent V2 activity > Conversational Experience settings (custom language and voice). Checked 2026-10-01.
[^85]: At system level the Virtual Agent V2 activity enables barge-in, allows infinite retries for invalid or no input, and uses # as the DTMF termination symbol. Administrators cannot change these. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Default system-level settings. Checked 2026-10-01.
[^86]: The Virtual Agent V2 system-level no-input timeout is fixed at 5 seconds and the inter-digit timeout at 3 seconds. Administrators cannot change either. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Default system-level settings. Checked 2026-10-01.
[^87]: Flow validation checks expressions for structural errors only and cannot evaluate a function or check that a variable resolves to the expected value. Source: [Build and manage flows with Flow Designer](https://help.webex.com/en-us/article/nhovcy4/Build-and-manage-flows-with-Flow-Designer), Flow Designer, validate a flow. Checked 2026-09-15.
[^88]: Flow Designer undo and redo covers the last 10 canvas actions. Source: [Build and manage flows with Flow Designer](https://help.webex.com/en-us/article/nhovcy4/Build-and-manage-flows-with-Flow-Designer), canvas actions, undo/redo. Checked 2026-09-30.
[^89]: For Webex AI agents, the voice set through Global_VoiceName must be supported by the engine that the AI agent uses. Source: [Use AI agents for customer interactions](https://help.webex.com/en-us/article/s0qro1/Use-AI-agents-for-customer-interactions), Voice channel > Virtual Agent V2 activity language and voice settings. Checked 2026-10-01.
[^90]: The Virtual Agent V2 activity takes its Handled outcome when the virtual agent execution completes. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Outcomes > Handled. Checked 2026-10-01.
[^91]: The Virtual Agent V2 activity takes its Escalated outcome when the call must be escalated to a human agent. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Outcomes > Escalated. Checked 2026-10-01.
[^92]: The Virtual Agent V2 activity's Errored output is the path the flow takes in any error scenario. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Error Handling > Errored. Checked 2026-10-01.
[^93]: Cisco's example reads virtual agent data in the flow with a Parse activity that applies a JSON Path expression (for example $.Params.Estimate_Agents_Count) to VirtualAgentV2.MetaData and stores the result in a flow or global variable. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Configure > Webex Contact Center flow configuration (Handled and Escalated paths). Checked 2026-10-01.
[^94]: Two unrelated expression syntaxes are used in the same activity configuration: the Pebble double brace form for flow variables and a JSON path form for parsing a response body. Source: [Build and manage flows with Flow Designer](https://help.webex.com/en-us/article/nhovcy4/Build-and-manage-flows-with-Flow-Designer), Flow Designer, Parse activity and HTTP Request content type. Checked 2026-09-15.
[^95]: Inference: because the Escalated outcome only exits the activity and the Dialogflow handoff signal performs no transfer, an 'escalation never reaches the queue' symptom is most likely in the flow after the Escalated outcome (Parse, Case and Queue Contact wiring) or in a missing handoff or transfer action on the bot side (inferred). Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Configure > Webex Contact Center flow configuration, read together with google-dfcx-fulfillment Live agent handoff and the n6gaghu Outcomes section. Checked 2026-10-01.
[^96]: In Cisco's example, the Escalated path parses the handoff payload into a flow variable and uses a Case activity to route to the matching Queue Contact activity. The Escalated outcome itself does not choose a queue. Source: [Configure Dialogflow CX Virtual Agent](https://www.cisco.com/c/en/us/support/docs/contact-center/webex-contact-center/221526-configure-dialogflow-cx-virutal-agent.html), Configure > Webex Contact Center flow configuration steps 4 and 5. Checked 2026-10-01.
[^97]: The output variable VirtualAgentV2.StateEventName stores the name of the custom event that the system receives back from the agent bot. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Activity Output Variables. Checked 2026-10-01.
[^98]: Flow Designer has no version control feature. The flow version number counts how many times the flow has been published, and version labels can be applied. Source: [Build and manage flows with Flow Designer](https://help.webex.com/en-us/article/nhovcy4/Build-and-manage-flows-with-Flow-Designer), flow versioning section. Checked 2026-09-30.
[^99]: In European regions the Webex AI Agent transparency disclosure is enabled by default, and administrators who disable it must acknowledge the risks and add audit comments before opting out. Source: [What's new for administrators in Webex Contact Center](https://help.webex.com/en-us/article/a1gx3h/What's-New-in-Webex-Contact-Center), What's New entry 'AI Transparency Compliance: Disclosure Notice for Webex AI Agent' (July 31, 2026). Checked 2026-09-30.
[^100]: Outside European regions, Webex AI Agent administrators choose whether to enable the AI transparency disclosure and can configure its message; Cisco recommends enabling it as a best practice. Source: [Webex AI Agent Studio Administration guide](https://help.webex.com/en-us/article/ncs9r37/Webex-AI-Agent-Studio-Administration-guide), section 'Update autonomous AI agent profile', AI transparency settings. Checked 2026-09-30.
[^101]: Webex AI Agent can play an AI transparency disclosure on voice channels that callers cannot interrupt and that plays before the welcome message. Source: [What's new for administrators in Webex Contact Center](https://help.webex.com/en-us/article/a1gx3h/What's-New-in-Webex-Contact-Center), What's New entry 'AI Transparency Compliance: Disclosure Notice for Webex AI Agent' (July 31, 2026). Checked 2026-09-30.
[^102]: Inbound telephony entry points are created in Control Hub under Services, Contact Center, Customer Experience, Channels. Source: [Set up a channel](https://help.webex.com/en-us/article/ewuay1/Set-up-a-channel), procedure Create a channel, navigation step. Checked 2026-09-24.
[^103]: An inbound voice entry point is created with channel type Inbound Telephony and a routing flow selected from a drop-down, and its timezone defaults to the tenant timezone. Source: [Set up a channel](https://help.webex.com/en-us/article/ewuay1/Set-up-a-channel), procedure Create a channel, Channel Type, Timezone and Routing Flow fields. Checked 2026-09-24.
[^104]: Calls reaching the Webex Contact Center media edge must use G.711uLaw or G.711aLaw. Source: [Set up voice channels for Webex Contact Center](https://help.webex.com/en-us/article/nhy3codb), PSTN Media and Signaling Requirements for Webex Contact Center. Checked 2026-09-16.
[^105]: Inference: documented constraints that could each produce silence or a non-answering virtual agent are a missing CCAI config or connector, a tenant region outside the listed CCAI regions, a non-u-law media path, and placing the activity after Queue Contact or in the Event Flow (inferred). Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Prerequisites and Limitations, read together with npbt02j regional availability note. Checked 2026-10-01.
[^106]: The output variable VirtualAgentV2.MetaData stores the JSON data that the system receives from the agent as part of fulfillment, custom-event handling or a transfer action, and can drive later flow logic. Source: [Use AI agents for customer interactions](https://help.webex.com/en-us/article/s0qro1/Use-AI-agents-for-customer-interactions), Voice channel > Activity Output Variables > VirtualAgentV2.MetaData. Checked 2026-10-01.
[^107]: When a call is transferred to a live agent, the virtual agent transcript appears in the Transcript gadget on Agent Desktop, which requires the Transcript gadget to be configured in the desktop layout. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Conversation transcript in Agent Desktop. Checked 2026-10-01.
[^108]: The Virtual Agent V2 activity's Termination Delay accepts 0 to 30 seconds with a default of 30. It lets the virtual agent complete its last message before the activity stops and the flow moves to the next step. Source: [Configure Virtual Agent-Voice in Webex Contact Center](https://help.webex.com/en-us/article/n6gaghu/Configure-Virtual-Agent-Voice-in-Webex-Contact-Center), Virtual Agent V2 activity > Advanced Settings > Termination Delay. Checked 2026-10-01.
[^109]: Cisco lists Virtual Agent and Agent Answers as available in the US, UK, Frankfurt, Australia, Japan, Singapore and Canada. Source: [Create a Contact Center AI configuration](https://help.webex.com/en-us/article/npbt02j/Create-a-Contact-Center-AI-configuration), Create a Contact Center AI configuration > regional availability note. Checked 2026-10-01.
[^110]: Webex Contact Center PSTN, sold by Cisco as a bundle with Webex Contact Center, is available only to customers in the contiguous United States and Canada. Source: [Set up voice channels for Webex Contact Center](https://help.webex.com/en-us/article/2dputx/Set-up-voice-channel-for-Webex-Contact-Center), PSTN options section, Webex Contact Center PSTN. Checked 2026-09-30.
