Twilio Changelog | Mar. 31, 2025

Twilio Virtual Agent with Dialogflow CX adds Session Resumption and Action Callbacks for Last Matched Intent and Sentiment

TLDR; Resume Sessions and control Google Dialogflow/Gemini Generative AI “Conversational Agent” Virtual Agent using Twilio’s Studio/TwiML Programmable Voice platform capabilities; preserve call context and capture important structured data from customer conversations with AI Agents.

Twilio’s <Connect><Virtual Agent> connector, TwiML and updated Studio Widget now support Session Resumption via a Session ID, for up to 30 minutes after a session has been initiated with Dialogflow (aka Google Gemini AI Conversational Agents). This cements Twilio Studio’s role as the central, flexible orchestration platform for Virtual Agents , where developers can easily create and start sessions with a Dialogflow Virtual Agent, pause that session, then meanwhile do useful things like send a 2FA text message or make a PCI-compliant Twilio <Pay> payment with the caller, and then resume the session between the caller and the Virtual Agent right where that caller left off – even with the additional call context of the messaging/payment step they just performed – without the caller having to repeat info or renavigate the AI Agent! See a demo video of Session Resumption in action, here.

At the end of the session with the AI Agent, or indeed at any call state transition with the bot, the Twilio Platform and Studio can also be configured to send webhooks (Action Callbacks) – in addition to turn-by-turn Status Callbacks already configurably – to the developer’s higher-level or downstream application (for example, a call center platform/agent application) with the contents of that session, such as last intent in the AI Agent matched by the caller, or the caller’s sentiment score at that last matched intent.

By thus allowing the preserving and passing along of call context from the virtual agent session to human agents and downstream applications, this enables both human agents to handle escalated calls more efficiently, as well as virtual agent to more effectively handle similar calls or calls from the same caller more effectively on subsequent calls.

Both of these new capabilities can layer on top of Google’s latest Generative AI capabilities and models – four methods of incrementally wading into the world of Generative bots for voice self-service automation, plus a new, fifth GenAI-based feature recently released in Dialogflow CX.:

Generative Fallback
Generators
Generative Playbooks
Data Store Agents
(new) GenerativeAI-based secondary language Bot training phrase generation/translation

These GenAI-based modes and enhancements complement and are in addition to Dialogflow’s more traditional Predictive AI-based flows, which are deterministic, and based on detected natural language spoken “intents,” via training phrases entered (or, now, with the above new feature, generated).

How can you take advantage of these new capabilities of Twilio Virtual Agent with Dialogflow?

Session Resumption and Action Callbacks are included in the functionality and per minute price of Twilio Virtual Agent with Dialogflow CX. Pausing a session is done in Dialogflow; configuring to resume a session in progress done in the newly added tab of the Connect Virtual Agent Studio Widget (or equivalent TwiML). Get started by following our easy one-click integration Onboarding Guide here.

Learn More: Resources

Voice Voice