
Organizations that have built and tested Dialogflow CX flows do not have to discard or rebuild them when adopting generative agents in CX Agent Studio. Because Google describes CX Agent Studio as the evolution of Dialogflow CX, an existing flow can be added directly as a flow-based agent while a new generative agent handles open-ended questions. For the Professional Agentic Architect exam, this architecture comes down to deciding when to preserve deterministic flows rather than rewriting them, understanding how control and variables pass between an agent and a flow, and configuring pages, forms, and routes inside the flow accurately.
When a CX Agent Studio agent combines open-ended conversation with a flow-based agent, control transfers between them at clear boundaries. The CX Agent Studio agent hands the conversation to the Dialogflow CX flow until the flow reaches END_SESSION, and then the conversation returns to the agent that started it.
Values move between the initiating agent and the flow through input and output variable mappings. This design lets a generative agent answer open-ended customer questions while a tested flow handles structured steps, without forcing either component to duplicate the work of the other.
Google recommends Dialogflow CX flows for deterministic logic, such as sequential data collection and validation, while generative agents suit open-ended conversation. Flow-based agents let one CX Agent Studio agent use both. A tested flow already enforces its steps in a fixed sequence, and CX Agent Studio invokes it directly, so that logic carries over as it is rather than being rebuilt.
This pattern sits on the customer-facing side of Google Cloud's agent options. CX Agent Studio is the evolution of Dialogflow CX and a low-code builder, alongside code-first frameworks such as ADK and the employee-focused Workflow Builder. Google lists Dialogflow ES as a legacy conversational agent product.
Inside a Dialogflow CX flow, a page is the unit of state, and its form lists the parameters the agent must collect at that stage of the conversation. Forms give a flow native slot filling, so collecting a set of values is something the flow is configured to do rather than something built in code. Each page carries its own form, so a multi-page flow can collect a different set of values at each stage.
Each required parameter in a form has an initial prompt, and the page keeps prompting for any required parameter that is still empty. Once every required parameter is filled, $page.params.status becomes FINAL, and a route that checks for that status can move the session to the next step. Because the form itself tracks which required values are still missing, the flow gets this behavior without any custom logic.
Routes in Dialogflow CX can be attached either to an individual page or to the flow itself, and a route's scope depends on both where it is attached and what triggers it. The scope rules are worth knowing precisely:
Event handlers follow their own logic. A no-match handler runs when the user's input matches nothing in scope, which makes it the place to respond to unrecognized input. Together these rules decide which routes and handlers are available at each turn, and an intent requirement is what keeps a flow-level route available on every page of a flow.
Our Professional Agentic Architect practice exams test CX Agent Studio flow-based agents alongside generative customer agents and Dialogflow CX form and route design, with detailed explanations that drill these distinctions.