
Choosing an agent builder on Google Cloud follows from two architectural questions: who the agent serves and who builds and maintains it. On the Professional Agentic Architect exam, builder selection usually comes down to separating internal employee agents from customer-facing conversational agents, and separating low-code tools managed by non-developers or analysts from custom agents written and maintained in code by developers. The three primary options are Workflow Builder in Gemini Enterprise, CX Agent Studio in Gemini Enterprise for Customer Experience, and ADK on Agent Runtime or Cloud Run.
Workflow Builder, the low-code builder formerly called Agent Designer, runs inside the Gemini Enterprise web app. It is built for agents that serve employees and are created and updated by teams with no developers. Users define and change agent behavior through natural language prompts or a visual editor rather than writing code.
Agents built in Workflow Builder can connect to sources such as Gmail and Google Drive, run on a recurring schedule for scheduled tasks, and be shared with people across the organization. Because it is designed for internal use inside Gemini Enterprise, Workflow Builder does not serve external customers over phone lines or web chat.
CX Agent Studio, part of Gemini Enterprise for Customer Experience, is designed for customer-facing conversational agents. Google describes it as the evolution of Dialogflow CX. Business analysts and customer service analysts can manage agents in CX Agent Studio through a low-code builder without relying on developers.
CX Agent Studio includes capabilities specifically built for customer interactions:
The boundaries of CX Agent Studio matter as much as its features. It is built for customer conversations rather than internal scheduled tasks. A web chat widget waits for someone to start a conversation instead of running on a recurring schedule, and internal employee agents do not need telephony or customer chat channels.
Organizations that have already invested in Dialogflow CX flows do not have to rebuild them when adopting generative agents. CX Agent Studio allows existing users to reuse their Dialogflow CX flows as flow-based agents while building new generative agents. This pattern appears on the Professional Agentic Architect exam when a customer-facing agent needs both generative responses and deterministic logic.
Google recommends flows for deterministic logic such as sequential data collection and validation. Rewriting an existing flow as instructions discards tested deterministic logic, and rewriting it in code with ADK creates a rebuild that teams can avoid. At runtime, the CX Agent Studio agent hands the conversation to the flow until it reaches END_SESSION, and then the conversation returns to the agent that started it. Input and output variable mappings pass values between the agent and the flow.
ADK on Agent Runtime or Cloud Run is designed for custom agents built, deployed, and maintained by developers in code. When developers own the agent rather than analysts or non-developers, we would lean toward ADK.
At the same time, ADK is a poor fit when non-developers need to update an agent on their own, because every change then depends on developers or IT. On the customer-facing side, building a custom ADK agent with its own telephony gateway requires developers and rebuilds features that CX Agent Studio already includes.
Several other options overlap with parts of agent design but solve different problems or add development overhead that a team cannot support:
When evaluating an agent architecture on Google Cloud, start by identifying who interacts with the agent and who will maintain it over time. Internal employee agents maintained by non-developers, including scheduled tasks and shared organizational agents, belong in Workflow Builder inside Gemini Enterprise. Customer-facing agents across phone and web chat maintained by analysts, especially when flows, human handoff, evaluations, traffic splitting, or Dialogflow CX reuse are needed, belong in CX Agent Studio. Custom agents that require developers to write, deploy, and maintain code belong on ADK running on Agent Runtime or Cloud Run.
Our Professional Agentic Architect practice exams test choosing an agent builder on Google Cloud alongside CX Agent Studio capabilities and Dialogflow CX flow reuse, with detailed explanations that drill these distinctions.