Dialogflow: agent building features and setup

Overview

Dialogflow is designed for building conversational agents across different channels, including intents, entities, and tests. Newer versions also support working with Gemini models. It is suitable when you need a question-and-answer agent or a defined conversation path, rather than just a raw model. Before publishing, check handoff to a human agent, what is stored from conversations, and how to test responses on your own scenarios rather than only on general examples.

Features

• Build conversational agents with intents and flows
• Evaluate and test agent responses
• Integrate Gemini models within Google's platform

Business uses

• Answer frequently asked questions on a website before handing off to an agent
• Extract inquiry details from a conversation and send them to a ticketing system

Getting started

• Build a test agent with a few intents and examples in Hebrew.
• Run test conversations and check that they hand off to a person when there is no match.
• Check which conversation fields are stored before connecting it to customers.

What to check

• A general conversational model may invent an answer if there is no approved source.
• Connecting to a live channel requires checking permissions and user identification.