Dialogflow CX: a new way to design conversational agents

Google yesterday launched Dialogflow CX in beta, a new system created for companies that need larger and more complex projects.

I tried Dialogflow CX, and here are some considerations.

• A visual builder that allows you to view conversational flows also appears on Dialogflow CX.

• The conversation is managed by “states” and transitions, and this offers excellent control of flows.
• The concept of “Page” is introduced. The pages manage the operations that the user performs within a flow: they represent the “states”.

• The “state handlers” are the elements that check the state at each interaction and establish if a message must be given, if a transition must take place, or other.
• Fulfillment expands. It no longer allows only the call of a webhook or of Cloud functions. In fact, it also allows you to send replies (even in a queue) and to modify the parameters of the conversation.

• Dialogflow CX also includes sentiment analysis.
• Machine Learning settings can be configured for each flow. For example, you can choose a standard or advanced type of NLU. The advanced one creates a more accurate model, but takes more time to train.
• As with Actions Builder, intent management has been simplified to make them a more reusable resource.

• Entity management remains faithful to Dialogflow ES. The slot filling phase also remains similar; it is called “form filling” and is associated with the pages.
• The mega agents are replaced by a great flexibility in managing flows, which can also be developed by different teams.

• The layout is very reminiscent of the Action Builder, even though the concepts are different. Maybe there are too many? Contexts, scenes, flows, pages …
• The system is in beta stage, so it does not have all the features, such as integrations.
• The Dialogflow that we knew until yesterday, today is known as Dialogflow ES (Essential).

The official documentation

https://cloud.google.com/dialogflow/cx/docs