Companies use Google Cloud Dialogflow ES to design conversational interface bots, like the bots which power Google Assistant. This integration allows Genesys Cloud to call Google Cloud Dialogflow ES bot actions in Architect call, chat, and message flows.

Google Cloud Dialogflow ES bots use Google machine learning. This machine learning implements natural language understanding (NLU) to recognize a user’s intent and then extracts prebuilt entities such as time, date, and numbers. With the evolving functionality of artificial intelligence tools, conversational interactions with computers are now mainstream. Contact centers are a natural progression into this world of virtual assistants.

When a customer can speak naturally, your company can better understand the customer’s intent and more quickly route the interaction to a highly skilled agent. The Google Cloud Dialogflow ES integration within Genesys Cloud enables customers to use NLU within inbound synchronous customer interaction flows.

Before you begin, review the Google Cloud Dialogflow ES documentation.

Create Consumer and Resource projects for multiple bots

Google Cloud Dialogflow ES only allows one bot per project. To see multiple bots, create a Consumer project, and then create the bots as Resource projects. Create a service account for the Consumer account and then provide that service account access to the Resource projects. For more information, see Using multiple projects.

Set the correct Bot language local in the Google Cloud Dialogflow ES console

Make sure to set the correct language locale for your bot in the Google Cloud Dialogflow ES console. Architect flows do not run in language codes only, such as “en.” Add the locale to the bot’s language. The locale includes the language and region code combination. You can also add multiple languages to make your bots multilingual. For more information, see Languages in Agent Settings.

Work with enabled environments for the Google Cloud Dialogflow ES bot

Best practice recommends, when you enable environments for a Google Cloud Dialogflow bot, that you do not remove intents or supported languages for the bot.

Implications of removing intents for a Google Cloud Dialogflow agent

If you choose an environment in the Call Dialogflow bot action and remove intents from the draft version of the agent, then the flow takes the failure path. This action occurs because the flow cannot reach the deleted intents. Therefore, even though the agent fulfills the intent, the flow takes the failure path.

To ensure that the Draft environment always includes the entire set of intents in the published environment, only add intents during the life cycle of an agent throughout various environments. Do not remove them. You can also export the agent configuration into a new agent and then set the intents as appropriate. Then, update the Architect flow to reference the new agent.

For example, if you treat Dialogflow agents in an add-only manner but want to remove an intent, create a new agent. Export the current agent’s configuration into a new Google Cloud Dialogflow ES agent, then remove intents and slots as appropriate. Then, reference the new agent from the Architect flow instead. Going forward, follow the add-only approach described previously.

Understand Google Cloud Dialogflow pricing

Google prices Dialogflow ES monthly based on the edition, pricing plan, number of requests, the total duration of audio processed, and the total duration of the interaction. For more information about Google Cloud Dialogflow ES pricing, see the Google Cloud Dialogflow pricing page.

Example use case

When a customer interacts through Architect IVR, the Dialogflow ES bot begins. The system asks the customer an open question, such as “How may I help?”.

After the customer responds, Dialogflow ES attempts to interpret the intent of the request and then decides the next step. For example, if the customer replies, “I want to place an order for delivery,” then Dialogflow ES prompts the user for the order and returns delivery status.

If Dialogflow ES cannot establish or understand the customer’s intent, the system routes the interaction to an agent.

After the task ends, Dialogflow ES asks if the customer needs any additional help. The customer can ask another question, request to speak to an employee, or indicate that they need no further assistance. If the customer needs no further assistance, the interaction ends.

If the customer wants to connect to an agent but faces a long wait time or the request is outside normal business hours, then the IVR routes the interaction appropriately.

For more information, see About the Google Cloud Dialogflow ES integration.