How to Create a Conversational CX Agent With Dialogflow
Conversational UX (CX) has taken the world by storm in recent years, and chatbots have become one of the most powerful forms of this user interface paradigm—so powerful that it’s now possible to create them with Google Cloud services like Dialogflow and Actions on Google. Dialogflow allows you to build Conversational CX agents from scratch, and this guide will show you how to do it step-by-step, giving you an understanding of what goes into creating great chatbots, along with building your own!
Introduction to Dialogflow
Dialogflow is an end-to-end conversational AI platform. It allows you to create and deploy chatbots, virtual agents, voice apps, or language understanding models that connect with your users through conversations. You can use Dialogflow's drag-and-drop interface to build conversational interfaces on top of your existing web or mobile app. Once you've built the app, you can publish it in the Google Assistant and Amazon Alexa marketplaces as well as in the Facebook Messenger Platform. And if you want to make changes later, updating them is easy too - just log into the API console and click refresh.
In addition, you can monitor conversations with Dialogflow Agents from the dashboard.
Creating Intents
The next step is to create intents. This is the what in I want to do something. For example, if I say I want to order an Uber, you would have an intent for that sentence. You can name these intents anything you want, but it's best to name them after what they are - in this case, order-uber. So click on Create Intent and then type in order-uber. Click on save and then create another one called check-status. Once again, go ahead and type in check status into the box. Hit save when you're done. Now scroll down and hit add cards. A card is an item of content that will be shown to your user. Here, we'll add 3 cards: -Order Status: To see the status of their order -Order History: To see their past orders -Contact Us: If they need assistance or have questions about their order. That's all there is to creating dialogs with Dialogflow!
Training and Testing Your Agent
1) Write your agent's intents and their corresponding phrases. This will make it easier for your user to understand how to interact with the bot.
2) Record yourself saying the intent phrases so that your agent can recognize them. This is called training the agent. You can train multiple users, too, if you want the bot to be able to differentiate between voices.
3) Test out your bot by chatting with it. If you don't have anyone else to talk with it, just chat with it using your own account. 4) Keep testing until you're happy with the results. You should also ask others to test out your bot and see what they think of it as well! Use A/B testing to compare different versions of your conversational CX agent to find which ones are performing best. Don’t forget that you can change or edit the parameters anytime, even after publishing it on Google Assistant or Facebook Messenger!
Integrating Your Agent With Other Platforms
Dialogflow can integrate with many different platforms. You can integrate your agent with Google Assistant, Amazon Alexa, Facebook Messenger, Slack, Twilio SMS and WeChat. This will allow you to connect your conversational agent directly with those platforms and all of the users who use them. For example, if someone asks your agent on Google Assistant what the weather is like in Los Angeles, it will be able to answer them by saying that Los Angeles has a high of 77 degrees today and is mostly cloudy. Additionally, if they ask for the forecast for tomorrow, they would be told that there's an 80% chance of rain tomorrow evening and a low of 67 degrees. There are other integrations as well! One allows you to generate up-to-date data feeds based on live data from Bloomberg or Reuters using their APIs. The other allows you to create chatbots that interact with text messages sent through SMS. That way, instead of having to call the customer service number for your airline every time you want to change a flight, you could just send them a text message. And when the customer service rep texts back asking what happened and why they needed to change their flight, your bot will tell them exactly what happened so that they know whether or not they need help rescheduling.
Managing Your Agent
A conversational agent or chatbot is created using an AI-powered Natural Language Understanding (NLU) engine, such as Google's DialogFlow. An NLU engine understands the context and intent of phrases in a text message and can reply with the appropriate response. For example, if someone says, I'm looking for new recipes, you might reply with Here are 10 recipes that use five ingredients. The following steps describe how to create a simple dialog flow agent:
1) Set up your project in the Dialogflow console
2) Add intents
3) Design your conversation flow by linking intents together
4) Test your conversation flow to make sure it works as intended.
5) Deploy your conversation flow on either GCP or AWS.
6) Integrate with other platforms to broaden its capabilities.
7) Monitor and analyze conversations between users and your agent to see what's working and what isn't
8) Repeat this process over time to build out more sophisticated dialogue flows with many different intents and contexts. 9) You'll also want to monitor your customer feedback so you know what types of messages they respond best to. 10) If you don't like something about the experience, go back and update it until it feels right!