April 11, 2018

Chatbots and Dialogflow

Chatbot and Dialogflow

Why Should You Build a Chatbot?

The goal of a chatbot is to create conversation with your potential or existing clients. To make that possible, you want to build a conversational experience, which is any voice or text-based experience that uses natural language understanding to communicate with users. User interface (UI) needs to be designed and developed as if it was a human being carrying out the conversation. Before you do that, you need to understand what the intentions of your interlocutor are.

A statistic from E-marketer and Gartner shows that by 2019, 33% of US internet users will use voice assistants. By 2020, more than 85% of customers interactions will be managed without a human and by 2021, 87% of US B2C marketers believe chatbots and virtual assistants will play a significant role in the marketing world.

Voice-based interfaces are also very convenient to a great number of users that cannot actually interact with touch devices for many different reasons. As we all know, technology evolves along with culture and you definitely want to keep up with these changes while satisfying your customers’ needs, such as being reachable everywhere at any time, knowing their likes, and responding instantly to their requests.

The actual need for chatbots relies on two crucial factors:

  1. Increase in smartphone devices.
  2. Shift in brand engagement.


Pitfalls You Want to Avoid While Building a Chatbot

Here are some common reasons why people get UI wrong and cause a bot to fail:

  • Artificial intelligence (AI) is still not that accessible.
  • Use cases are not that strong.
  • Some bots lack transparency.
  • Bots don’t understand context.
  •  They don’t communicate with existing business systems.
  • They try to handle too many things at once.
  • They lack proper human escalation protocols.

*Source: chatbot.fall

Here is an example of a bot functioning improperly:




How to Develop Your Bot Strategy

First, you want your interface to be delightful so that users can have a wonderful experience with your brand. There are many aspects you can touch and enhance with your bot. In this article we are going to deal with 3 huge opportunities:

  1. Costumer support
  2. Transactions
  3. Getting something done

Often these activities will overlap with each other if the bot is well designed.

Thanks to bots, you can move a potential client down to the conversion funnel and make them perform a purchase.

As for customer support bots, there are a lot of benefits you can get from them. On the business side, they are cost-effective, lower time to resolution, and let agency spare time while handling complex cases or dealing requests that users could be solving themselves through a simple search.

Moreover, automation can be available 24/7 whereas physical customer support cannot. On the one hand, building a customer care bot will let you help your customer in many ways: from simple troubleshooting down to handling transaction information. This is if you design a well-structured, conversational workflow. A customer support bot can also help you with simple instructions requests. 

Help customers Instantly


Talking about transactions, a bot comes in handy if you’re selling a complex product or service that needs to be understood before clients are prepared to buy it. You can guide people through what product really suits their needs by designing a good conversational interface. Here are some transaction related activities that a bot can help with:

Make transactions fast and seamless


As for the third case, getting things done, you need to work consciously to obtain a rich interactive experience, which can let you increase engagement, enhance brand love, and give your customers new ways to interact with your brand.

Here are some examples of what kinds of activities you could use to engage your client:

Delight customer in fun and useful ways


Dialogflow Comes into Play

One platform that you can use to build a professional and well-designed chatbot is Dialogflow. It is a Google-owned developer of human-computer interaction technology based on natural language conversations. It has more than 450k developers and it is used in more than 150 countries all around the world.

How to get started with a chatbot:

  1. Have a clear goal in mind of what you are going to use the chatbot for, the reason why you need it, and what its objective is.
  2. Scratch out your chatbot UI by writing a user journey. This allows you to have a concrete thing to anchor on when it comes to building your bot.

Preparing a possible conversation is technically called “happy path implementation”, which is a way to describe the fact that you are making sure everything goes correctly while keeping the conversation with your users.

This is what the platform looks like:

Dialogflow Intents


In the Intents section (which refers to what people really mean while saying something), you can add anything related to your business that a customer might ask you about. You only need to enter 10 to 20 “training phrases”, as Dialogflow calls them, to get great accuracy matching users’ intent. This is because Dialogflow has tons of business data stored in the back end, thanks to which, it handles machine learning for you.

Here you are dealing with ambiguousness, so that if any information is missing, Dialogflow goes ahead with the conversation and follows up with the user asking for additional information.

The next part of the conversation that needs to be taken care of is the Response, where you instruct the bot on how to present contextual information to users. Most of the path that Dialogflow is following to create your conversation is pre-built. You get all the common ways of training phrases set up. In particular, Dialogflow counts with 30 different languages and their variations, such as American English, Australian English, and British English.

When you build your first bot, you don’t want to bump into details but, don’t be afraid to test it on users so you can see their reactions and get useful feedback from them 🙂

When you want the bot to fulfill something from the back end of your database, you need to use a Dialogflow feature called Fulfillment, which is a way to proceed and send users specific information already stored in the system.

This was meant to be a first look at what Dialogflow has to offer in order to build a professional and well-designed chatbot. Thanks to Machine Learning, you can get a powerful natural language understanding that is a key factor for your bot to be efficient and useful.

For further information, you can visit the Dialogflow website.