Ultimate Guide to Google Dialogflow
A step-by-step guide to integrate AI solutions into business in order to reduce costs, improve automation, and offer a better user experience.March 24, 2021 | Big Data | Dialogflow | Google Cloud
In 2021, AI is all the rage, and it will continue to be so as businesses rush to integrate AI solutions into their business in a race to reduce costs, improve automation, and offer a better user experience. And one of the most common business implementations of AI is a conversational agent, also known as chatbots and virtual agents.
Today, conversational agents are more:
Acceptable: 1 out of 5 consumers is willing to purchase goods from a chatbot.
Cost-efficient: Deploying AI conversational agents can reduce time to provide customer service by up to 80% and reduce costs by up to 30%.
Widespread: 84% of eCommerce businesses are either actively working AI solutions into their business or have it as a top priority.
Useful: 40% of consumers use chatbots to look for offers and deals.
However, the main challenge isn’t the viability of conversational agents but with their implementation. The business value of conversational agents is well-established but making the most out of this technology is a bit more complicated.
Thankfully, there is a solution to the numerous complications of conversational agents, and it’s called Google Dialogflow.
Dialogflow is one of the most popular and feature-rich platforms for conversational agents. It makes building and deploying conversational agents far more simpler and eliminates the need for numerous tools and services as well as knowledge of artificial intelligence and machine learning. In fact, this platform does not have even require coding to get started.
What is Google Dialogflow?
Google Dialogflow is part of the vast Google cloud ecosystem of cloud-based services. It started out as a company called API.ai, specializing in natural-language-understanding and conversational AI. In 2016, Google acquired API.ai to form the backbone of its conversational agent platform. Today, Dialogflow stands out as one of the most well-built platforms for AI solutions and the only platform that you can use to build AI conversational agents, without any coding.
Although most of us think of text-based chatbots when the term conversational agent pops up, Dialogflow is capable of both voice and text-based agents. More importantly, the conversion between these two is quick and easy which means businesses do not have to invest significant time and money into these two channels separately. Furthermore, Dialogflow’s conversational agents are cross-platform which means the same agent can be deployed on Web, Android, and iOS.
The technology powering Dialogflow isn’t new at all but things have changed a lot in the last five years and conversational AI has become far more capable. Google has made major strides in the field including numerous major AI acquisitions, developing its own language model (BERT), and deploying its own digital assistant on over 400 million devices. The culmination of all this is Google Dialogflow.
How Google Dialogflow Works
At its core, Dialogflow is a Natural Language Understanding (NLU) engine paired with a powerful, visual builder that enables businesses to build and deploy smart virtual agents. Natural-language-understanding (NLU) and natural-language-processing (NLP) are both sub-fields of machine learning (which itself a sub-field of artificial intelligence) that are focused on teaching computers natural and life-like speech. Neither technology is new and in fact, older implementations such as speech recognition have existed for decades.
But Dialogflow has its own NLU framework. The framework is based on the following components:
Agents
Intents
Entities
Contexts
- asking a question
- finding the context in the current message
- finding the context in a previous message
Knowledge Base
An agent refers to the actual conversational agent that gets deployed and does the talking (Google Dialogflow calls them Virtual Agents). When it comes to agents, Dialogflow offers a lot of customizability options including writing your entire agent from scratch or using one of its pre-built agents.
Intents refer to user intentions and it’s critical for any conversational agent to identify what the user wants. Intents can change quickly and can also change the conversation’s flow. For instance, if a user starts a conversation by saying “hello”, the agent will identify the user’s intent as a greeting and will choose a suitable response. The process of choosing a response is done through “training phrases” or identifiable keywords or phrases in the user’s messages.
Dialogflow’s NLU engine works so well because it is constantly scouring the user’s messages for additional information. An entity is one such way. While intents specify the motivation (for instance, the user wants something), entities help define the parameters (for instance, what does the user want). By using both entities and intents, the Dialogflow agent is able to create a better understanding of the user request and give specific answers without asking too many questions.
Similar phrases have different meanings under different contexts, which means in order to give an accurate answer, the agent needs to understand the context. The conversational agent can define the context by:
Since Dialogflow creates AI conversational agents, it is capable of using machine learning to improve accuracy over time. This can be done through Knowledge Base (currently in Beta) which is a collection of all the knowledge documents (such as articles, FAQs, price sheets, product descriptions and catalogs, etc) that the developer enters in Dialogflow. The agent can then use this knowledge base to improve its answers.
Benefits of Google Dialogflow
No-coding Required
Telephony Support
Numerous Integrations
Visual builder
No hosting required
Language-support
Fulfillment
One of the main advantages of using Dialogflow over any other conversational agent platform is that it enables businesses to create extremely powerful and versatile agents without having to write a single line of code.
Furthermore, Google Dialogflow has a wide range of fully-functional pre-built agents that come with intents and entities pre-loaded. The only limitation is that these pre-built agents are only available in English and that it’s possible that the intents and entities may not be applicable to your exact business model.
Dialogflow’s voice-based agents aren’t limited to smartphones and web browsers but in fact, are capable of being accessed through a phone number. Dialogflow offers a range of useful features including native Interactive Voice Response (IVR) support, one-click integrations with supported telephony partners, and an upcoming feature called Phone Gateway (currently in Beta) that creates a telephone interface for your existing agents.
Another major benefit of using Dialogflow is the wide range of integrations it supports right out of the box. Some of the most popular platforms that Dialogflow conversational agents can be deployed to include Facebook Messenger, Slack, Google Assistant, Telegram, Skype, Twitter, Twilio, Kik. and Viber.
Unlike most other conversational AI platforms, Dialogflow does not compromise ease of usage for functionality. In addition to its powerful NLU engine, Dialogflow has a very user-friendly console that visualizes your conversational agent and how potential conversations can unfold. The visual builder is updated in real-time and makes it easier for decision-makers to understand the agent without going through pages of code.
Dialogflow is a completely cloud-based platform that means the business does not require any separate hosting for their conversational agent. More importantly, Google’s global data centers ensure low latency, consistent performance, and reliability at affordable costs.
Google Dialogflow’s conversational agents support more than 30 languages including English, Turkish, Danish, Dutch, French, German, Hindi, Polish, Ukrainian, Indonesian, Italian, Japanese, Korean, Norwegian, Portuguese, Russian, Spanish, Swedish, Chinese, and Thai as well as numerous other variants.
Fulfillment is another feature of Dialogflow that allows your deployed virtual agent to communicate with the cloud and achieve numerous functions including remotely generating all of the answers for the chatbot deployed on websites and smartphones as well as for those deployed through webhooks.
Google Dialogflow vs Other Conversational AI Platforms
Keeping aside its features (many of which are unique to Google Dialogflow), there are still many reasons why a business would choose Dialogflow over competitors. For instance, compared to other platforms like LUIS, Amazon Lex, and Watson, Dialogflow is the most user-friendly platform, has a useful Web Preview feature, as well as a free plan that is ideal for small-medium-businesses (SME) (and an affordable Enterprise plan too).
In addition to software features, Dialogflow also has real hardware differences, such as Tensor Processing Units (TPUs) which make a significant difference in machine learning projects such as conversational AI.
Related:The AI Behind Dialogflow - How Is It Different From Other Conversational AI Platforms
Differences between Google Dialogflow ES and CX?
Although there are only two variants of Google Dialogflow, the differences between the two can be confusing. There is a Dialogflow ES and a Dialogflow CX variant. Let’s start with Dialogflow ES.
Dialogflow ES
Dialogflow ES (short for Essentials) is the variant aimed towards small-to-medium businesses. It starts at $0.002 per request but has a free Trial version. Some of the features of the standard ES version are missing from the Trial version including Sentiment analysis. There is also a limit to how many requests can be made with the Trial version.
Dialogflow CX
Dialogflow CX is a more powerful version of Dialogflow that unlocks a host of additional features. Unlike ES, CX does not have a per-request cost - instead, it has a per-chat session cost which comes out to be $20 per 100 chat sessions. That said, all new customers receive a $600 credit for a free trial of Dialogflow CX.
Some of the main differences between ES and CX are:
- support for reusable intents and flows
- Visual Flow builder
- State-based data model
- Separate flows within an agent
- Number of agents per project (1 in ES versus 100 in CX)
Additionally, despite having more features, CX has a more-friendly and simplified interface that is better suited for creating complex conversational agents.
There is only one major limitation to using Dialogflow CX and it’s that it only supports English as of right now (but that will very likely change soon).
Business Value of an NLU-based Conversational Agent
There are hundreds of use cases and popular implementations of conversational agents across dozens of industries that make them worth many times the original investment. For instance, many businesses find conversational agents absolutely critical for providing customer service as NLU-based conversational agents ensure instant and 24/7 service which would be very costly with their human counterparts.
Additionally, conversational agents make personalization at scale extremely cost-efficient which is important to improve brand loyalty and customer satisfaction. And finally, conversational agents are an often overlooked sales channel as many conversational agents can have cart and checkout features and be deployed on popular messaging platforms such as Facebook Messenger and WhatsApp - making purchases even easier.
Therefore, the only difficult part of conversational agents is actually building them the way you want. Dialogflow provides everything a business needs but you’ll still need someone to put all of the features to use. This is where D3V, a Google-certified cloud-solutions provider, comes in. Businesses can use D3V’s talented Dialogflow engineers to create the most helpful customer service agent, the best shopping assistant, or the ideal sales representative. Book a free consultation with a Dialogflow expert today.
Author
Gopi Mishra
Director at D3V
Gopi holds a Master’s Degree in Computer Science from University of Texas at Dallas, specializing in Machine Learning and Artificial Intelligence and is a Google Cloud Certified Architect and Data Engineer. In his off hours Gopi loves swimming, writing,and Formula 1.
