What is a Chatbot?

First, some basic information about chatbots.

Overview of Chatbots

Chatbot is a coined word that combines the words "chat" and "bot," and is translated in Japanese as an "automated conversation program. Chatbots are used for online real-time communication by automated programs to respond to customer inquiries and internal help desks.

Chatbots can perform certain tasks and operations, and when responding to inquiries, they automatically respond to text and voice input without human intervention.

Although often confused with AI as a similar advanced technology, AI and chatbots are two different things. While chatbots are software that conducts human-like "conversations" using text and voice, AI learns and infers on its own based on data. Today, AI-based chatbots are becoming more mainstream in order to improve the accuracy of responses.

History of Chatbots

The history of chatbots can be traced back to "Eliza," which was created in 1966. At that time, it was a simple function that automatically provided simple "yes/no" responses to text input by humans. Chatbots began to be available in Japanese around 1980.

One of the reasons chatbots became common was the "Office Assistant" in Office 97, a help function where a character would answer questions about Office software functions.

As time went on, chatbots also appeared in smartphones, including the much-talked-about iOS Siri, an AI voice-activated chatbot that was introduced in 2011. In 2016, LINE and Facebook Messenger released APIs for chatbots. Voice-activated chatbots such as Cortana, Alexa, Google Chat, and many others have appeared, and are now very commonly used in daily life.

Background of Chatbot Attention

The spread of chatbots is largely related to the widespread use of the Internet. As online customer interaction has become more common, the need for chatbots has increased, and so has their attention.

Telephone and e-mail communication, which used to be the mainstream of customer support, is costly in terms of manpower, time, and effort. In order to increase customer satisfaction in today's increasingly busy society, real-time response has become more important, but there is a limit to the amount of human resources that can answer what you want to know faster and more accurately. Chatbots are becoming a reliable solution to questions and queries.

Latest Trends in the Chatbot Market

To what extent are chatbots impacting the economy?

The chatbot market is expected to grow at a CAGR of 23.5% from US$2.9 billion in 2020 to reach US$10.5 billion by 2026. If this forecast is correct, the market will grow at a rapid rate of approximately 3.6 times in six years. Among the factors contributing to the growth of the chatbot market are advances in technologies such as analytics, AI, and cloud computing. AI, in particular, is playing an important role in improving the accuracy of responses, as mentioned earlier.

For more information on AI chatbots What is an AI chatbot? Explaining in detail from basic knowledge to business utilization effects and introduction. for more information on AI chatbots.

Chatbots are becoming more accurate every day and can be used in a wide variety of industries. They are becoming as accurate as business demands and are establishing themselves as a trusted partner.

The rapid growth of chatbots is also due to a change in customer behavior. In recent years, an increasing number of consumers are becoming more self-service oriented and are not afraid of doing things on their own. With online shopping, Internet banking, and other services that touch all aspects of our lives, it is only natural that more and more consumers want to make their inquiries online.

At the same time, there is a need to provide customers with a personalized experience in order to resolve their inquiries. At the same time, the demand for chatbots as a contactless service has skyrocketed in the wake of the Corona disaster. In order to respond to Corona-related inquiries, municipalities have begun to introduce automated chatbot responses.

The trend to promote telework is also driving the use of chatbots. Chatbots are being adopted as a means of responding to customers without relying on in-house personnel, or as an alternative to in-house help desks.

Even public organizations are using chatbots to provide services, such as the National Tax Agency's use of chatbots to answer questions related to income tax returns and year-end adjustments. The use of chatbots is expanding in accordance with the needs of society.

Chatbot Structure

This section explains how chatbots work, their structure and types.

Chatbot Structure

In chatbots, an "application" and a system called a "bot" are linked through an API (Application Programming Interface). The general mechanism is that the system interprets questions, generates and selects answers, and then returns them to the application via the API.

APIs are interfaces between different software, programs, and web services. Within the bot, information stored in a database is analyzed according to logic to obtain an answer. Therefore, the accumulation of data is important for the chatbot to provide correct answers.

Types of Chatbots

There are three main types of basic mechanisms used in chatbots.

1: Scenario type (rule-based)

The first type follows a pre-defined scenario and automatically responds to the user's questions. The conversation proceeds in a selective manner, with branching leading to a final answer. The system responds to the user's choice of questions, so if a question is not very common, it may be impossible to answer it. Because of the limitation of possible question content, this method is effective for standardized questions such as "frequently asked questions.

As for the scenario type What is a chatbot scenario? What is a chatbot scenario? for more information on AI chatbots.

2: AI-powered type (machine learning type)

The AI engine analyzes the question and instantly calculates the statistically most appropriate answer, thus establishing a conversation. It is more flexible than the scenario type and can respond to a wide range of questions. It can also infer needs from users' questions, and link them to the next proposal. In addition to support desk support, it can be used for a wide range of other purposes, such as making proposals to customers on e-commerce sites and making reservations on behalf of customers.

3: Hybrid type

This is a response method that combines automatic responses by chatbots and human responses. For inquiries that are complex or require a personalized response, a manned response can provide the most reliable answer. Although this method is not fully automated and requires additional staff to respond, there is no need to worry that a response will be delayed.

Differences by Chatbot Internal Structure

There are several internal mechanisms for "how to answer.

▶Choice Type

The choice type is a mechanism used in the scenario type (rule-based type). The question is narrowed down by asking the user to choose one of the options. The advantage is that chatbots can be easily created as long as the question branches are set correctly. It is also characterized by the ability to guide users according to a pre-defined scenario.

? Dictionary type ?

The dictionary type is a system in which a combination of "keywords" and "responses" are registered. The AI uses analogies from the keywords to understand the content of the question and derive an answer, utilizing the AI's natural language processing function to analyze the content of the question and carry on a conversation in response to the corresponding keywords. The disadvantage of this system is that it requires a large number of keywords to be registered in advance, which requires human labor and time.

Choice & dictionary type

This is a hybrid type of choice type and dictionary type. It makes good use of the best parts of both types and enables superior responses. However, the disadvantages of each of the two types also occur, so care must be taken.

▶ Log type

Human conversation data is logged and used as a base for machine learning. By memorizing conversations, the system analyzes the context of the user's input and provides responses. The more log data that is accumulated, the more natural the conversation becomes and the more accurate the answers become. Conversely, if there is little data to learn, accuracy cannot be improved, and preparation takes time. Currently, AI-based log analysis is being developed to achieve natural conversations.

Usage Scenes of Chatbots

In what situations are chatbots effective? We will explain the main usage scenarios.

Improving Internal Business Efficiency

Chatbots can contribute to improving operational efficiency by answering various inquiries within a company. The following are some of the specific business areas where chatbots can be applied.

・Application to internal portals and business systems
Help desk for network, system and other tools
・Business manual reference and document search
・Personnel and administration related procedures and questions
・Internal information sharing

Back-office (administrative) departments such as information systems, accounting, general affairs, and human resources respond to questions from employees to their respective departments, but they tend to respond to many of the same questions. Whenever an employee asks a question, the person in charge must stop his or her work to respond.

In addition, if the person in charge has to be in the office to get an answer, it can slow down the work of the employee who asks the question. With the introduction of chatbots, questions can be answered 24 hours a day, allowing employees to focus on their work without unnecessary frustration.

For routine questions and answers, it is reasonable to use chatbots. Chatbots can help you search through a wide variety of company materials to quickly find what you are looking for.

FAQs" are an efficient system for managing frequently asked questions, but there are cases where the FAQs are not being used as expected. In such cases, there is a way to increase the usage rate by linking chatbots and FAQs. If the answers can be found instantly by simply typing in simple keywords, employees will feel more comfortable using the system.

Customer Response Leading to Customer Satisfaction

When applied to customer service, the following tasks can be substituted.

User support for handling instructions, problems, etc.
Real estate property inquiries
・Personalized guidance for tourists
・Accommodation and restaurant reservations
・Health and nursing care related questions
・Admission information, school guidance, and online classes
・Support center

E-commerce sites, real estate, tourism and reservation sites, medical and nursing care facilities, education-related facilities, support centers, as well as government websites that receive many inquiries related to procedures.

Especially in support centers that receive a wide variety of questions and complex inquiries, the combination of chatbots and manned responses can lead to improved customer satisfaction.

Effects of Chatbot Implementation

The following are possible examples of the effects of introducing chatbots.

▶ Improved efficiency of internal and external portal site operations

It can operate 24 hours a day, 365 days a year, and can receive inquiries 24/7. The ability to respond to questions regardless of time allows for a high level of user satisfaction and the ability to handle a large number of questions at one time.

▶ Improving efficiency of customer support operations

It contributes to solving the problem of operator shortages at call centers and to reducing labor costs. In addition to labor costs, the reduction of man-hours can facilitate operations.

It eliminates delays in response due to call waiting time, which tends to occur in call centers, and leads to improved customer satisfaction. For users, it will reduce the hassle and lower the hurdles to making inquiries. In addition, since the history is recorded as text data, it is easier to grasp the specific opinions of users in numerical form. This is expected to be utilized for customer marketing.

Chatbots can be easily deployed on a variety of platforms. They can increase contact with customers and respond to questions in a timely manner without missing the moment they arise.

Major Causes of Chatbot Introduction Failures

Although chatbots have shown great promise in the business world, there have been many cases of failure among companies that have already implemented chatbots. The main reasons for chatbot implementation failures are as follows.

▶The purpose is not clarified

There are many failure cases where the company's issues could not be solved after introducing chatbots, and this is often caused by the "unclear purpose of introduction". It is dangerous to think that the current situation can be improved only by introducing chatbots. Chatbots can be very effective depending on where they are applied, but they do not solve everything. It is necessary to consider clearly and in detail what the purpose of introducing chatbots is and what kind of results they are intended to achieve.

▶ Lack of understanding of current situation

If you do not see an increase in user satisfaction after implementation, it is likely that the deployment area is not "chatbot-friendly". If the questions are too complex, it is difficult for chatbots to respond to them alone. Consider the effectiveness of introducing chatbots based on a thorough understanding of the content and current status of inquiries at the site.

▶ Misreading customer demographics

If the introduction of chatbots does not reduce the number of phone inquiries, you may have misread the affinity between your customer base and chatbots. If many customers are not accustomed to online operations or are uncomfortable with text-based responses, the introduction of chatbots will have the opposite effect.

▶ Not raising awareness

In some cases, customers are not aware that chatbot support is being provided, and they continue to inquire in the traditional way. The design for making contact with users may not be well designed, or the design of the traffic line may be unfriendly. It is necessary to design according to the behavioral patterns of users and at the same time, to devise a way to make announcements.

? Failure to choose a chatbot

A major cause of failure is the wrong choice of chatbot. For example, there are cases where a service with many parts to be managed by the company is chosen, and the actual operation does not go well and the company gives up on the service. There are also many cases where the performance of the chatbot does not meet expectations, and users are unable to get answers due to problems with response to Japanese language fluctuations and response accuracy.

For more information on issues related to Japanese language fluctuations and shaky notation, and points to improve the accuracy of responses, please refer to the following How to improve the accuracy of chatbot responses? Introducing solutions for natural conversation" "What causes chatbot operation failures? for more information on AI chatbots.

▶Insufficient scenario setting

In the case of scenario-based chatbots, scenario setting is the key to success or failure. Existing inquiry and question data must be thoroughly analyzed and designed to provide the correct branching path.

▶If there was a problem in operation, no improvement was made

After the introduction of chatbots, it is necessary to analyze the data obtained from the operation. If the PDCA cycle is inadequate, it may be difficult to continue chatbot operation.

? Absence or lack of skills of operation staff

This is a case of inadequate systems on the part of the company. Problems may arise due to a lack of understanding of chatbots or a lack of qualified personnel to monitor and maintain the chatbots during operation. If you are concerned about your internal human resources, you need to choose a service with high maintainability.

If you want to learn more about failure cases What are the causes of chatbot operation failure? Learn from common cases to succeed" "What are the causes of chatbot operation failures? for more information on AI chatbots.

Key Points of Chatbot Implementation and Operation

The following is an explanation of key points when introducing and operating chatbots.

Key Points at the Time of Introduction

▶Get a better understanding of how chatbots work and the types of chatbots

Asking chatbots to be versatile is also a cause of failure. If you understand how chatbots work, you will be able to determine which parts of your business can be handled by chatbots.

▶Set conditions and accurately select a chatbot that meets your company's objectives

To ensure that the chatbot will meet your company's needs, you need to determine and prioritize the criteria for selection. The flexibility of customization must also be confirmed, since the target audience may be expanded after the introduction of chatbots.

? Check the chatbot's track record

When selecting a chatbot, check what industries the chatbot has been used for and how it fits into your company's business.

? Check the strengths of chatbots and their rationale

As we have seen, there are several types of chatbots. Each chatbot service has a wide range of strengths and characteristics. We will confirm the strengths of each and the rationale behind them, and make a choice that will lead to reliable results.

▶Conduct a simulation of the effects of introducing chatbots

Before actually introducing a chatbot, we will conduct a series of simulations of its implementation effects. At this time, it is important to carefully consider the attributes of the target users and the nature of the issues. Visualize the return on investment and consider specific implementation effects from a cost perspective as well.

▶Define the scope of work to be performed by chatbots

While the separation of tasks is done at the time of introduction, we will reconfirm whether the chatbot is appropriate for that task after further consideration. For example, even within the same support center, there are cases where it is preferable to separate the scope of chatbot operations from that of manned operations.

▶Appoint a person in charge of operation

We will check the skills required by the company in light of the support system for the chatbot service, and assign a suitable employee as the person in charge of chatbot operation.

Key points of the operation system

▶Establish a system to absorb feedback from the field

The first step in the operation of chatbots is to thoroughly familiarize internal and front-line employees with the chatbot system. It is a good idea to explain the purpose of introducing chatbots and, if possible, to include specific numerical targets. We will share with them what they should do and what points they should keep in mind when operating the chatbot.

We will periodically check the level of awareness among users and monitor how they feel about using the chatbot. We will establish a system to constantly collect feedback from the field, mainly from chatbot operators, and use this data as material for improvement.

▶Run PDCA cycle

Analysis and improvement of usage will be the pillar of chatbot operation. We will strive to improve user satisfaction by modifying scenarios and rules as necessary, and establish a system that enables internal checks to ensure that the PDCA cycle is functioning properly, rather than leaving it up to the person in charge.

▶Define KPIs for chatbot operation

The effectiveness of chatbot operation is measured by the "response rate," "correct response rate," and "resolution rate. We will calculate our own target values using the manned response as a reference, and conduct evaluation and improvement toward that goal.

Leading to Success by Taking Failures as a Reference Point

Chatbots can be expected to be highly effective if operated properly. On the other hand, if your company's business is not well-defined and chatbots are not properly selected, it may end in failure. The most important factors to consider when choosing a chatbot are the accuracy of the answers to questions and the maintainability of the service. To increase user satisfaction, the answers to their questions must be speedy and to the point. In addition, the issues that emerge during operation must be addressed and strengthened through repeated improvements. We would like to make careful preparations while referring to examples of failures, and aim for a chatbot operation that makes a significant contribution to our business.

As we have discussed in this article, it is important to understand the current situation and clarify the objectives when selecting a chatbot. In addition, the PDCA cycle based on data analysis and having a person in charge of operation will bring you closer to success.

HUE Chatbot" is highly resistant to Japanese language fluctuations and shaky notation, so a high level of response accuracy can be expected. In addition, the usage status and points for improvement are visualized on the dashboard, which is intuitive and easy to operate. High operability is an important factor directly related to ease of operation and improvement. Please take a look at the following materials for an introduction.