What are chatbot failures?

Let's consider the scenes in which chatbot operation is perceived as a failure.

The problem that tends to occur after the introduction of chatbots is that they are not used. The number of users does not increase even though you went through a lot of effort to introduce a chatbot. It is not uncommon to see a certain amount of usage due to the novelty of the chatbot after its release, but after a certain period of time, the usage of the chatbot declines.

Inaccurate answers provided by chatbots are also a serious problem. The user's problem is not resolved. It takes too long to reach an answer. Inconsistent answers to questions. Under these conditions, it is no wonder that the system is no longer being used.

Failure to contribute to operational efficiency as expected is also a possible cause of failure. The number of inquiries to the manned response does not decrease. There are too many operational problems, and cost-effectiveness has not been achieved. Not only is the response rate too low to be of any use, but complaints are also being generated, which is a real downfall.

These are problems that should not exist in business chatbot operations. In the next section, we will discuss specific examples of failures and their solutions.

Failure Cases and Solutions in Call Centers

No reduction in incoming calls and no effect seen

As long as the operation is successful, the introduction of chatbots in call centers can have a great effect. However, there are cases where chatbots are introduced for the reason that there are many simple inquiries, but the number of incoming calls does not decrease. The most likely cause is a misreading of the customer base. If many of the target customers are seniors who are unfamiliar with the Internet, they will be confused by the chatbot's operation and input.

It is also possible that the visibility of the chatbot embedded in the site is low. In this case, there may be a lack of efforts to expand awareness.

Efforts for Improvement

In order to increase the usage rate of chatbots, it is necessary to consider online measures, such as easy-to-understand screen displays for seniors and enhanced input support. It is important to improve the layout to match the user's flow line and make it visually easy to understand.

It is also a good idea to thoroughly announce the chatbot and emphasize the advantages of its ability to resolve issues more quickly than call centers.

Failures and solutions in online stores

Resolution rate does not increase

In an example of a manufacturer's online store that introduced chatbots to reduce the number of e-mails and incoming calls, the usage rate increased and the number of inquiries decreased, but the resolution rate and customer satisfaction did not improve.

In such cases, even if the accuracy of the answers is good, the explanations may be long and difficult to understand. In cases where there is a lot of content that cannot be conveyed in writing, it may be difficult for users to understand even if the answers are correct.

Efforts for Improvement

First, review the existing FAQs, and if there is room for improvement, change to simpler text and expressions. If the chatbot is not answering the question well, create a branch and add an FAQ. If the text alone is difficult to understand, it is necessary to direct the user to an image page with illustrations or photos. If the answers are multifaceted, scenarios and categorization should be created to increase the percentage of correct answers.

Failures and Solutions for Internal Portal

There was a gap between management and frontline thinking.

A common problem in companies is a discrepancy in the image of AI chatbots between management and front-line operations staff. Some senior management may assume that AI chatbots are all-purpose. However, operational personnel who know the reality of the situation are thinking of operating in certain areas, which causes discrepancies.

To meet management's expectations, it is necessary to convert all kinds of FAQs into data, but there is a risk that the accuracy of answers will be low due to excessive volume or lack of work. As a result, the system may fail to be completed and aborted due to lack of practical operation.

Efforts toward improvement

Company-wide promotion of understanding of AI chatbots is required. After gaining a common understanding, we will clarify the goal toward practical operation. It is necessary to determine the scope of work to be handled by the chatbot, identify high-priority "frequently asked questions," and so on. In addition, the initial scenario displayed by the chatbot and the menu will be reviewed in line with the goal setting.

Help Desk / Information Solution Failures and Solutions

Employees complained about inaccurate responses

If technical terms and abbreviations are not sufficiently registered in the dictionary, questions may not be answered. The question is being asked "because I don't understand," but the question itself is not being understood. This results in frustration for the user because a speedy resolution is not possible.

The problem may be on the chatbot side, such as difficulty in registering additional words in the dictionary, difficulty in tuning, and inability to fully handle shaky notation.

Initiatives for improvement

It is necessary to re-select a chatbot with excellent language processing capabilities and a provider that performs careful tuning. We will consider changing to a chatbot that can meet your company's needs by adding the following points to the selection criteria: "a service that allows easy addition to the dictionary" and "specifications that facilitate the PDCA cycle within the company.

Failures and solutions in back office

Details of required documents are not clear.

There have been cases where chatbots have been used to build systems for guiding employees through various procedures, but they have been unpopular because they do not understand the important parts of the system. As the story spreads among employees, the rate of inquiries to the person in charge will increase without using the chatbot to find out.

For example, since the system only tells the names of documents required for a procedure, details such as the specific place to obtain them and how to write them are not clear. No matter how much you try to find an answer, you will end up going back to the beginning again from the question branch. The number of man-hours required to ask a question is not enough to get information...etc.

Initiatives for improvement

We redesign the branches by prioritizing the most frequently asked questions based on the past question log data and working backwards. If the explanations are difficult to understand, it would be a good idea to make PDFs available so that they can be opened for visual confirmation as well. We will collect feedback from employees as needed and make modifications as we identify specific areas for improvement.

If the accuracy of responses to scenario-based chatbots is low, consider switching to AI-powered chatbots. Since it is possible to ask back and narrow down the answers, it is possible to design a branching design that does not add too many layers and does not go too deep. If there are many complaints such as, "I get very little information for the amount of time it takes to ask questions," this will lead to improvement.

Failures and Solutions in Web Marketing (Customer Service)

Chatbot responsiveness not leading to customer satisfaction

We have seen cases where chatbots have been introduced to handle reservations for hotels, restaurants, travel, etc., but they are unable to properly understand the intent of the user's question. In addition, if the chatbot is not smoothly guided to a manned response when it is stuck in replying, it can lead to a decline in customer satisfaction. If the conversation does not progress, the user's latent needs cannot be deciphered, and appropriate suggestions cannot be made.

Initiatives for improvement

Since it is difficult for a scenario-based chatbot to provide a sophisticated response that leads to a proposal, consider switching to an AI-based chatbot, which can be improved to provide appropriate suggestions as its learning accuracy is improved.

When reintroducing chatbots, it is recommended to choose a service with high natural language processing capability. In addition, we will strengthen our efforts to improve customer satisfaction by establishing a system to coordinate with manned responses.

Failure Cases are the Treasure Trove of Success Strategies

Chatbots are a promising technology that can solve a wide range of issues facing businesses. At this point, however, they are by no means a panacea. It is important to think about how to get the best out of chatbots, rather than demanding the same level of human interaction. If you are dissatisfied with the operation, identify the causes of failure and take remedial measures one by one. If it is difficult to improve the current chatbot, consider migrating to a chatbot that is better suited for your purposes to ensure the success of your chatbot operation.

You do not want to make the mistake of introducing a chatbot that does not solve problems or does not improve customer satisfaction. There are multiple factors that can cause failure, as discussed in this article. It is ideal to know them in advance and avoid failures, but it is also important to choose a chatbot that can be improved while in operation and respond to changes in the situation. HUE Chatbot" has a highly operable management screen and is characterized by its easy PDCA cycle. HUE Chatbot" can be said to be easy to improve for issues. Please take a look at the introduction materials.