Breakdown of chatbot costs

There are various costs involved in implementing and operating a chatbot. Let's take a look at the breakdown of costs for "introduction" and "operation.

Breakdown of costs for introduction

▶Initial cost
This is the cost when you sign a contract. Some services are free of charge, but in such cases, support for initial setup, such as maintenance of FAQs and support for creating answer scenarios, is often lacking. In the case of AI-type services, there may be an additional fee for initial training to ensure that the AI can understand fluctuations in notation and different phrases.

Even if there is no initial cost, if the support is not sufficient, it will be a heavy burden to introduce the system. Therefore, you must carefully assess not only whether or not there is an initial fee, but also whether or not the system can be operated smoothly after introduction. In particular, scenario creation and initial learning of AI are important factors for successful chatbot operation, so we recommend chatbots with full support even if you have to pay for it.

▶Customization cost
This is the cost for customization such as "changing the design," "displaying answers as images," "presenting multiple candidates," "linking to SNS," etc. The scope of customization varies depending on the service. The scope of customization varies depending on the service.

Cost breakdown for operation

▶ Monthly fees
There are two main types of fees: those that are fixed regardless of the amount used, and those that are charged according to the amount used.

▶ Customization cost
This cost is incurred when you want to change the design or make additional modifications to the scenario after the service is launched. These costs are also applicable when you want to add optional features such as a function that notifies the operator when a user asks a question to the chatbot, or a function to switch from the chatbot to an operator.

▶Consulting and customer support fees
Chatbots do not necessarily improve efficiency immediately after they are introduced. It is important to improve the accuracy of the answers as the operation continues, so the PDCA cycle will always be performed to ensure that users' questions can be resolved promptly. However, it is not easy to improve the accuracy of answers, as improvements must be made after aggregating and analyzing usage data. For this reason, paid support and consulting services may be used in some cases.

What are the estimated costs for introducing chatbots?

There are two main types of chatbots: scenario type and AI type. Scenario-type chatbots anticipate user questions and have pre-defined answers. On the other hand, AI-type chatbots are designed to provide natural, human-like conversation by having AI pick up keywords in response to open-ended questions entered by the user, understand the intent of the question, and return the most appropriate answer. Let's look at the general market prices for introduction and operation costs based on the assumption that these two types are available.

*For the difference between the scenario type and the AI type, please refer to " What is a chatbot scenario, and how it differs from the AI type and how to design it together ".

Cost of Chatbot Implementation

The initial cost, which includes customization and installation costs, varies from service to service, ranging from several hundred thousand yen to more than one million yen, although some are free. As a general rule, the more AI-type and highly customizable the service is, the higher the initial cost tends to be. However, there are cases where the initial cost is free but the monthly fee is high, or conversely, where the initial cost is expensive but the monthly fee is low, so it is important to consider the monthly cost as well.

Costs for Chatbot Operation

Most of the costs incurred after operating a chatbot are monthly fees, ranging from 50,000 to 100,000 yen per month for a scenario-type service, and from 100,000 to 500,000 yen for an AI-type service. Additional fees will be charged for customization and consulting services.

If you create your own chatbot

In addition to using an external service, chatbots can also be created from scratch or by using APIs. The following is an overview of the costs involved.

▶ If you create your own chatbot from scratch:
If you have an in-house programmer, you can create a chatbot with only human resources costs. However, it is necessary to perform all maintenance and management in-house, and human costs are required not only for production but also for operation.

▶Case of self-production using API
The following are examples of self-production of chatbots using API, which is an interface for linking software, programs, and web services.

Facebook Messenger
on Facebook and Instagram The advantage of this API for chatbots available is that it is basically free. Using this API, it is possible to automatically respond to posts from users on your Facebook page or Instagram. You can also send and receive images, videos, and voice memos with a dedicated application, but the disadvantages are that the places where you can use it are limited to specific SNSs and it cannot handle complex questions.

Google Dialogflow
This API is provided by Google. It is relatively easy to set up a chatbot on your own website. As a further advantage, chatbots can also work in conjunction with LINE, Skype, Slack, and other services. However, you will be charged based on the amount of usage. Another disadvantage may be that some plans do not support Japanese.

For more information on the advantages and disadvantages of in-house development or using an external service, please see " For Companies Worried about Introducing Chatbots! Questions to Consider When Making a Choice " for more information.

Points to consider when selecting a chatbot

So far, we have looked at the costs of both using an external system and creating your own. Although creating your own system is an option, we recommend that you use an external system if you consider the maintenance, customization, and operational support required. Therefore, we will explain the points to consider when selecting a chatbot based on the use of an external system.

▶Are the functions and services suitable for the purpose of using chatbots
Clarify the purpose for implementing a chatbot. For example, if the response is limited to questions such as inquiring about the location of a sales office or making reservations at a restaurant, not many functions are necessary.

On the other hand, if the purpose is to provide user support, such as explaining how to handle the company's products, services, troubleshooting, and functions, there are a wide range of questions that can be asked. Therefore, it will be difficult to respond to such questions unless the system is capable of setting up many scenarios.

In this way, setting up a chatbot with functions and services that match the purpose will make it easier to achieve a higher level of effectiveness.

▶Does it have a function to improve the accuracy of answers after operation?
Another important point in the selection process is whether or not the system has a PDCA cycle function and support system to improve the accuracy of responses according to user usage after the start of operation. Without this function and support system, customization and consulting costs may be incurred later, which may result in higher-than-expected costs.

▶ Whether programming knowledge is required or not
For example, if programming is required to create or update scenarios, you will have to outsource the work if you do not have the personnel in-house to handle it. This will incur extra costs and make it difficult to quickly create and update scenarios. Even if the chatbot is suited to your company's objectives, if it is time-consuming to operate, you will not be able to realize one of the advantages of chatbots, namely, increased efficiency. Non-programmable chatbots that are easy to manage are recommended.

▶Does it have ease of operation and efficiency?
It is also important to make sure that the registration and maintenance work of information necessary for the operation of the chatbot can be done smoothly. Specifically, consider the ease of viewing and operating the management screen. As mentioned earlier, it is also important to be able to manage the system with non-programming. However, even if the chatbot is non-programmable, if it does not have excellent operability, on-site work efficiency will suffer.

▶Does it have an adequate support system?
If the support system is not adequate, such as "inquiries are only made by phone and the response time is limited" or "it takes time to get a response," when a problem occurs, the response to users will be delayed and trust will not be gained. Therefore, be sure to check whether the support system is well-developed and can respond to problems promptly.

▶Don't make a decision based solely on initial and monthly costs.
If you make a decision based solely on the low cost of introduction, the accuracy of the answers may be low, and in the end, there may be cases where users' questions cannot be resolved without a human response.

Even if the cost is low, it will be difficult to improve efficiency with a chatbot that is limited in the number of questions and answers that can be set, and requires an additional fee for more than a limited number of questions and answers. In many cases, spending additional costs to provide a wider range of responses may end up being more expensive.

It is important to select a service that can achieve your objectives after considering cost-effectiveness, rather than selecting a service based solely on initial or monthly costs.

It is important to consider the introduction and operation of chatbots from a long-term perspective

Chatbots play a major role not only in responding to user inquiries but also in user support. However, if you do not select a service that is cost-effective, you may end up spending more than you initially expected. It is important to understand the cost of introducing and operating chatbots in advance. Then, it is necessary to know what the purpose of the introduction is and what functions are necessary for that purpose.

However, if you simply select a chatbot based on cost alone, you may end up selecting a chatbot that has "a small dictionary vocabulary" or "cannot answer questions with the same meaning if the phrasing of the questioner is different. Such a chatbot would have to register FAQs in every possible way, and as a result, the cost would be higher than expected. In this sense, we recommend that you select a chatbot after considering how effective it will be in the long run.

HUE chatbots are expensive to initially build, but the advantages of HUE chatbots are that they can be tuned, can respond to your company's unique language, and can be used with high accuracy. Also, operators can customize the chatbot design, for example, by changing the design of the chatbot from the management screen at no additional cost.

In addition, the technology developed at the Tokushima NLP Research Institute, a research and development institute specializing in natural language processing (NLP), is used to create chatbots that can respond to the unique "fluctuations" of the Japanese language. The number of registered vocabulary words is one of the largest in Japan (2.9 million), so even if you receive a variety of questions, you only need to register one FAQ to respond to it.

Please take a look at our introduction materials.