Iwasaki Tsushinki Co., Ltd. provides a service (voice bot) that automatically responds to inquiries made to call centers via voice. We spoke with a member of their Cloud Solutions Sales Department about the background and effects of adopting AmiVoice API Private.

Background and Challenges in Using Speech Recognition

Voicebot Service for Financial Institutions in Times of Disaster

A client, a financial institution, approached us with a request to launch a service that would automatically receive information on damage caused by disasters, and we decided to develop a voicebot service.
In the event of a disaster, there will be a flood of phone calls, which can lead to overflow calls (calls that cannot be answered due to a lack of operators or the number of calls exceeding the lines), and it is also expected that calls will need to be accepted outside of business hours.
The aim was to make it possible to respond to such situations without human intervention.

Initially, we considered proposing a voicebot on the public cloud, but there was nothing that met our requirements in terms of cost and functionality, so we decided to customize and provide our voicebot service, called IW-SmartAssistant.
To keep costs down, we decided to use cloud APIs for all of the telephone lines, speech recognition, and other components of the service, and adopted AmiVoice API Private, which uses a name engine for speech recognition.

IW-SmartAssistant management screen

Key Drivers of the Adoption

Reliable "Katakana Name Conversion" with High Recognition Accuracy

In this voicebot project, one of the key requirements was to recognize the caller’s name through speech recognition and display it in Katakana. Since each component was intended to be procured as a cloud service, solutions such as Google’s speech recognition were also considered.
However, our company has extensive experience using AmiVoice’s call‑center‑oriented package product, (AmiVoice Communication Suite), and we were already confident in its reliability and high recognition accuracy.
For that reason, when we considered the specific use case of “recognizing Japanese names and outputting them in Katakana,” AmiVoice immediately came to mind, and we were able to adopt it with confidence.
Another major factor was that the demo environment provided for our PoC allowed the customer to directly experience the recognition accuracy.

Impact of the Implementation and Future Outlook

Smooth Development Significantly Shortened the Initial Implementation Timeline

The AmiVoice API's client library and sample programs are available on GitHub, and by using these as reference resources, we were able to proceed with the implementation smoothly and quickly.
In addition, if any questions or concerns arose, there was an online support desk available, and we were able to receive thorough support.
Thanks to this, we were able to shorten the development period considerably from the original plan, which allowed us to devote sufficient time to dialogue with our customers and create a service that is in line with their business flow.

Automated Responses Improve Service Quality and Reduce Staff Workload

When a disaster occurs, there will be a large number of phone calls, but by using a voicebot to handle simple inquiries such as initial interviews, it will be possible to deal with overwhelmed calls and accept calls outside of business hours, which will lead to improved service quality for customers.
Additionally, by preparing voicebots in advance in anticipation of potential call surges, we believe that this will reduce the burden on call center staff and prevent them from panicking in emergencies.

Looking ahead, we plan to continue releasing communication‑related cloud services and APIs designed for call centers and office environments, following the launch of our communication‑cloud brand “Blue Commpaas” in April 2023. As a telecommunications equipment manufacturer with strong expertise in voice‑related services, speech recognition is an excellent fit for our business. Leveraging the development know‑how and experience gained through this project, we intend to actively incorporate AmiVoice into our future service offerings.