#Japan #AI #Tokyo #IVRy #conversation

Innovative AI Architecture: IVRy Secures Patent

IVRy, a company specializing in conversational AI technology, has announced that it has been granted a patent for a unique architecture that effectively separates the generation of follow-up questions in voice AI interactions. Located in Minato, Tokyo, the company, led by CEO Ryo Uenishi, aims to improve the engagement and efficiency of AI interactions in customer service settings.

This patent, classified under H04M, pertains to technology in telephone communications, centering around the automatic generation of follow-up questions based on contextual understanding. This technology aims to suppress hallucinations (misleading information generated by AI) while facilitating natural dialogue that aligns with the context of the conversation.

The Need for Better Interaction

In today’s customer service landscape—including stores and contact centers—there is a pressing need for AI agents and conversational AI to effectively handle ambiguous customer inquiries. Often, customer requests are vague or encompass multiple intents, demanding AI to autonomously ask clarifying questions before providing precise answers. However, traditional methods fall short when faced with unpredictable or undefined customer inputs, leading to incorrect responses and frustration.

To address this issue, IVRy has focused on creating reliable conversational AI services that combine multiple AI systems to minimize hallucinations. Nevertheless, the demand for more adaptive and flexible voice dialogue mechanisms has significantly increased. Conventional generative AI technologies struggle to balance the elimination of hallucinations with the necessity for spontaneous and context-sensitive responses.

The Unique Architecture of the Patent

The newly acquired patent introduces a dual-layer architecture for conversational AI, achieving both reliability and flexibility. Instead of having the AI generate the final response itself, the system only allows AI to create follow-up phrases to clarify customer intent. This innovative separation allows for faster intent identification, leading to quicker resolutions for customer inquiries.

Breakdown of the Technology

1. Reception and Intent Estimation: The AI comprehensively captures the customer's speech, accurately recognizing and estimating their intent. This expanded capability reduces the need for human intervention by autonomously identifying vague inquiries, thereby lowering operational costs.

2. Selection of Responses: The final answers are chosen exclusively from a defined dataset, eliminating the risk of hallucination by ensuring that the AI only draws from factually accurate information. This feature is particularly beneficial in high-stakes environments like contact centers, where misinformation can lead to trust issues and customer complaints.

3. Real-Time Follow-Up Generation: In scenarios with ambiguous intent, the AI dynamically crafts pertinent follow-up questions in real time, ensuring continuity in dialog despite unclear inquiries. Traditional AI systems often revert to generic responses or pass the dialogue to a human agent when faced with unexpected input. This patented technology simplifies the need for pre-programmed scenarios and reduces maintenance overhead.

This architecture allows the AI to engage in natural, contextually relevant dialogues while adequately addressing customer needs without information overlap.

Competitive Advantage of the Technology

IVRy’s patent, classified under H04M, fundamentally addresses the challenges faced by existing voice response systems and the current trend of using large language models (LLMs) in handling all dialogue. Traditional systems, often reliant on cumbersome hierarchical menus, delay accurate intent identification and may fail to respond appropriately when menus are bypassed. Although LLMs provide a comprehensive processing approach, they are plagued by related hallucination risks when user intents are unclear.

By maintaining a clear distinction between the intent identification and response selection processes, as well as real-time follow-up question generation, this innovative divide-and-conquer approach effectively reduces hallucination risks, advancing the field of voice response technology.

Commitment to Innovation

Ryo Uenishi, CEO of IVRy, emphasized the importance of such technology in enhancing customer service interfaces, stating that hallucination from AI poses significant risks. The newly patented technology for generating follow-up questions is instrumental in achieving a harmony between responsive dialogues and customer satisfaction. IVRy targets a robust application of this technology in the formal rollout of AI voice agents by 2026, aiming for increased operational trust and reliability.

IVRy is actively seeking engineers and specialists across various domains to join its team as the company continues to pioneer advancements in conversational AI.

For more details about career opportunities, please check this link. Additionally, learn more about IVRy’s conversational AI platform, ‘Aivry,’ here.

Company Overview

Founded in March 2019, IVRy, Inc. (Aivry), is located at 10F Sumitomo Realty & Development, Mita Garden Tower, 3-5-19 Mita, Minato, Tokyo. To learn more, visit IVRy.