Hakky List of Products and Services

We would like to introduce a case where TAPP Co., Ltd. developed and provided "Akari AI," a conversational AI that supports the anxieties of investment beginners. The company wanted to incorporate AI into its sales process in line with the trends of the times and achieve a web customer service unique to the AI era. However, they were unsure about the specific implementation methods and how to integrate it into their actual sales flow, and they had concerns about the market perspective and approach to their first AI development. In response, our company developed a dedicated conversational AI that learned from actual consultation cases accumulated during the company's asset management seminars and the question trends of users. This AI enables customer interactions equivalent to those of in-house sales representatives, regardless of time or place, significantly contributing to the realization of new web customer service in the AI era. [Case Overview] ■ Challenges - Methods for realizing "web customer service in the AI era" - Specifics on how to integrate into the "sales process" - Concerns about "cost and feasibility" regarding first-time AI development ■ Effects - Successfully increased the conversion rate to about three times that of immediately after release, reaching the expected KPI. *For more details, please download the PDF or feel free to contact us.

We would like to introduce a case where we developed and provided an evaluation system for "contest application attractiveness" using generative AI for the Public Recruitment Guide Co., Ltd. The company was manually evaluating approximately 2,000 to 3,000 new public recruitment applications each month with a few staff members, which was a significant burden in terms of labor. Additionally, there were issues with inconsistencies in evaluation criteria depending on the reviewer, leading to fluctuations in evaluations. In response, we verbalized the subjective evaluation criteria through interviews and achieved a highly accurate public recruitment evaluation using AI. This allowed for the complete automation of evaluations for approximately 3,000 applications per month, resulting in evaluations that are more accurate and consistent than those made by humans, significantly contributing to operational efficiency. [Case Overview (Partial)] ■Challenges - To convey the enjoyment of public recruitment to users and revitalize the contest industry. - To enhance the added value on the site through automated evaluations by AI. - To manage the burden of evaluating approximately 2,000 to 3,000 new public recruitment applications each month. *For more details, please download the PDF or feel free to contact us.

We would like to introduce a case study of a dialogue-based manga recommendation service developed using generative AI for Shueisha Inc. The company was considering a service to recommend over 5,000 manga titles they own, tailored to users' preferences. We provided an early prototype for the challenging requirement of defining a conversational experience. This allowed for a smooth transition to production without significant setbacks. We incorporated essential character information and the desired conversational experience into the prompts, enabling unique responses. [Case Overview] ■ Challenges - Considering a service to recommend manga based on user preferences. ■ Results - Reached a total of 200,000 recommendations within two months of launch. - Achieved unique response expressions through character-specific prompt engineering. *For more details, please download the PDF or feel free to contact us.

We would like to introduce a case study of a Proof of Concept (PoC) for diagnostic support AI that assists with diagnostic discrepancies in a beauty clinic for Virginia Corporation. The company was considering incorporating diagnostic support AI into their existing system, but they were unable to achieve practical accuracy through in-house development. Therefore, they defined requirements based on the actual counseling flow and successfully built an AI that meets the needs of the field. To properly train the AI on diagnoses that can lead to differing judgments, they conducted multiple interviews to redefine annotation standards that could be understood even by non-medical professionals. As a result, they succeeded in improving the accuracy of the AI model without increasing the amount of data. [Case Overview] ■ Challenges - They want to incorporate diagnostic support AI into their cloud business support system in the future. - They want to support the diagnostic discrepancies arising from the experience of physicians. - They could not achieve practical accuracy through in-house development. *For more details, please download the PDF or feel free to contact us.

We would like to introduce a case study of developing a data analysis chatbot for television production at Data Stadium Inc. The company was considering a system that would allow for easier analysis and was interested in value provision that combines a data infrastructure with a large language model. Therefore, we developed a tool that can acquire data related to baseball. This tool can display the desired information in a suitable format in text form, eliminating the need for complex analysis instructions. We focused on usability, including features that address specific needs such as correcting inconsistencies in notation. 【Case Overview】 ■ Challenges - Considering a system for easier analysis - Desired development of a tool to acquire baseball data in chatbot format ■ Effects - Improved usability through prompt engineering *For more details, please download the PDF or feel free to contact us.

We would like to introduce a case study in which we built a customer support chatbot system utilizing a large-scale language model for O.M. Network Co., Ltd. The company has been deploying a shift management system, but due to the centralized approach at headquarters and reliance on a rule-based bot, they were unable to provide adequate individual support. By trialing advanced AI models and leveraging accumulated conversation history, we achieved high-precision user support. Additionally, we established a system within the company to implement product improvements. [Case Overview] ■Challenges - Inadequate individual support for actual end users - Desire to build a flexible chatbot to improve customer satisfaction *For more details, please download the PDF or feel free to contact us.

We would like to introduce a case study of Blue Mo Securities Co., Ltd. regarding the transition to a data infrastructure that balances security and data utilization. As the company’s services grew, the volume of data processing increased, revealing issues with the existing data infrastructure. It became difficult to consider dependencies, making it challenging to understand the impact of one change on other processes. In response, we implemented automatic masking of personal information, achieving immediate anonymization. We built an environment where data could be utilized safely while simplifying permission management, significantly reducing management workload while maintaining security standards in the financial industry. [Case Overview] ■ Challenges - The complexity of data processing dependencies made it time-consuming to respond to changes. - It was difficult to balance strict personal information management with data utilization and operations. - Ensuring safety and usability in collaboration with partner companies. *For more details, please download the PDF or feel free to contact us.

We would like to introduce a case study in which we developed a foundation and dashboard to utilize various data that was lying dormant within Progrit Inc. Due to the lack of an integrated data foundation, data aggregation necessary for documents was being done manually. Additionally, as this was their first time building a full-scale foundation, they were looking for a partner to assist with the overall design. In response, we quickly built the infrastructure and pipeline for the data foundation, as well as a dashboard for visualization. We designed a user-friendly dashboard that does not require high literacy levels and also supported the selection of tools to reduce operational burdens. [Case Overview] ■Challenges - Data aggregation necessary for various documents is done manually each time. - Cross-departmental data analysis is time-consuming. - Accumulated data is not being utilized effectively. *For more details, please download the PDF or feel free to contact us.