Comix Inc. (Headquarters: Shibuya-ku, Tokyo; Representative Director: Akihiro Suzuki) provided its Generative AI Utilization Consulting service to the executive in charge of DX promotion at a Tokyo Stock Exchange-listed company offering comprehensive guarantee services. As a result of their support, including the live implementation of an internal tool in approximately three weeks, they have succeeded in semi-automating the handling of 700-800 internal inquiries per month regarding guaranteed products.

By leveraging past Q&A data, a system was built where AI automatically generates draft answers, with final decisions made by humans, creating a "semi-automated response" mechanism.

Background and Challenges

With a wide range of guarantee products such as rent debt guarantees, medical expense guarantees, nursing care expense guarantees, and child-rearing expense guarantees, the increase in products and frequent system revisions lead to a surge of questions from various internal departments regarding these guarantee products.

These inquiries amount to 700-800 cases per month. Currently, the responsible department manually answers each one, placing a significant operational burden on initial responses alone.

On the other hand, data from past inquiries and their answers has been accumulated within the company. "The answers already exist within the company. Can AI be entrusted with the part of retrieving them?" ── This was the starting point for this initiative.

Services Provided

1) "Semi-Automated Response" Using Past Q&A x Spreadsheets x Claude Generation GAS

Given that existing Q&A data could be utilized, handling inquiries was positioned as the highest priority theme likely to yield early results. The following mechanism was proposed and its implementation supported:

- Past inquiry and answer data was converted into a simple database in a spreadsheet. - GAS (Google Apps Script) generated with Claude Code automatically imports inquiries received via email. - AI automatically generates draft answers by referencing the accumulated data. - Final confirmation and response are carried out by a human representative.

Instead of aiming for full automation immediately, AI's role is limited to "drafting answers," with humans retaining judgment and responsibility. This small-start approach to semi-automation allows for gradual expansion of the application scope while verifying accuracy and reliability.

2) CIO Personally Engages ── Approximately 3 Weeks of Support Record

Another characteristic of this initiative is that the executive in charge of promotion personally learned Claude Code.

- Initial Regular Meeting: Business challenges were inventoried, and automation of inquiry response was set as the top priority theme. - Individual Hands-on Session: Environment setup for the Claude Code desktop version. Participants experienced automatically generating a GAS-linked spreadsheet with a single prompt and adding a "summary result email sending function" through conversation alone. - Regular Meetings: The "Design First" approach was personally practiced, where design documents were created through dialogue with the chat AI and then input into Claude Code. A web survey form to collect internal business issues was implemented live on the spot.

Rather than simply being taught how to operate the tools, the goal is to master the "method of entrusting work to AI" itself. This is the support style emphasized by Comix's Generative AI Utilization Consulting service. By personally engaging and practicing collaboration with AI, it serves as a model case for driving AI literacy improvement across the entire organization.

Inquiries about Generative AI Utilization Support can be made here.

Inquiries regarding this article can be made here.

Features and Strengths

- Reduction of Rework through "Design First"

- Humans Solidify Specifications First: Requirements are organized through discussions with chat AI and translated into design documents.

- Implementation is Entrusted to AI: When design documents are provided to Claude Code, implementation proceeds without further questions.

- Minor Adjustments Completed Through Conversation: UI adjustments are made with natural language instructions such as "Make the button's clickable area wider."

This is not a simple handover. Humans design based on business challenges, AI swiftly handles implementation, and humans finally verify the operation. This appropriate division of roles significantly reduces development rework and maximizes the performance of generative AI. This enables even non-engineers to rapidly create tools directly linked to their practical work.

Target Users and Usage Scenarios

Target Users

- Management / CIOs of companies handling guarantee services and a wide range of products. - Heads of business departments overwhelmed by internal inquiries. - Information systems departments seeking to promote the application of generative AI in business.

Example Usage Scenarios

- Indirect Operations: Internal FAQ support, initial reception and draft answer creation for various application procedures. - Field Operations: Internal inquiries regarding product specifications. - Organizational Strength Enhancement: Asset creation and utilization of past Q&A data and manuals scattered throughout the company.

Future Outlook

- Construction and Internal Deployment of Semi-Automated Inquiry Response System: Rollout for company-wide use along with educational videos. - Centralized Management of Product Master Data: A plan to consolidate the current state of scattered product information, terms, and business flows into "a single source of trusted master data" by utilizing data output from the core system. - Service Offering to External Parties: The mechanism refined through internal inquiries is also being considered for future deployment as a customer-facing service.

The target company and Comix aim to improve the quality of guarantee services and accelerate response speed by connecting AI utilization to actual business operations starting with specific tasks that have a clear scope of 700-800 inquiries per month. The policy is to ensure that the introduction of generative AI leads to the transformation of business processes, rather than ending with "the installation of tools."

Inquiries about Generative AI Utilization Support can be made here.

Inquiries regarding this article can be made here.

ーーー Comix Inc. Company Overview ーーーーーー

Headquarters Address: Kondo Building 2F, 15-4 Maruyama-cho, Shibuya-ku, Tokyo

Representative: Representative Director Akihiro Suzuki

Established: September 2007

Business Activities: Generative AI utilization productivity improvement support, sales support for BtoB businesses, freelance professional matching, etc.

Web: https://www.comix.co.jp/

Comix Academy: https://www.comix.co.jp/academy/

【Inquiries】

TEL: 03-5459-5394 MAIL: [email protected] Contact: Comix Inc. Narumi Adachi

FACT BOX

  • Source: PR TIMES
  • Category: 技術導入