mocomoco Successfully Develops Beta Version of New Generative AI Feature in 4 Months

A closed beta test for the new Knowledge AI Review feature of FRAIM Inc.'s next-generation AI cloud editor "LAWGUE", for which mocomoco had been commissioned to handle research and development since June, was conducted just 4 months after the project began.

About FRAIM Inc.

Under the vision of "Reinventing document creation," FRAIM Inc. provides "LAWGUE," a cloud document workspace utilizing the latest technologies, including AI. They aim to realize a world where workstyles are enriched through a next-generation document experience that allows for more accurate creation of the numerous documents handled in daily operations.

What is the Knowledge AI Review Feature?

The closed beta test for the new "Knowledge AI Review feature" of FRAIM Inc.'s "LAWGUE" service began in October, with a beta version scheduled for release by the end of 2024.

The Knowledge AI Review feature utilizes a company's internal knowledge and generative AI to automatically extract risks and generate proposed revisions.

For more details about this feature, please refer to the following article.

Next-Generation AI Cloud Editor "LAWGUE" Begins Closed Beta Testing for Knowledge AI Review Feature - FRAIM Inc.
https://lawgue.com/news/4785/

Project Participation

Commissioned with the research and development of the Knowledge AI Review feature, our company, under the leadership of FRAIM Inc., conducted R&D on algorithms using generative AI and developed a demonstration system using front-end technologies.

Based on the concepts, prototypes, and requirements shared by President Miyasaka and VP of Machine Learning Mr. Mizuno, we proactively investigated the latest research papers, proposed algorithms, comparatively verified various models, and developed the demonstration system. Subsequently, based on feedback obtained from hearings conducted by FRAIM Inc. with legal practitioners such as lawyers, we proactively proposed UI and internal algorithm improvements and implemented rapid modifications.

Because we were able to improve the demonstration system to easily collect feedback, and thanks to FRAIM Inc.'s precise instructions enabling high-speed improvement cycles, we reached the closed beta test in a short period of just 4 months.

Hurdles to Success

The difficult aspects of realizing this project were applying general-purpose LLMs to the legal field and accelerating processing speed while maintaining the performance of the AI review feature's algorithms.

Application to the Legal Field

Since our company lacked knowledge in the legal field, to facilitate smooth communication with FRAIM Inc., we proactively researched the necessary information. We read "Detailed Explanation of Legal Drafting, New Edition III" (Gyosei) by Masazumi Ishige to learn the formats and vocabulary of legal documents. Additionally, we used law school materials to study how legal professionals think regarding contracts.

Because the legal documents, such as contracts, targeted for development contained complex legal terms and clauses and were quite long, achieving sufficient performance using general-purpose LLMs required considerable ingenuity and trial and error. The frequent hearings with practitioners conducted by FRAIM Inc. and their shared feedback allowed us to set a clear direction for development.

Accelerating the AI Review Feature

To reduce the wait time from uploading a contract to displaying the review results, we devised various algorithm and implementation strategies, such as parallelizing processes. As a result, we achieved faster speeds without compromising review performance, with almost no impact on costs.

Technologies Used

AI Technologies: We implemented and comparatively verified diverse LLMs (API-provided models and open-source models) as well as classical natural language processing methods using Python.
Front-end Technologies: We used TypeScript, React, and Tailwind CSS.
Back-end Technologies: We used Python, FastAPI, and Uvicorn.
Cloud Technologies: We primarily used AWS, with partial use of Azure.

Comment from FRAIM Inc.

FRAIM Inc. President and CEO Go Miyasaka

Go Miyasaka, President and CEO

mocomoco provided us with high-quality deliverables at a speed that exceeded our expectations. Not only do they strictly adhere to deadlines, but projects often progress faster than anticipated, making them a company we can confidently entrust with our requests. They possess deep technical expertise, quickly understand instructions, and report, communicate, and consult at appropriate times, allowing for rapid responses to inquiries.

Furthermore, mocomoco has a high capacity for catching up with new technologies related to generative AI, fully covering our requirements from a technical standpoint. They always participate in meetings with a positive attitude, enabling lively exchanges of opinions and idea generation, making every meeting a highly productive and enjoyable time.

We are very much looking forward to continuing our product and research development together with mocomoco.

Summary

In the highly specialized legal field, our company successfully brought the research and development of a new AI feature from initiation to closed beta testing in a short timeframe. Regardless of the field, we will continue to rapidly acquire specialized knowledge to support the introduction of generative AI.

For consultations or inquiries regarding projects, please contact us via our inquiry page or at contact@mocomoco.ai.

Acknowledgments

Meetings with FRAIM Inc. were always enjoyable, and thanks to their precise and easy-to-understand instructions, our company was able to proceed with work smoothly. We would also like to express our deepest gratitude for their prompt and consistently attentive responses via Slack to our wide range of highly specialized questions and consultations covering legal practice, AI technology, cloud technology, and web technology. We look forward to your continued guidance and encouragement.