Issue
・The quality of internal manuals is poor, but due to the large volume, they have not been addressed.
・There are plans to translate the manuals in the future, so improving the quality of the original text is necessary.
・Currently, there are no rules, and there are differences in expression and quality depending on the author (development, design, maintenance).
Proposal Details
We proposed a manual quality improvement process utilizing generative AI.
1. Proposal of a hybrid model
・Use a text generation AI tool to perform a first rewrite of existing manuals
・Humans (writers) review and adjust the content to ensure final quality
2. Support for Pre-Implementation Preparation
・Analysis of existing manuals: writing style, structure, etc.
・Design of quality standards and writing rules
・Creation of sample manuscripts for AI training
Even when writing manuals using generative AI, it is necessary to design the quality standards required for the manuals.
Without designing quality standards, it is impossible to evaluate the quality of the generated manuals or measure the effectiveness of the initiative. Additionally, there may be inconsistencies among writers regarding how much to adjust the generated drafts, which could prevent standardization.
We explained to Company C the importance of "designing quality standards" and proposed a system that includes this aspect.
3. Phased Implementation Plan
・Verify effectiveness with a small start
・Decide whether to proceed with full-scale implementation based on the results
Expected Effects
If this proposal is implemented, the following effects can be expected.
・Collaboration between AI and humans enables raising the baseline quality of manuals and unifying notation rules
・Reduces variation in writing styles among authors, maintaining consistent quality regardless of who writes
・Enables efficient handling even during large-scale manual revisions
・Improves source text quality, stabilizing translation quality in multilingual deployments
While aiming to both reduce the labor required for manual creation and updates and improve quality,
it is expected to lead to the establishment of a sustainable document operation system in the medium to long term.
