Inspiration
GENBA Voice began with a small question from everyday work: what happens to concerns that feel too minor to formally report, but continue to create confusion, anxiety, or avoidable mistakes?
After many years working close to frontline operations, I have often seen people notice a problem but hesitate to speak because they do not want to blame someone, cause conflict, or make the issue sound larger than it is. I wanted to create a simple place where those small signals could be organized before they disappeared.
What it does
GENBA Voice helps workers turn a small workplace concern into a clearer starting point for conversation.
The user selects the situation and urgency, describes the concern without entering names or confidential information, and receives five structured outputs:
- what may be happening
- a possible system pattern
- a small step to try
- a calm message that can be shared
- points that still need confirmation
The purpose is not to make decisions for the workplace. It is to help people begin a constructive conversation without immediately blaming an individual. The interface supports both Japanese and English, including the form, dropdown options, guidance, and AI responses.
How we built it
I built GENBA Voice as a lightweight Google Apps Script web application using HTML, CSS, JavaScript, and the OpenAI API.
The browser sends the selected context, urgency, language, and concern to a server-side Apps Script function. The OpenAI API returns a structured analysis, which is then displayed as five separate sections in the interface.
The application does not intentionally store the user’s input. The API key is kept in Google Apps Script Properties rather than exposed in the browser. The interface also asks users not to enter company names, personal names, client information, or confidential information.
During Build Week, I connected the frontend to the API, added bilingual output, improved the mobile-friendly interface, and tested the complete analysis flow in both Japanese and English.
Challenges we ran into
The main challenge was connecting a simple frontline-facing interface to the OpenAI API while keeping the experience understandable and safe for people who may not be familiar with AI.
I also encountered a small but important bilingual interface bug. The main text switched to English, but the two dropdown menus remained in Japanese because their update function was not being called correctly. I traced the problem, simplified the language-switching flow, redeployed the existing web app as a new version, and confirmed that both dropdowns now switch correctly.
Another challenge was shaping the AI response so it would not blame or judge an individual. I structured the output around the situation, the possible system pattern, a small testable step, and a calm message for starting a conversation.
Accomplishments that we're proud of
I am proud that GENBA Voice became a working public prototype rather than remaining only an idea.
It now supports the full flow in both Japanese and English: selecting a situation, describing a concern, receiving structured AI guidance, and copying a calm message that can be used to begin a workplace conversation.
I am also proud that the application stays focused on a small and practical purpose. It does not try to replace managers, specialists, or human judgment. It helps someone take one careful first step with a concern that might otherwise remain unspoken.
What we learned
I learned that the quality of an AI application depends not only on the model, but also on how carefully the purpose, questions, boundaries, and output structure are designed.
I also learned that a small interface detail can affect whether people trust and understand the whole application. Testing the Japanese and English flows separately helped me find problems that were not visible from the code alone.
Most importantly, I learned that frontline experience can become a useful starting point for building with AI. I did not begin with a large technical plan. I began with a small discomfort I had seen repeatedly at work, then improved the prototype one step at a time.
What's next for GENBA Voice
The next step is to test GENBA Voice with a small number of real users and listen carefully to where the wording, choices, or results still feel unclear.
I would like to improve accessibility, refine the safety guidance, and make the generated suggestions easier to adapt to different kinds of workplaces. I also want to explore additional languages while keeping the interface simple.
There is still much to learn, but I hope GENBA Voice can grow as a small bridge between an unspoken frontline concern and the first constructive conversation about improvement.
Log in or sign up for Devpost to join the conversation.