AI Operating Layer
AI helps you reach understanding faster.
AI does not solve the problem by itself. It amplifies what already exists.
Knowledge inside NT is fragmented. It lives in people, chats, old presentations, spreadsheets, and personal folders.
"The company often starts from scratch, even when the right materials already exist somewhere."
We are not implementing "one AI tool." We are creating a corporate knowledge layer — a structured layer of internal knowledge that different AI tools can work on top of.
Files + Prompts + Ownership + Instructions = AI Operating Layer
Google Drive is not just storage. At the first stage, it is the simplest way to organize corporate memory into a manageable structure: by departments, owners, materials, and use cases.
Click folders to expand · Each level inherits and restricts access
Each key function gets its own directory. Each directory should be not an archive, but a working environment from which both people and AI can quickly produce useful output.
Each directory has an owner. This is not a "folder administrator." This is the person responsible for making the department's knowledge usable: relevant, clear, structured, and practical for AI-supported work.
agents.md isEach directory contains an agents.md file — the main instruction file for AI inside a specific department. It explains what the department does, which materials are primary, which style to use, and when human review is required.
# Department: Marketing## RoleYou are an AI assistant for NT Marketing.## Primary materials- brand_guidelines.pdf- tone_of_voice.md- campaign_templates/## Rules- Never invent statistics- Always reference brand positioning- Flag when human review is required
agents.md mattersThe same AI tool can produce very different results depending on context.
agents.md turns a folder from file storage into a working environment.
Employees no longer start from a blank page. They start from existing materials, examples, prompts, and department rules. AI helps create the first version faster, but it does not replace professional judgment.
The company gets more stable execution quality.
We create a shared AI Casebook — a collection of real examples where AI helped a department: the task, input materials, prompt, result, what worked, and what still required human review.
Not "ideas about AI," but actual working patterns.
We are not building deep automation yet. First, we need to understand which processes are clear, repeatable, and structured enough.
"Automating chaos is like putting autopilot on Helios' chariot: impressive, but someone will have to extinguish the city afterwards."
Main idea