Your business already has an interface for AI.
Files and folders. ICM turns the structure people have used for decades into a shared operating system for human and artificial intelligence.
H
Shared work
Files + folders
AI
AI is being hired one chat at a time.
Teams get impressive answers, but the work remains trapped in isolated conversations with no durable operating context.
Context reset
Every interaction starts from zero. The business pays the reset cost.
01
Context is repeatedly explained
People spend time reconstructing the company, the problem, the rules and the desired output.
02
Quality depends on the operator
The best result stays with the person who knows the right prompt, not with the organisation.
03
Learning disappears into chat history
Better instructions, corrections and decisions are not converted into reusable company knowledge.
The model is powerful. But the model is not your operating system.
ModelCapability
×
ContextUnderstanding
×
WorkflowRepeatability
The world's best model still needs to know where it is, what matters here, what it may use, and what a good output looks like.
Put humans and AI in the same workspace.
Folders define where the work happens. Markdown files explain the business, the stage, the rules and the expected result. Both sides read and improve the same material.
H
root/
├── AGENTS.md
├── CONTEXT.md
├── 01_VAULT/
├── 02_OPERATIONS/
└── 99_ARCHIVE/
├── AGENTS.md
├── CONTEXT.md
├── 01_VAULT/
├── 02_OPERATIONS/
└── 99_ARCHIVE/
AI
Five simple layers turn a folder tree into an operating system.
The first three layers route the work. The final two supply the knowledge and hold the product.
Layer 0AGENTS.md routes the workspace.
Layer 1CONTEXT.md gives the helicopter view.
Layer 2Folder context defines each stage contract.
Layer 3Reference material + skills form the factory.
Layer 4Working artifacts are the product.
Layer 0AGENTS.md
Layer 1Root CONTEXT.md
Layer 2Folder CONTEXT.md
Layer 3Reference material + skills
Layer 4Working artifacts
Every interaction can improve the product. And the factory that makes it.
Stable context that makes future work faster, safer and more consistent.
Each reviewed artifact reveals what the shared system should learn next.
The method is simple. Adoption is an operational transformation.
Work lives in people, inboxes and disconnected tools.
- Knowledge is implicit
- Ownership is unclear
- Review happens inconsistently
- AI is added as another isolated tool
Transform the work
Not just the tool
Work has a place, a contract and a review gate.
- Knowledge becomes explicit
- Folders clarify ownership and stage
- Humans approve visible artifacts
- AI works inside the operation
Start with one valuable workflow. Prove the model, then expand.
Map the work
Choose a recurring workflow with clear inputs, decisions, outputs and owners.
Structure the context
Separate trusted sources, reusable knowledge, instructions and active artifacts.
Run with review gates
Let people and AI share the work while humans inspect and approve each stage.
Improve the factory
Turn corrections and good examples into stronger context for the next run.
The first goal is not company-wide automation. It is one repeatable example of a better hybrid way of working.
The method is simple. Transformation is the work.
You can explore ICM directly through Jake Van Clief's community. When the question becomes how to redesign a real business operation around it, that is the workshop.
?