Business as Code - why Git instead of Notion and how to bring in AI

We don't use a wiki. We don't use Notion. We don't even have a shared Google Docs graveyard full of outdated guides.

We use a Git repository.

Sounds strange? Maybe. But ask yourself one simple question: Where does your company's know-how actually live today? In the heads of 2-3 key people? In Slack? In a document "Peter has somewhere on Drive"?

We took a different path.

Processes, knowledge, strategy, templates and personas live in a Git repository as versioned Markdown documentation. Structured so that both humans and AI can navigate it.

Because the difference between companies that merely experiment with AI and those that truly put it to work will not come down to which model they use. It will come down to whether their know-how is stored in a way that supports systematic work.

Two layers

In our case this produces two layers:

Brain: a Git repository as the company's internal operating system.

Smart Colleague and hands: AI agents connected to Gmail, Google Drive, Sheets and CRM.

What does this look like in practice?

When an agent receives the task "process payroll," it does not start from scratch. Using YAML metadata it finds the right process in the repository: who owns it, what state it is in, which procedure to follow. It loads only the relevant workflow, checkpoints and expected outputs. Then it reaches into Gmail for source documents, writes data to Sheets and saves output to Drive.

Other scenarios work the same way.

Connect AI to the CRM and just type: "Prepare a training proposal." The output is a finished DOCX following your company standards.

And when you need a new process, describe it briefly. Thanks to a clear structure, a consistently formatted process document with all required elements appears.

The fundamental difference

AI does not work on top of chaos. It works on top of a company system.

Git plays a key role here. Process changes go through a pull request, review and a history of edits. You see who changed what, when and why. Company know-how stops behaving like a pile of documents and starts behaving like a system you can govern, version and improve.

I believe this is where the AI-first company begins. Not the one where people occasionally generate something in a chat. But the one where AI truly knows how the company operates.