Building a Custom AI Brain That Actually Maps Your Entire Organization

A lot of people use Obsidian as their AI knowledge base, but one builder just showed off a fully custom alternative built in Claude Code in about a week. The tool maps an entire organization into an interactive graph of employees, agents, SOPs, tools, and departments, all linked together with clickable nodes. It also plugs into a custom chat interface where those connections surface automatically during conversations.

What the Tool Actually Does

The demo walks through a web application called "Conducting AI" that serves as an organizational brain. The main view is a large interactive node graph where each node represents a specific entity: a person, an AI agent, an SOP, a software tool, a department, or a project. Every node is color-coded by type and connected to related nodes through visible relationship lines.

The left sidebar shows a filterable list organized by category. In the demo, the creator clicks into a "Sales" department and the interface displays the full breakdown: the demo company, its Sales AI Copilot, Sales AI Brain, Sales Autonomous Agent, individual team members (Michael Torres, Omar Hassan, Rachel Green, Samantha Reese, Lisa Chang, David Park), and a long list of sub-agents like "Account Risk and Expansion," "Automated Follow-Up Sequences," and "Meeting Preparation & Briefing."

The graph view supports multiple layout modes. One view arranges everything in concentric circles, grouping nodes by department and function. Categories visible in the sidebar include Management, Growth, Admin, Supply Chain, Automation, Leadership, Analytics, Human Resources, and Compliance. The system tracks 757 entities and 1,047 connections in the demo dataset.

Why This Beats Obsidian for Enterprise AI

The creator's argument is straightforward. Obsidian works well as a personal knowledge graph, but it was not designed for multi-user organizations with AI agents. This custom tool solves three specific problems:

Permission structures become visual. When an employee uses the AI brain, you can see exactly what they have access to based on their node connections. The graph makes access boundaries explicit rather than hidden in config files.
Human-agent teams are first-class objects. Departments contain both human employees and AI agents on the same level. SOPs connect to both. This means the system models how work actually gets done in an AI-augmented org, not just how humans are organized.
The chat interface inherits the graph. The demo shows a custom chat UI where agents, employees, and data sources are all referenceable. When you mention an SOP or a tool in conversation, the brain already knows the full context and connections.

The Build Details

The entire application was built using Claude Code. The creator frames it as a "custom beta for conducting AI software" and mentions a roughly one-week build timeline. The interface runs as a web app (visible at localhost:3000 in the demo) and includes a structured dashboard view with high-level frameworks, business reference models, and a color-coded matrix that appears to map capabilities across departments.

The dashboard view shows a grid-based layout with multiple card sections covering operational domains. Each card contains structured data fields, and the color-coded rows in the matrix suggest a maturity or coverage assessment across business functions.

Key Takeaways

A custom organizational AI brain can be built in about a week with Claude Code, replacing what many people try to hack together in Obsidian.
The graph-based approach maps real organizational complexity: departments, humans, AI agents, SOPs, tools, and the connections between all of them.
Permission structures and access boundaries become visible when the knowledge graph is purpose-built for multi-user AI environments.
Linking the knowledge graph to a chat interface means every conversation has full organizational context without manual referencing.
The creator is considering releasing the tool for free, which could make this accessible to teams that lack the resources to build their own.