Tribal Knowledge System

The Hidden Knowledge Problem

Every organization has knowledge that exists only in people’s heads:

Institutional Knowledge

“We never count demo units as inventory”Unwritten rules specific to this organization

Experiential Knowledge

“We tried that approach in 2019, and it doesn’t work with our suppliers”Learned through doing, never documented

Operational Knowledge

“Sales from the Miami branch always come in a day late”How things actually work day-to-day

This knowledge is critical for correct analysis. Without it, AI will give technically correct but practically wrong answers.

The Three-Level System

Superatom captures tribal knowledge at three levels, each with different scope and application:

Level 1: Global Knowledge

Scope: Applies to every query for every user Purpose: Organization-wide definitions, rules, and conventions

Examples

Knowledge Node	Effect
”Sales always exclude returns and refunds”	All sales queries automatically filter out returns
”Fiscal year starts April 1”	Date calculations use Apr-Mar fiscal year
”EMEA region includes UK despite Brexit”	Geographic groupings remain consistent
”Inactive products have status code ‘D‘“	Product queries can correctly filter active items

How It Works

Without global knowledge, “Q3” might be interpreted as calendar Q3 (Jul-Sep) instead of fiscal Q3.

Level 2: User Knowledge

Scope: Applies to queries from specific users Purpose: Role-based views, personal preferences, access restrictions

Examples

User/Role	Knowledge Node	Effect
West Coast Manager	”Only show data for regions CA, OR, WA”	All queries automatically filtered to West Coast
Finance Team	”Include pending transactions”	Queries show uncommitted data
Executive	”Always compare to same period last year”	Automatic YoY comparison
Warehouse Staff	”Show quantities in cases, not units”	Unit conversion applied

How It Works

The same question from a different user would show different data based on their knowledge nodes.

Level 3: Query Knowledge

Scope: Applies to specific types of questions Purpose: Define business terms, calculation methods, domain-specific logic

Examples

Term/Query	Knowledge Node	Effect
”deadstock"	"Items with no sales in 12 months where On-Hand Cash Value exceeds COGS”	Specific calculation applied
”overstock"	"Inventory exceeding 90 days of average sales”	Days-of-supply calculation
”margin"	"(Revenue - COGS - Shipping) / Revenue”	Includes shipping in calculation
”active customer"	"Placed order in last 180 days”	Customer status definition

How It Works

Knowledge Node Structure

Each knowledge node contains:

knowledge_node:
  id: "kn_overstock_definition"
  type: "query"  # global | user | query
  trigger: "overstock"  # when to apply

  content:
    description: "Items with inventory exceeding 90 days of sales"
    calculation: |
      - Structural products: Rolling 90 days COGS > 0
      - Decking products (MajorMinorCode = '0103DEC'): Rolling 180 days COGS > 0
      - Specialty products: Rolling 120 days COGS > 0

    key_columns:
      - "FactInventoryDaily.OnHandInventory"
      - "FactSalesInvoice.AverageCostExtended"
      - "DimItem.ProductType"

    formula: "OnHandInventory - (RollingPeriod * COGS)"

  metadata:
    created_by: "[email protected]"
    created_at: "2025-11-15"
    updated_at: "2026-01-21"
    applies_to: ["inventory queries", "deadstock analysis"]

The Knowledge Graph

Knowledge nodes connect to form a graph that captures organizational understanding:

This graph enables:

Relationship traversal - Understanding how concepts connect
Context inheritance - Child nodes inherit parent knowledge
Conflict resolution - More specific nodes override general ones

Creating Knowledge Nodes

From the UI

Navigate to Knowledge Base → Knowledge Graphs
Click + Create Node
Select level (Global, User, Query)
Define trigger conditions
Add description and rules
Save and test

From Conversations

When the AI gives an answer that needs organizational context:

User corrects: “Actually, we don’t count returns in sales”
System prompts: “Would you like to save this as a rule?”
User confirms scope (global, user, query)
Knowledge node created automatically

How Knowledge Flows

Priority Order:

Query-level knowledge (most specific) wins
User-level adds personal/role context
Global-level provides baseline rules

Why This Matters

The Problem with Generic AI

Scenario	Generic AI Answer	With Tribal Knowledge
”Show sales”	All transactions	Excludes returns, demos, internal transfers
”Q4 revenue”	Oct-Dec	Jan-Mar (fiscal year)
“Top customers”	By raw revenue	By margin-adjusted lifetime value
”Overstock items”	Undefined error	Calculated per product category rules

Business Impact

Accuracy

Answers reflect how YOUR organization actually works

Consistency

Same terminology means the same thing for everyone

Onboarding

New employees get organization-aware AI from day one

Knowledge Preservation

Institutional knowledge survives employee turnover

Curating for Any Domain

The Tribal Knowledge system allows Superatom to be curated for any industry and any organization:

Supply Chain
Retail
Finance
Healthcare

Lead time calculations
Safety stock formulas
Supplier rating methods
Logistics cost allocation

Next Steps

Knowledge Base Platform

Managing knowledge in the platform

Coding Agents

How agents use tribal knowledge

​The Hidden Knowledge Problem

Institutional Knowledge

Experiential Knowledge

Operational Knowledge

​The Three-Level System

​Level 1: Global Knowledge

​Examples

​How It Works

​Level 2: User Knowledge

​Examples

​How It Works

​Level 3: Query Knowledge

​Examples

​How It Works

​Knowledge Node Structure

​The Knowledge Graph

​Creating Knowledge Nodes

​From the UI

​From Conversations

​How Knowledge Flows

​Why This Matters

​The Problem with Generic AI

​Business Impact

Accuracy

Consistency

Onboarding

Knowledge Preservation

​Curating for Any Domain

​Next Steps

Knowledge Base Platform

Coding Agents

The Hidden Knowledge Problem

The Three-Level System

Level 1: Global Knowledge

Examples

How It Works

Level 2: User Knowledge

Examples

How It Works

Level 3: Query Knowledge

Examples

How It Works

Knowledge Node Structure

The Knowledge Graph

Creating Knowledge Nodes

From the UI

From Conversations

How Knowledge Flows

Why This Matters

The Problem with Generic AI

Business Impact

Curating for Any Domain

Next Steps