📋

Technical Details

Deep-dive into QUAD implementation: Agent patterns, flow docs, hierarchy rules

The 2 Dimensions: Business + Technical

Every circle operates along two axes: Business (understanding requirements, stakeholders, value) and Technical (code, systems, infrastructure). The B/T ratio defines each circle's focus.

B

Business Dimension

  • • Understanding stakeholder needs
  • • Translating requirements to specs
  • • Prioritization and value assessment
  • • User experience and workflows
  • • Communication with non-technical teams
T

Technical Dimension

  • • Writing and reviewing code
  • • System architecture decisions
  • • Database design and optimization
  • • Infrastructure and DevOps
  • • Performance and security

B/T Ratios by Circle

Management
B 80%
T 20%
Development
B 30%
T 70%
QA
B 30%
T 70%
Infrastructure
B 20%
T 80%

Shared vs Dedicated Resources

MORE DEDICATEDMORE SHARED
Management
Circle
Dedicated
per project
Development
Circle
Mostly Dedicated
can share
QA
Circle
Mostly Shared
across projects
Infrastructure
Circle
Always Shared
across directors

Organizational Structure

GROUP
(e.g., Wealth Management)
Director A
Project 1
Project 2
Director B
Project 1
Project 2
Director C
Project 1
Project 2
SHARED RESOURCES
Infrastructure Circle
ALWAYS SHARED
QA Circle
USUALLY SHARED
Enabling Teams
OPTIONAL

Enabling Teams (Optional)

Enabling Teams are NOT counted as QUAD Circles. They are optional support groups that provide specialized expertise across ALL 4 circles.

3 Core Enabling Teams (Almost Always Needed)

1

Architecture Group

ROLES
Solution ArchitectDomain ArchitectDatabase ArchitectCloud Architect
AI AGENT
Design Agent
Reviews PRs for architecture violations, suggests patterns
When needed: Large systems, microservices, complex integrations
2

Security Team

ROLES
Security EngineerPen Tester
AI AGENT
Security Scanner Agent
SAST/DAST scans, vulnerability detection, dependency audit
When needed: Any production app, especially with user data
3

Compliance Team

ROLES
Compliance OfficerAuditor
AI AGENT
Compliance Checker Agent
Policy validation, audit trail generation, data handling checks
When needed: HIPAA, SOC2, GDPR, PCI-DSS regulated industries

4 Extended Enabling Teams (Add Based on Project Type)

4

Data/Analytics Team

ROLES
Data EngineerData ScientistBI Analyst
AI AGENT
Analytics Agent
ETL monitoring, query optimization, anomaly detection
When needed: ML features, dashboards, data pipelines, reporting
5

UX/Design Team

ROLES
UX DesignerUI DesignerUX Researcher
AI AGENT
Design System Agent
Component consistency, accessibility validation, design tokens
When needed: Consumer apps, complex workflows, brand consistency
6

Platform/Tools Team

ROLES
Platform EngineerDX Engineer
AI AGENT
Tooling Agent
CI/CD optimization, developer experience metrics, build performance
When needed: Large orgs, internal developer tools, CI/CD standards
7

Accessibility Team

ROLES
A11y SpecialistA11y Tester
AI AGENT
A11y Scanner Agent
WCAG compliance scanning, screen reader testing, color contrast
When needed: Government contracts, public apps (WCAG/ADA/Section 508)

Key Points

⚙️Enabling Teams are OPTIONAL - Not every project needs them
🔄Support ALL 4 Circles - They don't belong to one circle
🤝Always SHARED - Across all directors/projects (like Infra Circle)
🤖Each has AI Agent(s) - Automation for repetitive checks
📊Core vs Extended - Start with Core 3, add Extended as needed

Hierarchy Rules

Level 1
BASE RULES
QUAD Platform - Universal do's and don'ts
Level 2
COMPANY-WIDE
Entity/Org level policies
Level 3
DIRECTOR
Program/Portfolio level
Level 4
TECH LEAD
Team level
Level 5
TEAM MEMBER
Personal preferences

Each level can ADD restrictions, never CONTRADICT upper levels

Example Rules

LevelExample Rule
Base Rules"Never commit secrets to Git"
Company-wide"All PRs require 2 reviewers"
Director"Use Java 21 only for this program"
Tech Lead"Follow our naming conventions"
Team Member"I prefer dark mode in IDE"

Agent Class-Object Pattern

CLASS (Template)

Story Agent (Template)
  • - Capabilities defined
  • - Base rules configured
  • - No connection yet
  • - "Ready" state
Cannot execute tasks

OBJECT (Instance)

Your Story Agent
  • - Connected to platform
  • - App password set
  • - Local machine context
  • - User-specific rules
Can execute tasks
→
assign + setup = object

Estimation Agent Pipeline

The Estimation Agent is unique - it orchestrates multiple agents in a sequential pipeline to produce accurate, confidence-based estimates.

💻
Code Agent
Files, components, complexity
→
🗄️
DB Agent
Tables, migrations, queries
→
📊
Flow Agent
Screens, APIs, integrations
→
🎯
Estimation Agent
Aggregates all, final estimate

Pluggable Design: Agents can be added/removed (e.g., add Security Agent, remove DB Agent if no DB changes)

Confidence Deductions

High code complexity (>50 files)-15%
Database migration required-10%
External API integration-10%
New technology/framework-15%
Cross-team dependency-10%
Unclear requirements-20%
No similar past stories-10%

Confidence Additions

Similar story completed before+10%
High test coverage in area+5%
Clear requirements+5%
Same developer did similar+5%
Formula: 90% base + additions - deductions
Range: 30% (min) to 95% (max)
ESTIMATION RESULT
COMPLEXITY
Octahedron (8)
CONFIDENCE
78%
EFFORT
5-6 days
BREAKDOWN BY AGENT
• Code Agent: 3 files, 2 components, medium complexity
• DB Agent: 1 new column (user_preferences.theme)
• Flow Agent: 2 screens affected, 1 API endpoint
⚠️ RISKS
• CSS variable system not standardized
• No existing theme context in app

Human Override with AI Learning

1
AI Presents
Complexity: 8, Confidence: 78%
2
Human Reviews
[Accept] or [Override]
3
Override
Human: 6, Reason: 'Done before'
4
Story Done
Actual: 5 days
5
AI Learns
Human was closer! Adjusting...

Override requires a reason (tracked for retrospectives). AI adjusts models based on who was more accurate.

Flow Document Template

# FLOW: [Feature Name]
## Overview
| Field | Value |
| Flow ID | FLOW_XXX_001 |
| Owner | Management Circle (BA/PM/TL) |
| Status | Draft / Active / Retired |
## Step 1: [Action Name]
### UI
| Element | Type | Validation |
### API
POST /api/endpoint
### Database
SELECT * FROM table
### Test Cases
| TC001 | Happy path | 200 |

One Source of Truth Architecture

GIT REPOSITORY
(Version Controlled)
/docs/flows/
/src/code
/tests/specs
↓
QUAD WEB APP
Flow viewer, API browser, DB schema, AI chat (RAG), Test dashboard
↓
Confluence (sync)
Jira (sync)
Swagger (auto)

All tools sync FROM Git (single source). Never edit in Confluence directly.