While FLOW emphasizes fluidity and adaptation, it still provides a concrete framework for organizing work. This framework is designed to be lightweight, flexible, and optimized for hybrid human-AI teams.
Overview
The FLOW framework consists of four key components that work together to enable continuous delivery of value:
Roles
Fluid, capability-based responsibilities
Artifacts
Living documents that guide work
Ceremonies
Lightweight synchronization points
Metrics
Real-time flow indicators
Roles
In FLOW, roles are not job titles but fluid responsibilities that team members (human or AI) take on based on the needs of the moment and their capabilities.
Stream Guide
Purpose
Maintains the vision and coherence of an objective stream, ensuring that all work contributes to meaningful user outcomes.
Key Responsibilities
- Define and refine objectives within the stream
- Prioritize objectives based on user value and dependencies
- Communicate stream vision to implementation agents
- Monitor stream health and adjust course as needed
Who Can Be a Stream Guide?
Typically humans with domain expertise, though AI agents can assist with data analysis, priority recommendations, and maintaining stream documentation.
Implementation Agent
Purpose
Transforms objectives into working software through code, tests, and documentation.
Key Responsibilities
- Select objectives from streams based on capability match
- Implement solutions that meet objective criteria
- Ensure code quality through tests and documentation
- Collaborate asynchronously with other agents
Human vs AI Agents
Human Agents Excel At:
- Creative problem solving
- User experience design
- Complex architectural decisions
- Stakeholder communication
AI Agents Excel At:
- Boilerplate and repetitive code
- Test generation and coverage
- Documentation maintenance
- Code refactoring and optimization
Quality Guardian
Purpose
Ensures that all work flowing through the system meets quality standards without becoming a bottleneck.
Key Responsibilities
- Define and maintain quality gates
- Review critical path implementations
- Automate quality checks where possible
- Mentor implementation agents on best practices
Automation First
Quality Guardians focus on building automated quality gates (tests, linters, security scans) rather than manual reviews. Human review is reserved for architectural decisions and critical user-facing changes.
Artifacts
FLOW uses minimal, living artifacts that are continuously updated rather than comprehensive documents that quickly become outdated.
Objective Streams
Example: E-commerce Platform Streams
{
"streams": [
{
"id": "user-experience",
"guide": "@sarah",
"objectives": [
{
"id": "responsive-checkout",
"status": "in-progress",
"agent": "claude-1",
"started": "2024-01-15T10:00:00Z",
"value": "Reduce cart abandonment by 15%"
},
{
"id": "personalized-recommendations",
"status": "queued",
"dependencies": ["user-analytics"],
"value": "Increase average order value by 20%"
}
]
},
{
"id": "platform-stability",
"guide": "@michael",
"objectives": [
{
"id": "database-optimization",
"status": "in-progress",
"agent": "@john",
"started": "2024-01-14T08:00:00Z",
"value": "Reduce query times by 50%"
}
]
}
]
}Flow Metrics Dashboard
A real-time dashboard showing the health and velocity of all streams:
Flow Velocity
Active Agents
Stream Health
Quality Score
Ceremonies
FLOW minimizes ceremonies, focusing only on essential synchronization points. All ceremonies are optional and should only occur when they add value.
Stream Synchronization
Purpose
Align implementation agents with stream objectives and resolve dependencies
Format
- Frequency: As needed (typically 1-2 times per week)
- Duration: 15-30 minutes maximum
- Participants: Stream Guide + active Implementation Agents
- Async Option: Can be done via collaborative documents
Agenda
- Review completed objectives (2 min)
- Identify blockers or dependencies (5 min)
- Clarify upcoming objectives (5 min)
- Adjust priorities if needed (3 min)
Quality Gates
Purpose
Automated checkpoints that ensure work meets quality standards before flowing downstream
Types of Gates
🤖 Automated Gates
- Unit test coverage ≥ 80%
- Zero critical security issues
- Performance benchmarks met
- Documentation complete
👤 Human Gates
- UX review for user-facing changes
- Architecture review for new patterns
- Security review for sensitive data
- Accessibility compliance check
Metrics
FLOW emphasizes actionable metrics that help teams understand and improve their flow, rather than vanity metrics that just look good on dashboards.
Flow Velocity
Definition
The rate at which objectives flow from conception to production
How to Measure
Flow Velocity = Completed Objectives / Time Period
Example:
- Week 1: 5 objectives completed
- Week 2: 7 objectives completed
- Week 3: 6 objectives completed
- 4-week average: 6 objectives/weekWhat Good Looks Like
- Steady or increasing velocity over time
- Predictable completion rates
- No artificial spikes at "sprint ends"
Stream Health
Definition
A composite metric indicating how well a stream is flowing
Components
Objective Age
How long objectives stay in progress
Blocker Frequency
How often work gets blocked
Quality Gate Pass Rate
First-time pass rate through gates
Agent Utilization
Balance of work across agents
Summary
The FLOW framework provides just enough structure to enable effective collaboration between humans and AI agents while maintaining the flexibility to adapt to changing needs. By focusing on continuous flow rather than fixed iterations, teams can deliver value more naturally and efficiently.
Next, we'll explore how to implement FLOW in your organization, including migration strategies from existing methodologies.