Version: 1.0.0
Purpose
AI Orchestrated Development (AOD) is designed to be platform independent.
Rather than depending on a specific AI provider or model, AOD defines a methodology that any sufficiently capable AI can follow.
This document describes the capabilities an AI should possess to effectively participate in the AOD methodology.
Core Principle
AOD is compatible with AI systems, not AI brands.
As artificial intelligence evolves, new models should be able to adopt the AOD methodology without modification to the methodology itself.
Minimum Capabilities
An AI participating in AOD should be capable of:
Maintaining context across extended conversations.
Reasoning through complex business and technical problems.
Following structured workflows.
Creating and maintaining documentation.
Recognizing relationships between documents.
Identifying contradictions and inconsistencies.
Explaining recommendations.
Asking clarifying questions.
Maintaining project state throughout multiple phases.
Supporting iterative refinement.
Recommended Capabilities
For the best experience, an AI should also support:
Large context windows.
Document generation.
Markdown formatting.
Code generation.
Architecture design.
Diagram generation.
File analysis.
Long-running conversations.
Reasoning across multiple governance artifacts simultaneously.
Expected Behavior
An AOD-compatible AI should:
Guide rather than dictate.
Ask questions before making assumptions.
Protect architectural integrity.
Recommend improvements.
Challenge unnecessary complexity.
Maintain consistency across governance artifacts.
Reference previous decisions.
Keep discussions focused on business outcomes.
Recognize when a project should return to an earlier phase.
Platform Independence
AOD intentionally avoids platform-specific implementation.
The methodology should remain compatible with future AI technologies without requiring changes to the standard.
Organizations may use different AI platforms for different stages of a project while continuing to follow the same AOD lifecycle.
Compatibility Evaluation
An AI should be evaluated on its ability to:
Complete Discovery.
Generate governance artifacts.
Maintain consistency between documents.
Recommend architecture.
Create implementation plans.
Validate completed work.
Guide product evolution.
Support long-term project continuity.
Known Limitations
Some AI platforms may have limitations including:
Limited context windows.
Restricted file handling.
Reduced reasoning capability.
Limited document management.
Conversation length restrictions.
Vendor-specific policies.
These limitations may affect the user experience but do not prevent adoption of the AOD methodology.
Best Practices
For the best results:
Use a new conversation for each project.
Maintain governance artifacts outside the AI platform.
Review each phase before approval.
Preserve project documentation in version control.
Treat the AI as an engineering partner rather than a code generator.
Future Compatibility
The AOD methodology is expected to evolve alongside advances in artificial intelligence.
As new AI capabilities emerge, the methodology may incorporate additional guidance while preserving the core principles and lifecycle established by the AOD Standard.
The goal of AOD is not to standardize artificial intelligence.
The goal of AOD is to standardize how humans and artificial intelligence collaborate to build exceptional software.