SOP-1040-01-AI_Requirements-Definition
Title: Requirements Definition
Version: 1.0
Effective Date: 2025-01-30
Approved By: Chief Technology Officer (CTO), AI IRB Chair
Scope: Enterprise-Wide AI/ML Projects
1. Purpose
This Standard Operating Procedure (SOP) defines the process for Requirements Definition within the AI Systems Development Life Cycle (AI-SDLC). It ensures that all business, regulatory, ethical, and technical requirements for AI-related projects are systematically identified, analyzed, documented, and approved, including integration with the AI Institutional Review Board (AI-IRB) as needed.
2. Scope
This SOP applies to all AI-focused software and system development projects at the organization that require formal requirements capture. It addresses:
- Elicitation, analysis, documentation, and validation of requirements
- Alignment with AI IRB guidelines for ethical, compliance, and risk considerations
- Traceability from concept to design, development, and validation
All departments—Product Development, Engineering, Quality Assurance, Operations, Technical Support, and the AI IRB—must follow this SOP for any AI-related project or feature release that goes through the AI-SDLC.
3. Roles and Responsibilities
| Role | Responsibilities |
|---|---|
| AI Project Sponsor | Champions the project, obtains funding, and ensures high-level alignment with AI/ethical guidelines. |
| AI IRB Liaison | Reviews potential AI ethics impacts, organizes IRB reviews, ensures compliance with AI IRB directives. |
| Product Manager | Leads the business vision, compiles user/business requirements, and coordinates with AI IRB Liaison. |
| AI Business Analyst | Elicits, documents, and validates all business and user requirements with an AI lens. |
| AI Data Scientist | Advises on feasibility, data requirements, and defines Model Validation Metrics (e.g. F1, AUC) for acceptance. |
| Quality Assurance (QA) | Reviews requirements for testability, ensures acceptance criteria are well-defined and trackable. |
| AI Ethics/Compliance | Monitors regulatory, ethical, privacy, and fairness aspects in the requirement definitions. |
| Operations | Contributes operational, infrastructure, and deployment constraints or requirements. |
| Technical Support | Communicates user needs, logs known constraints, and receives training for future user support. |
| Program Manager | Oversees schedule, budget, resources, ensures synergy among cross-functional teams. |
| Development | Provides input on technical constraints and ensures feasibility for system or platform requirements. |
4. Definitions
- AI-IRB: An internal board that reviews AI solutions for ethical compliance, data governance, and potential biases.
- SOW (Statement of Work): A high-level document that sets out the project scope, deliverables, acceptance criteria, and timeline.
- Business Requirements: High-level objectives describing what the organization or customers seek from the AI solution.
- System Requirements: Detailed functional and technical specifications for building AI/ML solutions, including data ingestion, model training/inference, performance metrics, and compliance.
- Ethical/Regulatory Requirements: Requirements derived from AI IRB guidelines, data privacy laws, fairness, interpretability, or other compliance standards.
- Authorized AI Agent: A validated AI system or subsystem identified within the Mind Matrix as having the authority to execute specific SDLC or operational tasks. |
5. Procedure Activities
Below is the end-to-end procedure for AI Requirements Definition:
5.1 Initiate Requirements Definition
- Project Sponsor: Communicates project concept to AI IRB Liaison and obtains initial approval to proceed with scoping.
- Program Manager: Sets up the initial project record in AI-SDLC tracking system.
- Product Manager, AI Business Analyst, and Lead Data Scientist: Begin high-level business requirements and Hypothesis Definition drafting.
5.2 AI IRB Preliminary Review
- AI IRB Liaison: Evaluates if the project needs immediate AI IRB attention (sensitive data, high-risk use case).
- If needed, the Liaison convenes an AI IRB quick check.
- If accepted: proceed.
- If clarifications: revise scope.
5.3 Requirements Elicitation
- AI Business Analyst: Conducts interviews and workshops with stakeholders (Product Dev, Data Scientist, Ethics, Ops).
- Technical Support: Provides user feedback, prior incidents or requests relevant to this new AI project.
- AI Data Scientist: Outlines technical feasibility, data needs, possible AI approaches, constraints.
- AI Ethics/Compliance: Identifies any data governance or interpretability mandates, including specific controls for recursive self-improvement subroutines.
- Collate results into a Draft Requirements Document (business & system requirements).
5.4 Analyze & Document Requirements
- AI Business Analyst: Structures user stories or use cases, includes acceptance criteria for AI functionality, risk analyses (bias, drift).
- Product Manager: Confirms alignment with business goals, updates potential release timeline in the SOW.
- Operations: Validates any infrastructure or deployment constraints.
- AI IRB Liaison: Ensures compliance concerns are integrated.
5.5 Requirement Validation
- Quality Assurance: Reviews requirements for testability, completeness, and clarity; ensures acceptance criteria are unambiguous.
- AI Data Scientist: Double-checks feasibility for ML tasks, data availability.
- AI Ethics/Compliance: Final check for fairness, privacy, or regulatory concerns.
- Operations: Confirms readiness or calls out missing operational details.
- Organize a formal Requirements Review Meeting with all roles. Document any changes or clarifications.
5.6 Approve System Requirements
- Product Manager: Presents final system requirements to Project Sponsor and AI IRB Liaison for sign-off.
- Project Sponsor: Provides final authorization, acknowledging resource and schedule implications.
- AI IRB Liaison: Grants final ethical/regulatory acceptance if needed, utilizing Exochain Peer Reviews to verify governance adherence.
- Program Manager: Marks requirements “Locked-Down.” Further changes proceed via formal change control (SOP-ChangeCtrl).
5.7 Requirements Baselining & Next Steps
- AI Business Analyst: Publishes the baselined Requirements Document in the repository.
- Program Manager: Informs relevant teams that development/design can proceed, referencing the approved requirements.
- Quality Assurance: Prepares test strategy (SOP-1210) referencing these requirements.
- Operations: Begins refining capacity or infrastructure planning (SOP-1002) if new demands are discovered.
6. References
- SOP-1000-AI: Program/Project Management for AI
- SOP-1002-AI: AI Capacity Management
- SOP-1041-AI: Detail Design Procedure
- AI IRB Guidelines: Organizational standard for ethically reviewing AI solutions
7. Document Control
| SOP Version | Date | Description | Approved By |
|---|---|---|---|
| 1.0 | 2025-01-30 | Initial release (AI-specific). | CTO, AI IRB Chair |
Supersedes: None
Retention: Maintain this SOP and all subsequent versions for minimum 5 years after decommission of the AI system.
Revision History
| Version | Revision Date | Summary of Changes | Author/Editor |
|---|---|---|---|
| 1.0 | 2025-01-30 | Initial AI-SDLC version | AI Governance Team |
End of SOP-1040-01-AI
@startuml
title SOP-1040-01-AI: Requirements Definition (Sequence Diagram)
participant “AI Project Sponsor” as SPONSOR
participant “AI IRB Liaison” as IRB
participant “Program Manager” as PRM
participant “Product Manager” as PDM
participant “AI Business Analyst” as BA
participant “AI Data Scientist” as DS
participant “AI Ethics/Compliance” as ETH
participant “Operations” as OPS
participant “Quality Assurance” as QA
participant “Technical Support” as TS
participant “Development” as DEV
’ Step 1: Sponsor Initiates Concept
SPONSOR -> IRB: Present new AI project concept
alt IRB requests clarifications
IRB -> SPONSOR: Clarification needed
SPONSOR -> IRB: Provide clarifications
IRB -> SPONSOR: Preliminary acceptance
else IRB approves
IRB -> SPONSOR: Preliminary acceptance
end
’ Step 2: Program Manager sets up project record
SPONSOR -> PRM: Notify of IRB acceptance
PRM -> PDM: Request business requirements drafting
PDM -> BA: Begin high-level requirements
BA -> TS: Gather known user constraints
TS -> BA: Provide user feedback
’ Step 3: AI IRB Preliminary Review
BA -> IRB: Submit early risk/ethical info
alt IRB concerns
IRB -> BA: Adjust scope or data approach
else IRB no concerns
IRB -> BA: Proceed
end
’ Step 4: Requirements Elicitation
BA -> DS: Confirm ML feasibility, data needs
DS -> ETH: Flag any data privacy or fairness issues
ETH -> BA: Provide compliance requirements
OPS -> BA: Provide operational constraints
BA -> BA: Draft Requirements Document
’ Step 5: Analyze & Document
BA -> QA: Request testability review
QA -> BA: Review acceptance criteria
BA -> PDM: Validate alignment with business goals
PDM -> BA: Confirm timeline in SOW
BA -> OPS: Confirm operational details
’ Step 6: Requirements Validation
QA -> DS: Check feasibility & data readiness
DS -> QA: Acknowledge feasibility
ETH -> BA: Check final ethics/regulatory compliance
BA -> IRB: If big changes, re-check compliance
OPS -> BA: Confirm final environment constraints
BA -> BA: Prepare final version for sign-off
’ Step 7: Approve System Requirements
BA -> SPONSOR: Present final requirements
SPONSOR -> IRB: Final ethical sign-off (if required)
alt IRB sign-off needed
IRB -> SPONSOR: IRB approval
end
SPONSOR -> PRM: Approve & lock down scope
PRM -> DEV: Communicate “Locked-Down” requirements
@enduml
Short textual explanation: This sequence diagram walks through the SOP-1040-01-AI requirements-definition process. It begins with the AI Project Sponsor presenting the new AI concept to the AI IRB Liaison, proceeds through the Program Manager setting up the project, and then details the roles (Business Analyst, Data Scientist, Ethics, Operations, Quality Assurance, etc.) collaborating on eliciting, analyzing, and validating requirements. Finally, the Sponsor and AI IRB give final sign-off on the “locked-down” requirements.