I wrote this practical guide to make my academic research on TWAF accessible to practitioners. It explains how to assess value and trust together and use the AI Decision Gate to make deployment decisions.
Enterprise AI requires a decision framework that connects the business case with the conditions for responsible operation. A promising pilot needs measurable value, suitable data, accountable owners, and controls proportionate to its intended use.
The Two Wings AI Framework brings these requirements into one assessment. Value and trust develop in parallel, with governance spanning both wings and the AI Decision Gate.
The business case and the operating conditions
The value wing examines whether the initiative addresses a meaningful business need, whether people and processes are ready to adopt it, and whether its economics support investment and scale. The trust wing examines the data foundation, model and agent ownership, assurance, runtime controls, human oversight, and compliance evidence.
Assessing the two together gives decision-makers a fuller view of readiness. An initiative with demonstrated value still requires appropriate operational assurance. An initiative with strong controls still requires a business case that justifies investment.
Two foundations, twelve elevations
AI-Worthy Processes form the foundation of the value wing. AI-Ready Data forms the foundation of the trust wing. Each wing comprises six elevations that organize the assessment from its foundation through deployment and scale.
The value elevations cover process suitability, people and change readiness, value modeling, pilot evaluation, benefit ownership, and portfolio economics. The trust elevations cover data readiness, AI visibility and ownership, model and output assurance, agentic and runtime control, human oversight, and compliance evidence.
Data and AI work progress in parallel. AI use cases inform data priorities, while improvements to data quality, meaning, and traceability support the development and assurance of those use cases. My separate published work, the AI-Ready Data Diagnostic (ARDD), assesses this data foundation. Its dedicated article introduces the model and links to its overview and research paper.
Governance and the AI Decision Gate
Governance establishes the strategic bounds for AI: value appetite, trust thresholds, and forbidden zones. It assigns decision rights and accountability across both wings, and defines the evidence required at the AI Decision Gate.
For each use case, the process is to register the initiative, capture evidence, assess the twelve elevations, reach a verdict on value and trust, and issue an auditable decision record. Deployment and scale require sufficient evidence on both wings. Other outcomes call for strengthening assurance, revisiting the business case, or discontinuing the initiative.
The assessment supports decisions throughout the lifecycle. Changes to data, models, agent permissions, operating conditions, or obligations can require a renewed review.
Methodological foundations
The framework synthesizes data management, enterprise architecture, and AI governance disciplines. Its foundations draw on DAMA-DMBOK, TOGAF, the NIST AI Risk Management Framework, ISO/IEC 42001, and the EU AI Act.
It provides a structure for assessment and decision-making. Outcomes depend on the quality of the evidence, the assessment criteria, and the controls implemented for each use case. My full research paper, available on SSRN, sets out the framework and its provenance.
Applying the framework
Begin with a defined use case and its strategic context. Assign accountable owners, establish the value and trust criteria, and collect evidence for the elevations relevant to the decision. Review the findings at the AI Decision Gate and document the outcome, conditions, and required actions.
Explore the full framework, or discuss an AI governance or readiness engagement.