HR7 was developed to identify those structural conditions before major AI investments are made by analysing a structured historical reference population of 121 independently documented historical AI systems spanning healthcare, autonomous systems, financial forecasting, enterprise AI, natural language processing, computer vision, and industrial AI applications. Rather than evaluating model accuracy alone, HR7 identifies recurring structural deployment constraints that influence whether AI systems remain reliable under real-world operational conditions. Together with Predictive Feasibility Assessment (PFA), HR7 evaluates whether the currently available operational evidence supports reliable AI development for a defined objective.
Many AI systems demonstrate strong performance during development and controlled evaluation. However, real-world deployment introduces additional challenges that are not always captured by benchmark performance alone. HR7 was developed to identify recurring structural factors that influence whether AI systems remain reliable, robust, and operationally viable under real-world conditions.
HR7 uses documented AI systems as a structured historical reference population to identify recurring operational patterns.
The framework identifies recurring limitations related to validation, transferability, environmental change, governance, and operational integration.
HR7 provides the evidence foundation used by PFA to assess whether current operational evidence supports reliable AI development.
The HR7 Historical Reference Framework is grounded in a structured population of independently documented AI systems across multiple application areas. The purpose is not to rank systems, but to identify recurring structural factors that influence deployment outcomes.
Across historical AI systems, recurring deployment limitations appear in several structural categories.
Performance observed in controlled conditions may not automatically transfer to independent operational environments.
Changing operational conditions can affect reliability through concept drift and non-stationary behaviour.
Training data limitations can reduce robustness when systems encounter different populations or conditions.
Operational success depends on how AI systems integrate into existing human and technical workflows.
Regulatory, organisational, and accountability requirements influence deployment feasibility.
Systems may perform differently when moved across domains, environments, or operational contexts.
HR7 and PFA are complementary components. HR7 provides the historical evidence foundation. PFA evaluates whether the currently available operational evidence supports AI development for a defined objective.
HR7 Historical Reference Framework and Predictive Feasibility Assessment (PFA) are not designed to replace existing AI evaluation, reporting, or governance frameworks. Existing frameworks provide important evaluation layers related to reporting quality, clinical evaluation, implementation monitoring, and governance. HR7 addresses a complementary question: What structural limitations are most likely to determine whether an AI system remains reliable once deployed in real-world operational environments?
HR7/PFA does not determine whether artificial intelligence is possible. It evaluates whether the currently available evidence justifies developing a reliable AI system for a defined objective.
The complete validation of the HR7 Historical Reference Framework and Predictive Feasibility Assessment (PFA) has been publicly archived through Zenodo. This publication documents the complete validation programme, including retrospective validation cases, historical validation cases, prospective frozen validation cases, and the integrated HR7/PFA assessment framework.
HR7 provides the historical evidence layer. PFA evaluates current operational evidence and translates findings into structured deployment guidance.