• integrated_82#QC1

In one sentence, RAIDT is a run-level evidence framework for responsible governance of generative AI in organisational work that produces a run-level evidence pack and a score profile for each configured use.

That sentence is academically important because it compresses the framework's full logic without losing its defining commitments. First, RAIDT is concerned with governance of actual use, not with model description in the abstract. Secondly, its unit is the run, meaning one configured use in a specific task and context, which is why the papers insist on run as the unit of governance. Thirdly, the framework is evidence-centred: governance claims are not accepted merely because an organisation has policies, model cards, or general controls, but because a recorded run can later be reconstructed, checked, and scored.

The sentence also captures RAIDT's dual output logic. The run-level evidence pack is the proof object that records prompts, configuration, retrieved context where relevant, outputs, and checks. The score profile is the measurement layer that expresses governance readiness across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability), interpreted through anchors 1=missing / 3=partial / 5=audit-ready. In short, the one-sentence definition works because it names both the ontology of the framework and the practical artefacts through which it operates.

A healthcare service asks GenAI to summarise a chest-pain consultation note. Using the one-sentence definition, staff immediately know what RAIDT requires: treat that single configured use as the governed object, create a run-level evidence pack, and produce a score profile rather than relying on a fluent summary alone.

The evidence pack would include the prompt constraints, model and tool settings, any retrieved material, the output, and the recorded oversight decision. Reviewers could then judge whether the run handled uncertainty, escalation triggers, and safety checks appropriately. The sentence is useful precisely because it translates directly into action: capture the run, preserve the evidence, and score the governance quality of that use.

• 00-RAIDT_Wording_v2
• 11-RAIDT_Academic_Logic_M_v11