• integrated_82#Q4.17

Across the three papers, RAIDT is presented as the operational evidence layer that connects high-level governance frameworks to inspectable organisational practice. The NIST AI RMF supplies the managerial structure of Map, Measure, Manage and Govern, while the NIST GenAI Profile sharpens that structure for generative use cases in which risk depends on prompt structure, retrieval context, tool calls, adaptation layers and human review. RAIDT responds by treating run as the unit of governance: one configured use for one task, at one time, in one context. In that sense, RAIDT does not compete with NIST. It gives organisations a way to operationalise the documentation, oversight, monitoring and post-hoc review that the RMF and GenAI Profile increasingly expect where harms arise through situated use rather than model description alone.

More specifically, RAIDT translates those expectations into a run-level evidence pack and a score profile grounded in the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability). The policy pathways paper maps Responsibility mainly to Map and Govern, Auditability and Traceability to Measure, Dependability to Manage, and Interpretability to Govern as stakeholder-facing intelligibility. The audit-accountability paper then explains why this is more than a logging exercise: the purpose is reconstructability, contestability and comparable review. Accordingly, RAIDT uses anchors 1=missing / 3=partial / 5=audit-ready and treats influence methods as governance interventions, so that prompts, retrieval, alignment and adaptation are governed as evidence-bearing parts of use. RAIDT therefore functions as the missing run-level proof object through which organisations can show how NIST AI RMF and the GenAI Profile are enacted in practice.

Consider a public-service eligibility team using GenAI to interpret benefit rules for a claimant. Under the NIST AI RMF, the department must map the decision context, measure whether the output is grounded, manage error and appeal risk, and govern the workflow through oversight. The GenAI Profile makes the run-time character of that task especially important because the answer may depend on the exact prompt template and the retrieved policy clause. RAIDT therefore requires a run-level evidence pack containing the prompt version, model version, retrieval snapshot of the rule passage, output hash, and recorded reviewer check.

The resulting score profile lets managers see whether the advice was merely fluent or genuinely governable. If retrieval evidence is absent, Auditability and Traceability fall towards the anchors 1=missing / 3=partial / 5=audit-ready, prompting escalation or redesign. If structured prompting and retrieval are logged as influence methods as governance interventions, the organisation can later justify the advice to internal audit or to a claimant who challenges the outcome, showing concretely how NIST expectations were implemented in one case.

• 10-RAIDT_Policy_Pathways_M_V50
• 14-RAIDT-Policy-Motivation_M_v11
• 16-RAIDT-Audit-Accountability_M_v05