• qa_deck_100#slide 100 · Empirical programme, calibration, procurement, and assurance

In RAIDT, assurance carries procurement promises into ongoing oversight by using the same evidentiary object before and after acquisition. Procurement teams can ask suppliers for a run-level evidence pack, scoring rubric, and score profile during tendering, but those materials are valuable only if they remain usable once the system is deployed. RAIDT solves that continuity problem by linking ex ante supplier claims to ex post review of actual runs. The contract or tender can specify what evidence must be retained, what controls must be documented, and what minimum governance thresholds are expected across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability).

Once deployment begins, those promises are no longer rhetorical. Internal audit can sample runs like transactions; complaint handlers can inspect disputed uses; managers can compare workflows, teams, or vendors; and post-market monitoring can build time-series evidence about whether promised controls are genuinely operating. This is why RAIDT functions as an operational hinge between procurement and supervision. It turns vendor promises about transparency, oversight, and monitoring into reviewable records for each material use. It also matters that influence methods as governance interventions are documented over time: if a supplier changes prompting, retrieval augmentation, LoRA/PEFT, or RLHF-type controls, the governance consequences appear in the run record and score profile rather than being hidden behind a stable-looking procurement narrative.

A public-service authority procures a generative assistant to help caseworkers draft decision letters. In the tender, the buyer requires a run-level evidence pack for each consequential use and a score profile that shows whether the workflow meets agreed thresholds. That procurement promise is not left in the filing cabinet. Months later, when a citizen challenges a letter, the authority can inspect the exact run, including the prompt template, retrieved guidance, output hash, and reviewer checks.

If the supplier had promised strong oversight but the retained evidence shows no meaningful review step, the gap is demonstrable. If Traceability weakens because retrieval snapshots are no longer preserved, the organisation can escalate the issue through supplier assurance and internal governance, rather than relying on vendor reassurance. RAIDT therefore carries procurement commitments forward into live oversight, exception handling, and organisational learning.

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