• workshop_dense_100#slide 30

In the RAIDT papers, GenAI in organisations means generative AI embedded in work processes where outputs influence records, recommendations, decisions or stakeholder communications. It is therefore broader than a standalone chatbot. It includes drafting, summarising, explaining, recommending and communication tasks that sit inside organisational routines and may later be relied upon by staff, managers, auditors or affected stakeholders. The key point is that the relevant object is not only the model in abstraction, but the configured organisational use of that model in context.

That definition matters in RAIDT because organisational GenAI is socio-technical and run-dependent. A single output can be shaped by prompt templates, retrieved documents, active tools, alignment controls, adapter layers and human review. RAIDT therefore uses run as the unit of governance and treats the run-level evidence pack as the proof object for one configured use. From that evidence, the framework derives a score profile across the five pillars (Responsibility, Auditability, Interpretability, Dependability, Traceability), so governance readiness is inspectable rather than merely asserted.

This is why GenAI matters in RAIDT: it turns routine knowledge work into a governance-intensive activity. The framework does not ask only whether the system was useful. It asks whether a particular organisational use can later be reconstructed, reviewed, compared and challenged. The anchors 1=missing / 3=partial / 5=audit-ready make that judgement operational, and they reflect the wider RAIDT claim that influence methods as governance interventions must be evaluated for evidentiary quality as well as task performance.

A cybersecurity operations centre uses GenAI to triage alerts, summarise log evidence and draft a recommended severity label for an analyst. This is plainly GenAI in organisational work: the output is not casual text generation, but decision-adjacent material that may influence incident handling and later reporting.

In RAIDT, the case matters because the team must be able to show how that recommendation was produced. A proper run-level evidence pack would record the prompt template, model and deployment version, retrieved artefacts or tool-call traces, the generated summary, and the analyst?s accept, amend or reject decision. The resulting score profile then shows whether the run was responsible, auditable, interpretable, dependable and traceable enough for operational reliance, rather than assuming that a useful-looking output was automatically governable.

• 11-RAIDT_Academic_Logic_M_v11
• 15-RAIDT-IS-Governance_M_v07