S12.05 - Sector_playbooks

S12.05 ? Sector playbooks

flowchart LR
    A[Cross-sector GenAI use
Different risks, workflows and oversight needs] --> B[RAIDT
Run-level evidence framework]
    B --> C[[Sector playbooks
Domain-specific operationalisation]]
    C --> D[Run-level evidence pack]
    C --> E[Five-pillar score profile]
    C --> F[Sector-specific review guidance]
    D --> G[Reviewer reconstruction]
    E --> H[Governance readiness]
    F --> I[Organisational learning]
    J[Healthcare] --> C
    K[Finance] --> C
    L[Education] --> C
    M[Public services] --> C
    N[Cybersecurity] --> C

? Star S12 - Programme Architecture and Supervisory Navigation

Star context: Clarifies how RAIDT can remain a single coherent run-level governance framework while still being translated into domain-specific forms for supervision, implementation, empirical testing and policy discussion.

Academic picture

Definition / background

Sector playbooks are domain-specific implementation guides that apply the same RAIDT evidence logic and scoring logic to different organisational settings such as healthcare, finance, education, public services, supply chain or cybersecurity. They do not alter the conceptual core of RAIDT. Instead, they translate the core into the vocabulary, workflows, risk priorities and review expectations of a particular sector.

Conceptually, the idea arises from a common problem in AI governance: a framework may be general enough to travel across contexts, but if it remains only at the level of principles, organisations struggle to use it in live settings. RAIDT addresses this by treating the run as the unit of governance and requiring evidence that a configured use of generative AI can be reviewed in context. Sector playbooks extend that operational logic by showing what counts as relevant evidence, which risks deserve emphasis and how the same five pillars should be interpreted in different domains.

This matters because sector variation is real, but core governance fragmentation is risky. If every domain invents its own scoring basis, the framework loses coherence. If a framework ignores domain differences, it becomes superficial. Sector playbooks sit between those extremes. In RAIDT, they preserve the common structure of run-level evidence packs and score profiles while tailoring implementation to sector-specific tasks, controls, accountability routes and supervisory concerns.

They therefore belong inside RAIDT as an implementation layer rather than as a separate theory. The run-level evidence pack remains the core practical output, the five-pillar profile remains the comparative governance output, and sector playbooks explain how those outputs should be produced and interpreted in a field-sensitive way.

Why this concept matters

Sector playbooks solve a translation problem. Organisations often understand that AI governance should be responsible, auditable and traceable, but they do not know what those commitments mean for a specific deployment in a specific sector. A generic governance statement is rarely enough for a hospital team, a university administrator, a financial compliance reviewer or a cyber defence unit deciding whether a run was acceptable.

Without sector playbooks, two predictable problems appear. First, users may mistake implementation examples for the core framework and assume RAIDT is only for one sector. Second, organisations may adopt RAIDT in name while producing inconsistent evidence, because they do not know how to interpret the same governance expectations across different operational realities. The result is weak comparability, weak reviewability and weak audit readiness.

Sector playbooks matter because they let RAIDT move from abstract principle to operational governance without losing conceptual unity. They show how the same framework can travel across domains while remaining evidence-led, reviewable and contestable.

Key idea: Sector playbooks matter because they make RAIDT usable in real domains without changing RAIDT's core run-level evidence model.

What this item enables

Translation of RAIDT's common governance logic into sector-specific practice without rewriting the framework itself.
Consistent production of run-level evidence packs across different organisational domains.
Sector-sensitive interpretation of the five-pillar score profile while retaining comparability of structure.
Better supervisory explanation of how RAIDT scales beyond one paper, dataset or implementation case.
Clearer empirical testing because domain use cases can be examined through a shared evidence model.
Stronger organisational uptake by giving practitioners examples, templates and decision pathways they can actually apply.
Better policy discussion because the framework can be shown to operate in context rather than only in principle.

Practical example / likely audience question

Audience question

Are sector playbooks the real substance of RAIDT, with the framework itself acting only as a generic wrapper?

Answer

The concern behind this question is understandable because practical examples often look more concrete than the abstract framework that supports them. If a healthcare playbook or finance playbook appears detailed and actionable, an examiner or practitioner may conclude that the playbook is the true method and RAIDT is only a loose umbrella.

That interpretation is incorrect. The direct answer is that sector playbooks are not the core of RAIDT. The core is the run-level evidence framework: the run as the unit of governance, the evidence pack as the review artefact and the five-pillar score profile as the comparative governance output. Sector playbooks are implementation pathways that show how this same logic is operationalised in particular domains.

A practical example makes the distinction clear. Suppose RAIDT is applied to a hospital triage-support drafting tool and to a financial complaints summarisation system. The sectors differ in workflow, accountability, sensitivity and review thresholds. However, both still require evidence about the configured run, both still produce an evidence pack, and both still generate a score profile across Responsibility, Auditability, Interpretability, Dependability and Traceability. The playbook changes the interpretation and evidence emphasis, not the underlying RAIDT method.

RAIDT handles this better than a generic AI governance approach because it does not stop at saying that sector context matters. It gives a stable run-level governance structure and then allows sector playbooks to translate that structure into context-specific practice.

Practical example in RAIDT terms

Consider a healthcare use case in which a generative AI system drafts discharge summaries for clinicians. The run-level issue is not simply whether the model is capable in general, but whether this configured use, at this time, for this workflow, under this level of oversight, is governable and reviewable.

A healthcare sector playbook would specify the evidence needed for that run: task definition, user role, prompt template, model version, review workflow, escalation route, known clinical boundaries, record-handling controls, output correction history and justification for use in a clinical documentation setting. It would also show how the five pillars should be read in that context. Responsibility might focus on clinician accountability and approval thresholds. Dependability might emphasise error tolerance and workflow robustness. Traceability might require stronger logging and reconstruction because patient-facing records are involved.

The result is improved governance readiness. Instead of relying on a general claim that the organisation uses responsible AI, RAIDT can assemble a run-level evidence pack and score profile that are legible to healthcare reviewers. The sector playbook does not replace the framework; it makes the framework governable in a high-stakes domain.

Detailed link to RAIDT

Sector playbooks link to RAIDT in four ways.

First, they protect RAIDT's core idea by showing that one run-level governance framework can travel across multiple domains without collapsing into unrelated sector-specific methods.

Second, they connect directly to the run because each playbook clarifies how a specific configured use of generative AI should be described, evidenced and reviewed in context.

Third, they shape the production and interpretation of the evidence pack and the RAIDT score profile by indicating which evidence types, review criteria and sector expectations should be foregrounded.

Fourth, they strengthen reviewability, contestability, audit readiness and organisational learning because they make it easier for internal and external reviewers to understand why a run was acceptable, weak or in need of intervention within a particular field.

Sector playbooks ? Run-level evidence ? Evidence pack ? RAIDT score profile ? Governance readiness

This chain matters because RAIDT becomes practically transferable through sector playbooks while remaining conceptually unified.

Link to the five RAIDT pillars

Responsibility

Sector playbooks help specify who is accountable for a run in a given field, what approval structure is expected and where human judgement must remain decisive.