# Investigate Provision layout note

## Status

Implementation note for the preferred Phase 6 `Investigate Provision` screen represented in `phase-6-govuk-disciplined-workbench-wireframes.html`.

Earlier three-column/notebook-side-panel experiments are superseded by the preferred workbench layout described here.

## Main comparison layout

The main investigation area uses three columns:

1. source provision;
2. generated Akoma Ntoso provision;
3. a narrow supporting-evidence rail.

The source and generated provision should remain the dominant simultaneous comparison. The evidence rail contains compact current context such as machine evidence, QA context, review state/history and routes to comparison/problem exploration.

The evidence rail is not the investigation notebook and should not become a dense case-management panel.

## Mandatory QA requirement panel

When the reviewer enters from a mandatory QA recommendation, a compact **Mandatory QA requirement** panel appears above the source/output comparison.

It should show:

- that the recommendation is mandatory;
- current QA state (`Not checked` or `Addressed`);
- a concise explanation of why the area was surfaced;
- the two explicit completion actions:
  - **Checked — no further action required**;
  - **Checked — review action required**.

Merely viewing the provision does not address a mandatory recommendation. QA completion remains separate from the normal semantic reviewer conclusion.

Advisory QA recommendations require no comparable completion panel/action; advisory context may still be shown where useful.

## Source/output context

Source and generated provision remain simultaneously visible. Local expansion may reveal the containing section or modification hierarchy.

Broader context is opened separately:

- source context on eISB where reliable deep linking is available;
- generated context in the separate read-only rendered transformed-Act reference view.

Raw XML/provenance remains available on demand.

## Investigation notebook

The **Investigation notebook** is full-width below the comparison area and may be vertically resizable/collapsible.

It contains working material such as:

- observations;
- comparison cases;
- AI suggestions;
- counterexamples;
- working hypotheses.

Notebook material is exploratory. It does not automatically become a reviewer observation, semantic conclusion or problem association. Explicit actions such as **Record as reviewer observation** promote material into durable review evidence where appropriate.

## Supporting investigation tools

Machine evidence, existing-problem lookup and comparison tools should remain compact and subordinate to the main source/output task. They may appear in the evidence rail and/or restrained expandable controls rather than replacing the main comparison.

The reviewer should be able to reach:

- fuller machine evidence and provenance;
- existing reviewer-problem lookup;
- similar/contrasting provisions;
- reviewer-directed search where available;
- review history.

## Durable conclusion action

The normal semantic conclusion model remains:

- Acceptable transformation;
- Source-caused issue;
- New problem identified;
- Existing problem;
- Reviewer observation only.

These outcomes are distinct from the mandatory-QA completion actions. A reviewer does not need to record **Acceptable transformation** merely to demonstrate that a mandatory QA inspection occurred.

State-changing conclusions should use an explicit decision/confirmation step. Leaving without a durable conclusion leaves the investigation unresolved.

## Visual hierarchy

The intended hierarchy is:

1. Act/provision/run and retained navigation context;
2. mandatory-QA requirement panel where applicable;
3. persistent source / generated AKN / supporting-evidence comparison area;
4. full-width investigation notebook;
5. explicit semantic reviewer-conclusion action.

This keeps QA completion visible without making it the semantic identity of the investigation, and preserves the source/output comparison as the central reviewer task.