Skip to main content
RILayer

Evidence & validation

Prove what works. Record what does not. Scale only what the evidence supports.

RILayer is at proof-of-concept stage. A controlled demonstrator exists and runs preconfigured scenarios. Controlled live validation in an operating environment has not yet begun.
Examine YT-WM-001View validation method

Evidence position

A demonstrator is evidence of a mechanism — not evidence of live outcomes.

The current browser build is a controlled technical artifact. External pilot evidence remains the next validation step, not a claim already made.

Claims bounded
Failures retained
Remediation recorded
Human sign-off

Evidence is built in bounded lanes

Controlled proof is kept separate from live operating validation.

RILayer separates what the current demonstrator can show from what still requires evidence in a real operating environment.

Lane 1 — Controlled Demonstrator / Youth Transition

Status: CURRENT CONTROLLED PROOF

The existing demonstrator and prior Youth Transition/PAVRA work provide a separate design, usability and governance evidence lane. They do not establish general transferability, live organisational outcomes or market validation.

Lane 2 — Bounded Live Operating Validation

Status: PRE-ACTIVATION

RILayer is preparing a separate bounded live operating environment to test whether the governance mechanism can surface material issues, preserve provenance and uncertainty, and support meaningful documented human judgement before action.

Controlled live case capture has not begun under this lane.

Reference proof-of-concept · YT-WM-001

The Chloe Taylor case is a bounded controlled evidence artifact.

A fictional adult early-career case begins with a simulated machine recommendation: DO NOT PROGRESS. The original simulated machine input is retained for audit continuity, but it is not a RILayer score or a validated measure of candidate suitability. The purpose is to test whether RILayer can govern the reasoning between that recommendation and accountable human action.

What the current proof demonstrates

  • The original simulated machine recommendation remains visibly separate from governed human judgement.
  • Evidence is separated from inference, missing information and restricted factors.
  • Decision Readiness is treated as a governance condition rather than a score of Chloe.
  • Governance Risk is surfaced as risk in the decision process, not risk about the person.
  • Professional, system and information boundaries remain explicit.
  • The case is routed rather than automatically decided.
  • Employer and candidate decisions remain separate and attributable.
  • The Decision Evidence Record preserves the machine input, evidence change and human reason.
Launch the demonstratorView canonical reference record

What this does not prove

  • candidate outcome effectiveness
  • organisational performance improvement
  • live employer adoption
  • cross-sector transferability
  • production-scale reliability or security

These exclusions are part of the evidence discipline. They define the boundary between what has been demonstrated, what is under validation and what remains a future hypothesis.

Current maturity

Proof-of-concept, with live operating validation not yet begun.

RILayer is at proof-of-concept stage. A controlled demonstrator exists and runs preconfigured scenarios. Controlled live validation in an operating environment has not yet begun.

Current priorities are improving first-use comprehension, strengthening security and authentication controls, making governance and reporting outputs clearer, and testing the proposition through bounded discovery and pilot relationships.

First-use comprehension
Security and authentication controls
Governance and reporting outputs
Bounded discovery and pilot relationships

Validation discipline

A six-stage route from claim to governed evidence.

A credible validation programme records not only what passed, but also what failed, why it failed, what changed and whether the rerun supports a narrower or wider claim.
  1. 01

    Define the testable claim

    State the specific decision-process change a controlled test is intended to examine.
  2. 02

    Freeze the baseline

    Record the current workflow, evidence, authority, risk pattern and expected outcomes before intervention.
  3. 03

    Run bounded cases

    Apply the controls to representative and stress-test scenarios without treating proposals as authorised decisions.
  4. 04

    Record deviations

    Capture failures, ambiguities, boundary breaches and unsupported outputs rather than hiding them.
  5. 05

    Remediate and rerun

    Correct the relevant logic, evidence or workflow and test failed cases again.
  6. 06

    Decide what the evidence supports

    Approve, limit, adapt, pause or reject wider testing based on retained evidence.

Failure is validation evidence too.

Deviation, remediation and rerun evidence should remain visible rather than being removed from the story because a test did not support the original expectation.

Future validation scenarios

Other contexts are hypotheses for adaptation — not current validation evidence.

The scenarios below show where the same governance architecture may later be tested. They are not current validation evidence and should not be read as claims of transferability.

Future scenario 01

Human–AI decision governance

AI-assisted workflows, copilots, dashboards, automation and agentic systems.

Validation question

Can human ownership, evidence requirements and escalation remain visible when automated outputs influence action?

Future scenario 02

Operational escalation

Service delivery, incidents, complaints and operational continuity.

Validation question

Can evidence, routing and decision ownership improve consistency without adding unnecessary bureaucracy?

Future scenario 03

Financial risk & underwriting

Credit, underwriting, exceptions and regulated risk decisions.

Validation question

Can override rationale, uncertainty and authority boundaries remain reviewable under pressure?

Future scenario 04

Health, care & people-sensitive judgement

Health, care, wellbeing and safeguarding-adjacent organisational settings.

Validation question

Can governance structure strengthen escalation while preserving professional responsibility?

Evidence before expansion

Scale only what the retained evidence can defend.

The next external conversation should test the mechanism and the validation design — not assume wider rollout is automatically justified.
View YT-WM-001Discuss a Bounded Problem