Human accountability remains
RILayer does not make consequential decisions for organisations or individuals.
Governance
AI may propose. Governed controls determine what may proceed.
RILayer is designed around a simple principle: consequential governance should not depend solely on an AI system following instructions correctly. Human accountability, professional boundaries and reviewable evidence must remain visible.
Governance position
RILayer governs the judgement structure around decisions.
It does not become the decision-maker, professional authority, statutory service, clinical tool, legal adviser or financial adviser.
Human accountability
Professional boundaries
Governed routing
Reviewable evidence
Core governance principles
The infrastructure preserves responsibility rather than replacing it.
RILayer is useful only when its role remains clear. It structures the environment around interpretation, evidence, boundaries and action while preserving the authority and duties already held by the organisation.
Professional responsibility remains
Clinical, legal, financial, HR, safeguarding, regulatory and statutory duties stay with authorised professionals.
AI may propose
AI can help interpret information and surface possible routes, but it does not authorise action.
Governed controls determine what may proceed
Evidence, Decision Readiness, Governance Risk, boundaries, authority and routing controls define the permitted route.
The organisation owns the outcome
Decision ownership, evidence retention and later review remain visible and attributable.
Boundaries are explicit
RILayer is not therapy, diagnosis, legal advice, financial advice, clinical judgement, careers guidance or automated authority.
Governance should make responsibility more visible, not give the organisation a new reason to avoid responsibility for its decisions.
Professional and information boundaries
Six boundaries define where RILayer must stop, route or preserve authority.
Where RILayer is tested, its role is to structure evidence, assumptions, Decision Readiness, Governance Risk, boundaries, routing and accountability before action. The relevant organisation and authorised professionals retain responsibility.
01
Input boundary
What information, AI output, concern, request or pressure is entering the judgement environment?
02
Interpretation boundary
What is known, assumed, inferred, missing or being treated as more certain than the evidence supports?
03
Authority boundary
Who is authorised to act, approve, escalate, pause or request further evidence?
04
Professional boundary
Which regulated, statutory, clinical, legal, HR, safeguarding or sector responsibilities must remain outside RILayer?
05
Evidence boundary
What must be retained so the organisation can explain and review what happened?
06
Information boundary
What candidate- or person-controlled information must not become decision evidence merely because it exists?
Decisions remain owned by the organisation and its authorised people.
Governance in the demonstrator
What RILayer may do — and what it must never silently become.
YT-WM-001 turns the governance principles into a controlled case. A simulated ATS recommendation remains an input, not an authorised decision, and the case cannot silently move from adverse machine output to human action without evidence and boundaries being made visible.
AreaRILayer mayRILayer must not
Evidence
RILayer may
expose evidence, assumptions and missing or conflicting information
RILayer must not
invent missing evidence or treat unsupported inference as fact
Machine output
RILayer may
retain the machine score or recommendation as a visible input
RILayer must not
turn a machine score into authority or hide it behind a replacement RILayer score
People
RILayer may
protect dignity, agency and relevant information boundaries
RILayer must not
determine a person's worth or capability
Professional judgement
RILayer may
surface boundaries and route a matter for proportionate human review
RILayer must not
make the final recruitment, careers or professional decision
Authority
RILayer may
make Decision Readiness, Governance Risk and routing conditions explicit
RILayer must not
bypass professional, organisational or statutory responsibility
Information
RILayer may
retain authorised evidence required for review
RILayer must not
convert protected or irrelevant personal information into suitability evidence
Governance beyond the prompt
Consequential control cannot depend only on whether an AI model follows instructions correctly.
AI can help interpret information and surface possible routes. Governed controls determine what actions are permitted, when human judgement is required, what evidence must be retained and what must be escalated or paused.