The platform · Governance & trust

Built for trust from day one.

Rubric-aligned scoring. Evidence packs. Human review pathways. Accessible by design. Five design choices, made on purpose.

Five design choices

Commitments, not features.

Each one is a choice we made early about how the platform should behave in high-stakes contexts.

CHOICE 01

Rubric-aligned scoring

Every score is anchored to a criterion the partner can see. Published before the scenario, not after. No black box. No "AI says you passed."

CHOICE 02

Evidence packs

Every assessment is exportable, auditable, defensible. Per scenario and per cohort. Structured for accreditor, regulator, funder, audit.

CHOICE 03

Human review pathways

The assessor remains the decision-maker. The platform produces evidence; the human reviews it, agrees or disagrees, signs off.

CHOICE 04

Accessible by design

WCAG 2.1 AA minimum. Colour-independent meaning on every traffic light. Focus indicators. Keyboard. Screen reader. Reduced motion.

CHOICE 05

Fairness reviews

Rubrics audited for criteria that might disadvantage candidates. Persona library reviewed for cultural safety and stereotype. Bias monitored, not denied.

Bias testing

Tested across the axes that matter.

Scoring is reviewed for systematic disadvantage across accent, dialect, disability-related speech differences, and cultural communication style. The audit is documented; we don't pretend the audit is the cure.

Where a criterion is producing a disparate result without a defensible reason, the criterion is rewritten. Where a persona is encoding stereotype, the persona is reviewed by lived-experience advisors before it goes back into rotation.

How the persona engine handles diversity
Bias review · summary
Accent · 12 regional variants REVIEWED
Dialect & code-switching REVIEWED
Disability-related speech difference REVIEWED
Cultural communication style REVIEWED
Gender & age REVIEWED
Trauma-informed framing REVIEWED
Human review pathway

Higher stakes, more human review. Rule, not exception.

In higher-stakes contexts, the human review pathway is not an option — it's the design.

Conversation captured

Voice or text. Recorded with consent against the rubric.

First-pass score

Per criterion. With the moment that earned each score.

Assessor review

Agree, disagree, annotate. Calibration data alongside.

Sign-off recorded

The signed record becomes the audit artefact. Versioned.

What we won't do

Some commitments only make sense in the negative.

Six explicit things. Each one is the rule, not an exception we negotiate.

We won't make hiring decisions for you.

Where Attuniq is used in screening, candidates produce evidence and humans make the decisions. The platform is not the decider.

We won't replace the assessor.

The clinical educator, the trainer-assessor, the QA lead, the case manager — they remain accountable. Attuniq makes their evidence trail defensible.

We won't claim our scoring eliminates bias.

Rubrics can encode bias. Personas can encode stereotype. We audit them; we don't pretend the audit is the cure.

We won't deploy in high-stakes contexts without human review.

Higher stakes, more human review. That's the rule, not an exception we negotiate.

We won't black-box a score.

Every score has a criterion. Every criterion has a definition. Every assessor touch is logged.

We won't store data we don't need.

Data minimisation by design. Learner data belongs to the learner. Partner data belongs to the partner.

Audit, evidence, accreditation

For the partner whose conversation starts with their compliance lead.

Audit logs

Every score, every assessor touch, every rubric change, every persona configuration. The log is part of the evidence pack, not a separate report.

v

Version control

A score produced eighteen months ago is reviewable in the form the rubric existed at the time. Audit doesn't get harder because the platform improved in the interim.

Retention policies

Aligned to the regulatory timelines that apply to each partner sector. Explicit about what is kept, for how long, and who can access it.

A

Accreditation alignment

Claimed only where substantiated. We don't claim ASQA recognition. We don't claim AHPRA endorsement. Where the platform aligns to a standard, we say partial or full and we evidence the claim.

Accessible by design

WCAG 2.1 AA is the minimum, not the ceiling.

An EAL learner with low literacy, a clinical educator using screen-reader-only navigation, and a contact centre agent on a low-spec device all get the same product. Not a stripped version of it.

The brand guide treats accessibility as a core requirement, not a section. The platform reflects that.

WCAG 2.1 AA
Screen reader
Keyboard nav
Reduced motion
Captions
Transcripts
Big text
Focus rings
44px targets
Colour + label
Plain English
12 accents
A11y
core
Colour-independent meaning. Every traffic light is paired with a label, an icon, and a position — meaning is never carried by colour alone.
Touch targets meet 44 × 44 px minimum on mobile. Focus indicators visible on every interactive element.
Captions and transcripts ship with every audio scenario. Reduced motion is honoured.
Plain English in learner-facing surfaces. Especially the workplace English context, where readability is load-bearing.
Privacy & security

Not just compliance theatre.

Data minimisation is the default. Learners own their records by default. Partners control their cohort's data. Hosting region, encryption posture, and certifications held — and certifications in progress — are stated honestly on the security page.

We don't say "bank-grade encryption". We say what we use, where, and how it's reviewed. If a partner needs more than what's stated, the partner conversation is the next step.

Read the security page Read the privacy policy
Read more

The same posture, stated as a brand-wide commitment.

The principles page is where the trust posture is stated for the brand, not just the platform.