Policy is not evidence
A policy may describe intent. BiasLens asks what the organisation can actually demonstrate about one real system.
Evidence-led algorithmic accountability
Know what your evidence supports — and what it does not.
BiasLens helps organisations assess one AI system at a time, separate evidence from assumption, identify bias risks, document uncertainty and build an accountable evidence trail.
If AI is influencing decisions about people, opportunities, access, recruitment, learning, financial services or essential services, confident claims are not enough. Your organisation needs evidence it can explain.
Evidence, not assumptionSystems, not peopleTraceable findingsUncertainty preserved
The market problem BiasLens is built to solve
Many organisations can name the AI tools they have purchased. Fewer can show what evidence supports how those systems influence decisions, which groups may be affected, what remains unknown or whether vendor assurances have actually been substantiated.
Evidence is separated from severity
A serious concern with weak evidence is not presented as a proven outcome. The distinction stays visible.
Known unknowns become governable
BiasLens helps turn uncertainty into a documented question that can be investigated, owned and acted on.
Who BiasLens is for
BiasLens is designed for organisations where AI influences decisions, opportunities, access, eligibility or participation.
AI Governance / Responsible AI
Understand what systems are in use, what evidence exists and where your governance position is strong or weak.
Risk / Compliance
Build traceable documentation rather than relying only on policy language or supplier assurances.
HR / People
Examine AI-assisted recruitment or workforce systems without confusing outcome differences with legal conclusions.
Procurement / Vendor Governance
Turn vendor claims about fairness, accessibility and oversight into questions your organisation can document.
Accessibility / Disability Inclusion
Ask whether disabled people are visible in the training, testing, interface and outcome evidence that supports a system.
Executive / Board Oversight
Move from high-level assurance to a clearer record of evidence, limitations, controls and unresolved risk.
What BiasLens can assess
BiasLens works best when an organisation brings one defined AI-enabled system, workflow or decision process into view. Scope is confirmed during qualification.
Example systems and workflows
- AI-assisted recruitment and candidate screening
- Workforce decision support
- Education, learning and assessment systems
- Financial, insurance or eligibility systems
- Public-sector or essential-service decision workflows
- Third-party AI tools affecting customers, employees or applicants
Questions BiasLens helps surface
- What evidence do we actually have about this system?
- Which affected groups are visible in the evidence?
- What vendor claims remain unverified?
- Where could preexisting, technical or emergent bias arise?
- Which outcome differences need further investigation?
- What should be documented for governance and oversight?
Start with one system. Build stronger assurance over time.
The commercial pathway is deliberately clear: diagnose the evidence gap, assess where needed, document what the organisation owns and reassess when the system changes.
Evidence Readiness Diagnostic
A focused front-door engagement for one AI-enabled system or decision process.
- System and decision-context summary
- Evidence and documentation inventory
- Bias-pathway and affected-group visibility review
- Known gaps and immediate governance questions
- BiasLens Evidence Readiness Brief
System Bias Assessment
A deeper assessment using the BiasLens methodology and traceable evidence discipline.
- Bias-risk findings and evidence strength
- Fairness analysis where appropriate
- Accessibility and affected-group considerations
- Limitations and known unknowns
- Recommendations and documented rationale
Algorithm Defence File
An organisation-owned evidence record showing what was assessed, what is known and what action followed.
- Evidence present and evidence absent
- Findings, rationale and controls
- Limitations and unresolved questions
- Actions taken and next steps
- Governance evidence — not a promise of legal immunity
Continuous Assurance
Periodic reassessment as models, populations, vendors, evidence and use contexts change.
- Reassess material changes
- Track emerging evidence and Bias Drift
- Review changes in affected populations
- Maintain stronger governance over time
- Embed evidence discipline into AI operations
Why BiasLens is different
BiasLens is designed to improve the quality and honesty of the evidence an organisation relies on — not to manufacture false confidence.
Uncertainty is preserved
“Not sure” is not silently converted into “No risk”. Missing evidence remains visible.
Fairness signals are handled carefully
An outcome difference can justify investigation. It does not automatically establish causation or unlawful discrimination.
Small samples are guarded
BiasLens uses cautious methodology guardrails to reduce false certainty and privacy risk. They are not legal safe harbours.
Findings are traceable
Classifications are designed to record rationale, confidence, limitations and recommended next steps.
Accessibility is an evidence question
A successful demonstration of one accessible interaction path is not the same as formal accessibility conformance.
Systems, not people
BiasLens assesses systems, processes and aggregated outcomes. It is not employee monitoring, productivity surveillance or individual scoring.
How a BiasLens engagement works
The aim is to move one real system from vague concern toward documented evidence and an explicit next decision.
1
Choose one system
Identify one AI-enabled system or decision process that matters.
2
Review evidence and context
Examine available documentation, controls, affected groups and known gaps.
3
Document findings clearly
Record what is supported, what is uncertain and what requires investigation.
4
Act and reassess
Strengthen controls, deepen assessment or monitor material change over time.
Proof and trust
BiasLens is being taken to market with the same evidence discipline it asks of clients.
Proof assets
- BiasLens Methodology Note — public methodology and claims boundaries.
- Two-minute problem-to-evidence demo — fictional-data walkthrough and narration.
- Fictional-data recruitment case study — end-to-end evidence walkthrough using a fictional AI-assisted recruitment scenario. Download PDF.
- Sample Algorithm Defence File — fictional organisation-owned governance evidence record. Download PDF.
Boundaries stated openly
- BiasLens does not prove discrimination.
- BiasLens does not replace legal advice.
- BiasLens does not turn incomplete evidence into reassurance.
- BiasLens does not assess people as individual risk objects.
Why this product exists
I have lived for more than thirty years with the consequences of institutions making assumptions about disabled people. BiasLens comes from a simple conviction: when a system can affect someone's opportunity, livelihood, access or participation, assumptions are not enough. Organisations should be able to show what their evidence supports — and what it does not.
Ready to assess one AI system?
Complete a short qualification form first. We will use it to understand the system, decision context and evidence question before deciding the most appropriate BiasLens engagement.