Public-interest research for organisational accountability.
Observed is a public-interest organisational research platform. It uses AI-assisted analysis of publicly available information, benchmarked against peer-reviewed academic research and recognised good-practice frameworks, to identify patterns in organisational behaviour that may indicate harm to workers, communities or public trust.
Why Observed exists
Individual cases get heard. Systemic patterns often do not.
Harmful organisational practices can be difficult to see clearly when public information is scattered across media reporting, workplace review platforms, public registers, funding records, regulator material, organisation-owned publications and stakeholder submissions.
Observed brings those public signals into a structured research process. The work compares what is publicly visible against what academic and professional research says good practice looks like.
The mission is not to accuse. It is to make the comparison available to the people who need it, while staying within clear evidence, fairness, privacy, legal and publication boundaries.
What Observed is
Applied organisational research meets public evidence analysis.
Observed sits closer to an academic research centre, public-interest analyst or organisational risk research function than to journalism, private investigation or legal advocacy.
Evidence-led
The work begins with public signals and source classification, not predetermined conclusions. Public information is assessed before any benchmark comparison is developed.
Benchmark-based
Observed applies research frameworks and good-practice standards that exist independently of the platform. The benchmark generates the standard.
Proportionate
Findings are source-cited, confidence-rated, limited and balanced. Contradictory signals are included where they exist.
What Observed is not
Clear boundaries protect the work and the people affected by it.
Observed is designed for public-interest accountability, not personal campaigns, private disputes, unsupported allegations or reputational attacks.
Observed does
- Analyse lawful and publicly available information.
- Compare public signals with academic and professional benchmarks.
- Identify patterns, gaps and accountability questions.
- Use AI-assisted workflows under human review.
- Apply suppression thresholds before named publication.
- Offer right of response before named findings are published.
Observed does not
- Act as a private investigator, law firm or regulator.
- Make findings of legal wrongdoing.
- Publish unsupported allegations.
- Operate as a smear campaign or revenge platform.
- Use private, unlawful or unattributable material.
- Go beyond what the evidence can responsibly support.
The organisation is named. The allegation is not made. The academic benchmark does the talking.
How research begins
Three pathways. One methodology.
Observed can initiate its own research, assess client-initiated requests, or complete commissioned research and publication support. All pathways pass through the same evidence, ethics, legal and human-review gates.
Platform-led research
Observed identifies organisations, sectors or issues of public-interest concern and completes its own research using the published methodology. No client is required.
Client-initiated research
Individuals, communities, advocates, unions, funders, boards, professional advisors or other stakeholders may request analysis of a named organisation, sector or issue.
Commissioned research
Where suitable, Observed may provide research, benchmark comparison, report generation, content production and publication support, subject to the same methodology threshold.
AI governance
A nine-pass research architecture, with human sign-off before anything leaves the system.
The AI analyst supports research consistency, classification and comparison. It does not make legal findings, infer motive or publish outputs. The human reviewer remains accountable for publication.
Signal collection
Collects public signals by source type and extracts classification categories.
Privacy scan
Removes named individuals and applies minimum sample suppression thresholds.
Legal risk scan
Flags active proceedings, potential defamatory republication and vulnerable-person risks.
Signal classification
Maps signals to dimensions such as leadership, governance, culture and complaint patterns.
Benchmark comparison
Compares classified signal clusters against academic and good-practice frameworks.
Gap analysis
Identifies benchmark gaps and accountability questions, not verdicts.
Confidence and limits
Rates findings based on source diversity, hierarchy and sample strength.
Contradiction check
Searches for public signals that contradict, moderate or balance the concern.
Human review gate
No output leaves the system without human review against the checklist.
Founding focus
Workplace harm and organisational culture risk.
Observed’s founding focus is workplace harm and organisational culture risk. This area has serious public-interest relevance, a strong academic research base and clear implications for workers, families, communities, boards, funders, unions, regulators and policymakers.
The work does not begin by labelling an organisation. It begins by asking whether public signals align with research indicators of psychosocial risk, leadership dysfunction, governance weakness, cultural harm or accountability gaps.
The founder brings postgraduate HR study focused on workplace bullying, combined with around 20 years of business analysis and process consulting experience across New Zealand and Australian organisations.
Named-organisation findings are suppressed unless signals are drawn from at least three independent source types. Multiple anonymous reviews alone do not meet the threshold, regardless of volume.
This protects against single-source patterns, thinly sourced findings, small-sample identification and confidence ratings that exceed what the available evidence can support.
Legal and ethical controls
The model is designed around caution before publication.
Observed does not treat publication as automatic. Publication only proceeds where the evidence, methodology, right of response, proportionality and human-review thresholds are met.
Public evidence only
Every signal must be publicly available, lawfully accessible and attributable. No exceptions.
Right of response
Named organisations receive a response window before named findings are published.
Conflict register
Potential conflicts are recorded and assessed before work proceeds beyond initial screening.
Corrections process
Published outputs can be clarified, corrected, updated or withdrawn when new evidence requires it.
Research process
The tranche process is a quality-gate system, not just a billing model.
All work moves through staged quality gates. Initial fit assessment tests whether the issue has public-interest relevance and belongs within the Observed evidence model. Evidence Check tests whether enough lawful public material exists. Research development fills information gaps. Benchmark comparison produces the core gap analysis. Right of response and publication review determine whether named publication is appropriate.
The process can result in a matter proceeding, narrowing, pausing, being declined, being referred elsewhere or being held for future sector-level research.
Built first as a serious methodology, not a volume content machine.
Observed is being developed through internal funding, proof-of-concept research, benchmark-library development, workflow testing and legal/ethical refinement before scaled commercialisation.
The initial phase is focused on validating the methodology, building the AI-assisted research architecture, curating the benchmark library, establishing the conflict register, testing the human review gate and preparing publication safeguards.
Future services may include paid evidence checks, organisation-specific analysis, information gap work, sector trend reports, stakeholder briefings, publication packs, subscriptions and access to report libraries or benchmark dashboards.
Observed also reserves capacity for limited public-interest access where people or communities affected by serious organisational harm cannot access paid support. Free access does not lower the evidence standard or bypass legal review triggers.
Have a public-interest concern that needs structured analysis?
Observed can assess whether an issue fits the evidence model, public-interest purpose and methodology threshold before any deeper research or publication decision is made.
Public evidence. Academic benchmarks. Better accountability.
