Human-controlled. Technology-enabled. Evidence-led.

AI can accelerate research. It cannot establish whether a material claim is safe to rely on.

The question is not who produced the information. The question is whether the evidence is strong enough to rely on. Aperture independently tests material AI-assisted claims, citations, sources, contradictions and gaps while people remain accountable for the final finding.

AI reliance

The question is not whether AI was used.

The question is whether the evidence is strong enough for the decision that follows. Aperture tests the material proposition when generated information becomes organisational reliance.

AI USE → MATERIAL CLAIM → RELIANCE → CONSEQUENCE → EVIDENCE TEST

Explore AI Reliance Verification

The controlled system behind the output

A report is the visible end of a much larger process.

Aperture Research Works does not begin with a prompt and end with a polished document. Its current controlled operating framework extends across formal acceptance, evidence, regulatory, testing, quality, human-review, release and lifecycle controls.

a controlled internal frameworkProtected internal operating architectureHuman review and approval

The public framework explains the principles sufficiently for clients to evaluate the service. Detailed non-public methodology, workflows, templates and controls remain proprietary and confidential.

What technology may assist with

  • Source discovery and comparison
  • Chronology construction
  • Document and structured-data review
  • Contradiction detection
  • Calculations and analytical support
  • Drafting and visualisation
  • Quality-control assistance

The operating rule

Technology-enabled. Human-controlled. Human-evaluated. Human-accountable. Human-approved.

Technology does not become the final authority.

The tools may change. The evidentiary standard does not. New models, agents or software can be adopted without making the method dependent on any one vendor or technology.

Artificial intelligence is not treated as evidence.

Model memory, generated recollection, apparent familiarity, repetition or confidence cannot establish a material fact. Material factual propositions in professional work must have an identifiable evidentiary basis.

Human judgement remains responsible for materiality, fairness, source weight, professional boundaries, specialist escalation, conclusions and final approval.

AI-generated research can itself require evidence assurance.

Aperture can test generated claims, citations, omissions, independent corroboration, unsupported inference and whether the final output is stronger than its underlying evidence.

AI Research & Citation AuditSee a Real Claims & Evidence Audit: MH370

The MH370 case is not presented as an AI-generated research audit. It is used to show the real claim-tracing, independent-corroboration, contradiction and evidence-gap controls that also matter when AI-assisted research is reviewed.

Client outcome

What this means for you

Aperture helps turn scattered, disputed or uncertain information into a clearer record of what can and cannot responsibly support your decision.

See what is established

Separate verified facts and supported findings from claims, repetition, inference and assumption.

See what could change the answer

Keep contradictions, missing evidence, alternative explanations and specialist dependencies visible before reliance.

Know the responsible next step

Understand what the evidence supports now, what it does not support and whether further work is proportionate.

Human accountability remains the control point.

Discuss a Matter

Human control and professional accountability

Research is conducted under human control, with technology used to support evidence discovery, organisation and analysis. Final evaluation and conclusions remain human responsibilities.

HUMAN AUTHORITY

AI may assist. It does not make the final regulated or rights-affecting decision.

AI may help with research, comparison, organisation, contradiction detection and analysis. It may not autonomously determine criminality, legal liability, employability, creditworthiness, tenant or insurance eligibility, medical eligibility, final sanctions status or another material rights-affecting conclusion concerning a natural person.

AI-assisted decision evidence validation

THE QUESTION IS NOT WHETHER A SENTENCE WAS WRITTEN BY AI.

The question is whether the material factual proposition beneath the analysis is sufficiently supported for the intended reliance.

  • Do cited sources exist and support the proposition attributed to them?
  • Have the correct entity, period and jurisdiction been matched?
  • Do apparent sources trace back to one originating claim?
  • Were important primary records omitted?
  • Has interpretation been presented as established fact?
  • What stronger authoritative or private evidence is still required?
Explore the Capability
Discuss a Matter
V168