Understand the work before you rely on it.

No. Aperture does not claim to produce absolute or permanent truth. We test a defined material proposition against the lawful evidence available within an agreed scope and time period. The result distinguishes what is supported, what is not established, what remains uncertain and what further expertise or evidence may still be required.

Fact-checking may form part of the work, but the service is broader. A fact can be technically accurate and still be incomplete, outdated, dependent on one source or unsuitable for the decision being made. Aperture examines whether the material information is strong enough to carry the intended reliance.

No. Aperture evaluates the evidence and explains the responsible limits of reliance. The client or authorised decision-maker retains the decision. We do not exercise powers we have not been given, replace a court or regulator, or make a regulated professional determination outside our scope.

No. We do not buy, approve or make the decision for the client. We provide a decision-ready evidence record showing what may responsibly be relied on and what may not. The person or institution with the relevant authority remains accountable for the decision.

A clear question keeps the investigation proportionate and reviewable. It identifies what must be established, which evidence matters, what could change the answer and when the work is sufficient. Broad curiosity can produce a large volume of information without resolving the decision that actually matters.

A question is consequential when a wrong answer could materially affect money, safety, reputation, rights, governance, strategy, publication, procurement, a transaction or another significant outcome. The level of work should match both the likely consequence and the uncertainty in the evidence.

No. Aperture is designed for selective, evidence-intensive matters. Automated screening and standard background checks may be more appropriate for large volumes of routine checks. Aperture is strongest when a specific proposition, conflict, gap or decision requires independent examination and a traceable finding.

Human accountability remains attached to the scope, source weighting, materiality decisions, contradiction testing, specialist escalation, conclusion and final release. Tools may assist with search, sorting, comparison, calculation and production. They do not become the approving authority.

It means the quality, independence, relevance, completeness and limitations of the evidence are proportionate to the consequence of relying on it. A weak source may support an initial lead but not a major transaction. A plausible claim may justify further enquiry but not final approval.

Information is something available to read, hear or receive. Reliance begins when that information influences a real action, approval, publication, appointment, purchase or other decision. Aperture works at the point where information is about to acquire consequence.

Every material finding should state: 1. What is established? 2. What is not established? 3. What remains uncertain? 4. What level of reliance is justified for this decision? This keeps fact, inference, uncertainty and decision use separate.

No. “Supported” means that the available evidence supports the proposition within the stated scope, evidence cut-off, assumptions and limitations. New information, a better source, a changed fact or specialist analysis may later alter the finding.

Yes. A claim may sound reasonable or fit the known context while still depending on weak, incomplete, outdated or circular evidence. Plausibility can justify investigation. It does not by itself justify material reliance.

Interpretation built on an unverified premise can make an error more persuasive rather than less. Aperture first tests the material factual foundation, source origin and contradictions. Interpretation follows only after the record is sufficiently clear for the intended purpose.

It is a structured record that allows an authorised decision-maker to see the question, material claims, sources, contradictions, gaps, assumptions, uncertainty, conclusion and reliance limits without having to reconstruct the investigation from scattered research.

No. More information can add noise, repetition and false confidence. The aim is not maximum volume. The aim is the strongest proportionate record for the defined decision, with dependent repetition, weak sources and irrelevant detail kept from distorting the result.

AI reliance risk arises when an AI-generated claim, summary, recommendation, score or conclusion influences a real decision without independent testing of the evidence beneath it. The material issue is not merely whether AI was used. It is whether the resulting information is strong enough for the consequence that follows.

It becomes a business input when a person or system uses it in procurement, due diligence, management, governance, investment analysis, publication, compliance, strategy or another operational decision. At that point, fluent output is not enough. The material proposition and its evidence trail need to be tested.

It is the examination of material claims connected to an AI-assisted decision. The work may trace sources, test vendor assertions, identify missing or contradictory evidence, examine calculation and benchmark claims, distinguish generated synthesis from original evidence and state the level of reliance the available record can support.

No. AI can improve discovery, comparison and production. The problem begins when generated information receives more authority than its evidence warrants. Aperture focuses on the material claim, the evidence beneath it and the decision that may follow.

Yes, within an agreed scope. This may include claims about performance, accuracy, coverage, benchmarking, security, compliance, source quality, human oversight or operational benefit. Technical, legal, cybersecurity or specialist conclusions may still require appropriately qualified reviewers.

Yes. The review can focus on the material output being relied on, its source trail, omitted context, contradictions, calculations and decision use. Access to the underlying system may improve the review but is not always necessary if the relevant output and evidence record are available.

Yes. We can test whether material claims in an AI-assisted diligence product are supported, whether sources are independent, whether important contradictions were omitted and whether the stated conclusion exceeds the record. This does not replace legal, financial, technical or regulatory specialists where their judgment is required.

Yes. The focus is the proposition the decision-maker is being asked to rely on. We can separate source facts from generated interpretation, test decisive assumptions, identify missing evidence and state what the record can responsibly support before the recommendation is adopted.

Yes. Sampling, source tracing, omission checks, contradiction testing and material-claim verification can be used to assess whether the summary fairly represents the underlying record. The exact method depends on document volume, access, risk and the intended reliance.

Yes. A post-deployment review can examine whether AI output has begun influencing material decisions, which claims are being treated as fact, whether those claims have been independently tested and where monitoring, escalation or specialist control may be needed.

Not always. Some assignments can be completed from the output, prompts, citations, source documents, decision record and vendor materials. Other questions may require logs, technical documentation or controlled system access. The scope will identify what can and cannot be established from the available material.

The absence of a usable source trail is itself a material limitation. Aperture may search for independent evidence supporting the proposition, but it will not treat the generated output as proof. If the evidence cannot be reconstructed, the finding will restrict or refuse reliance.

Not by itself. A benchmark may test a different population, task, environment or definition of success. Aperture can examine the benchmark's source, methodology, independence, assumptions and relevance to the intended use. Specialist technical validation may still be required.

No. Governance defines roles, policies, approvals and controls. Evidence verification tests whether a material proposition is actually supported. The two can reinforce each other, but a policy document does not establish that a vendor claim, generated answer or decision input is reliable.

No. The client pays for the agreed investigation and deliverable, not for a preferred conclusion. Material contrary evidence remains visible. If the evidence does not support the expected answer, the finding says so.

The Central Question is framed so more than one answer remains possible. Material supporting and opposing evidence is retained, alternative explanations are tested where relevant and the conclusion is limited to what the record supports. Client preference does not change the evidence classification.

Potential conflicts are considered before acceptance and monitored during the matter. Aperture may request more information, narrow the scope, use safeguards or decline the engagement. Acceptance is not complete until fit, lawfulness, conflict and capacity have been reviewed.

Not automatically. Absence may be meaningful only if the missing evidence would ordinarily be expected to exist, the search was capable of finding it and plausible explanations for non-discovery have been considered. The conclusion must distinguish “not found” from “did not happen”.

Negative evidence is the meaningful absence of an expected record, event, trace or relationship. It requires careful reasoning about what should exist, where it should appear and whether the search was sufficient. Silence in one source is not automatically proof of absence.

Uncertainty is stated directly through evidence gaps, competing explanations, confidence language, assumptions, unresolved contradictions and reliance limits. It is not converted into confident wording merely to make the report sound decisive.

Yes. New material evidence, corrected records, changed facts or specialist findings may alter the assessment. Where continuing reliance matters, the client can request an update, monitoring scope or formal reassessment using a new evidence cut-off.

Selective acceptance protects quality, independence and professional boundaries. A matter must justify independent verification and be lawful, sufficiently defined, free of disabling conflict and feasible within available capacity. Work that does not meet those conditions is declined or redirected.

Send the question, what you are about to rely on, why it matters, the relevant deadline and what happens if the information is wrong. A short factual description is enough for the initial fit review. Do not send a sales pitch or a large confidential evidence pack at first contact.

Do not initially send identity documents, privileged material, medical records, credentials, unlawfully obtained data, very large document sets or information unrelated to the question. Aperture will identify an appropriate transfer route if the matter is accepted for further assessment.

No. They are entry points, not packaged promises. Two matters with similar labels may require different evidence, professional boundaries, jurisdictions and outputs. The final scope follows the actual question and intended reliance.

Yes. Relevant pages may provide Share, Copy link and Email link controls where those actions genuinely help the reader. Sharing public content does not change its copyright status, create an engagement or authorise reliance beyond the limitations stated on the page.

Use them to recognise the closest evidence problem, not to select a rigid package. Open the situation, compare its evidence question and specialist boundary with your decision, then send the material question so Aperture can define the smallest responsible scope.