For CTOs, SaaS buyers, operating partners and technology investors
Know which AI claims actually apply to the product you are buying.
Know which AI claims actually apply to the product you are buying.A vendor has quoted accuracy, safety, cost savings or model capability statistics that your team is expected to rely on.
A benchmark result may use a different model, test population or operating environment than your proposed deployment.
A real person scopes, researches, checks and signs off the evidence work. Technology assists; it does not approve findings.
The decision on your desk
Approve the vendor’s claims for this use case, require more proof or stop the procurement?
Your organisation is close to approving an AI supplier whose sales deck promises accuracy and savings with limited method disclosure.
Stop unrelated benchmark scores from being treated as deployment proof.
Identify the one or two missing tests worth requiring before signing.
Give decision-makers a clear written basis for challenging an AI sales claim.
The actual work
What our people will examine for you.
Each line connects to a material question in your decision. We test what the record supports and what it does not, rather than producing a generic AI summary.
Freeze the exact promise, product version, deployment configuration and intended use.
Check benchmark design, denominators, comparator, date and data leakage controls.
Separate the provider’s own evidence from independently reproducible results.
Review disclosed failure modes, human oversight requirements and data-use claims.
Test simple savings and error-rate calculations against their stated assumptions.
Prepare concrete vendor evidence requests and go/no-go conditions for the buyer.
Your deliverables
A completed dossier your team can act on.
We turn the collected evidence into named findings, identified gaps and next actions. A human reviewer remains accountable for the scope, interpretation and release.
- Vendor-claim evidence matrix, up to ten claims
- Benchmark and method limitations memo
- Data and operating-dependency question list
- Quantitative assumption and denominator checks
- Unresolved-risk register and procurement conditions
- Human-reviewed buyer-facing assessment
Complete 11-page example of report structure, evidence reasoning, calculations and limitations. The illustrated case is fictional, not a previous client assignment.
Low-friction engagement
Know the price. Send a short brief. Receive the acceptance terms in writing.
Tell us the decision
Send the name, your deadline and the 2–3 claims or records you need resolved. No phone call, budget questionnaire or data room on first contact.
A real person checks fit
We check conflict, legal permissions, sources and competence. You receive a written scope, start date and the fixed fee or a clearly explained change if your matter falls outside the published limits.
We deliver evidence you can inspect
Our people test the sources, check calculations, record contradictions and deliver the report. You receive one written clarification round within the agreed terms.
The standard review
One clear fee. No surprise expansion.
Included scope: One vendor, one defined use case, ten material claims and up to 30 supplied source objects. Documentary assessment, not independent model testing, cybersecurity certification or performance guarantee unless separately scoped.
Timing: Agreed after the vendor evidence pack and decision date are identified.
For regular users: Request a reusable written standard-scope arrangement. Each matter remains separately priced at its standard approved fee, with no automatic discounts or subscriptions.
Read the complete engagement boundaries and ordering process ↗Ready to proceed?
Request the AI Vendor Performance & Claims Audit.
You're asking our team to confirm and accept this $2,950 standard bounded review. No payment is taken here. Your message reaches a person, not an automated research checkout.
If your email app does not open, write directly to info@apertureforensic.com with the service name and your decision question.
Before you send
Questions that matter to your purchase
Is the $2,950 fee fixed?
It is the proposed standard fee for the defined scope, pending final written commercial approval. We confirm the actual facts, jurisdiction, records, permitted work and final terms in writing before accepting it. We never add unapproved work to the bill.
Can I use my own AI system instead?
You can use AI for a preliminary review. What you commission here is human-directed independent source work, documented contrary-evidence testing, dated attribution, reviewed calculations and a defensible evidence handoff. If your internal process can already do that adequately, there may be no reason to buy this service.
Can you investigate a confidential matter?
Start with a short non-sensitive outline. A human reviewer confirms conflict checks, lawful source access, privacy safeguards and an appropriate transfer method before we receive confidential files.
Can my team use the review repeatedly?
Yes, subject to each matter being accepted. Repeat matters may use a reusable written scope and simpler onboarding, where appropriate. Each matter retains its standard approved fee and separate acceptance; no automatic discount or subscription applies.