A real KYB case,
step by step.

Meridian Payments Ltd — a UK-incorporated payments firm with operations in Singapore — submits a corporate application through your branded onboarding portal. The applicant enters company details, declares two directors and three beneficial owners, and uploads a certificate of incorporation, an articles of association, a register of members, and a recent utility bill. The application lands in First Mile Labs as a single case record, timestamped and stamped with the applicant's IP and user agent. No analyst has touched it yet.

Within seconds of submission, every uploaded file is classified by document type and routed to the appropriate AI extraction pipeline. The certificate of incorporation yields the company number, incorporation date, and registered office; the register of members produces a structured table of shareholders and ownership percentages; the utility bill is parsed for the address and date of issue. Confidence scores are attached to every extracted field. The case dashboard now shows a fully populated company profile — without an analyst opening a single PDF.

Meridian's company number is checked against the public company register in real time. The platform compares the live register data — directors, persons with significant control, filing history, charges, and active status — against the documents the applicant submitted. Two of the declared directors match the register exactly. The third — a recently appointed non-executive — is not yet filed on the register, and the case timeline records the discrepancy for analyst attention. The platform also reconciles the declared ultimate beneficial owners against the PSC register, flagging one UBO whose ownership percentage differs by 4 points from the public filing.

Every individual on the case — directors, beneficial owners, and the legal entity itself — is screened against global sanctions lists, PEP databases, and adverse media sources via the AML screening engine. Meridian itself returns clear. Two directors return clear. The third director matches a Singapore PEP record — a senior government adviser sharing the same name and date of birth — with a 91% confidence score. Adverse media surfaces a 2023 Financial Times article naming one beneficial owner in a civil dispute. Both hits land on the case as structured alerts, scored and attributed to their source.

Before the case reaches an analyst, the AI triage layer reviews each screening hit and proposes a disposition. The PEP match is graded "true match — confirm and escalate," with a short rationale citing the matching DOB and country of residence. The adverse media article is graded "false positive — different individual," after the AI compares the article's named individual against the date of birth, nationality, and address on file. Every triage decision is attached to the hit with the model version and prompt version used, ready for human review.

The platform identifies who on the case needs identity verification based on your configured policy — in this case, all directors and any UBO above 25% ownership. Five unique individuals are identified after the dispatch pipeline dedups people wearing multiple hats (one director is also a 30% UBO and is checked once, not twice). For each person the platform first runs an instant government data check — a non-document, server-to-server lookup against authoritative databases using the name, date of birth, and address already on file, requiring no action from the customer. Three are confirmed immediately with nothing to do. The remaining two — including Meridian's Singapore-based non-executive director, whose country isn't covered — fall back automatically to a branded one-time link to scan a passport and complete a liveness check on their phone. Results flow back automatically.

With documents, registry data, screening results, and IDV outcomes in place, the risk engine calculates a composite score using the weights your institution has configured. Meridian scores in the high band, driven by the confirmed PEP exposure, the cross-jurisdictional ownership structure, and the payments sector. The risk band map says high-band cases must go to an analyst with full EDD before a decision can be reached — so the platform routes the case to your queue rather than auto-approving it.

Because the case is high-risk and involves a PEP, the EDD engine generates a per-case checklist: source of funds for the company, source of wealth for the PEP director, an organisation chart confirming the ownership hierarchy, and the Singapore regulator's licence number for the payments operation. The applicant receives a single, AI-drafted outreach email listing each item with the regulatory reason it is required, rather than a series of disjointed requests from different team members. The case timeline records the request and waits.

Once the EDD evidence arrives and is verified, the analyst opens the case to a single screen: company profile, directors, beneficial owners, screening hits with AI triage, IDV outcomes, EDD evidence, and the composite risk score with a factor-level breakdown. The PEP match is confirmed and adjudicated with a written rationale. The analyst approves the application, attaches a decision note, and the platform stamps the decision with the analyst's identity and a timestamp. Every input — every document, every screening run, every email — is preserved as a permanent, regulator-ready audit record against the case.