For the person whose name is on the compliance line. A proposed agent that surfaces FAIS Key-Individual duties, PCI scope, AML/FIC obligations and SARB directives — explained and cited — so the accountable human reads faster. It drafts and points; it never substitutes for the sign-off.
This agent is part of a proposed roster of four role agents for payments executives and consulting teams. It is scoped here so partners can pressure-test the brief before any of it is built. The one agent that is live today is the research & regulatory-watch agent — the editorial engine that keeps the knowledge base these role agents would read from current and human-moderated.
A FAIS Key Individual, a compliance officer or a money-laundering reporting officer carries personal, non-delegable accountability. When a SARB directive lands or a PCI standard is revised, they have to know what it means for their licence and their obligations — quickly and correctly.
The compliance & KI advisor is a proposed agent that does the briefing half of that job. It surfaces the relevant duty — FAIS KI responsibilities, PCI DSS scope, AML/FIC requirements, the latest SARB directive — explains it in plain terms and cites the source. It is a research and triage tool for the accountable person. It is explicitly not the person, and it never signs anything.
The agent is a thin reasoning layer over a thick, moderated knowledge base. It does not free-associate; it retrieves expert-approved leaves, reasons over them, cites the source, and hands a draft to a human.
// Compliance advisor — brief the human, never replace them 1. ASK “What changes for my KI duties under the new SARB directive?” 2. RETRIEVE pull compliance/ + training/fais-key-individual, map the directive to the affected obligations 3. EXPLAIN plain-language summary with the primary source cited 4. FLAG mark what needs a human compliance decision — sign-off, filing, interpretation, attestation 5. HAND OFF the accountable person reads, decides and signs. The agent has no authority and keeps none.
Accountable for the rendering of financial services under a licence — who must stay current on their duties and the regulator’s posture.
Tracking AML/FIC obligations, suspicious-transaction reporting and the moving regulatory perimeter.
Mapping DSS scope and obligations across an estate that changes faster than the standard does.
What the Key Individual is personally responsible for, where the lines sit, and what a change in the regulator’s stance shifts.
Which obligations bite for a given flow or estate — framed against the compliance/pci-dss leaf, not invented.
Customer due diligence, record-keeping and reporting duties surfaced and explained with the FIC source cited.
New directives mapped to the obligations they touch, so the accountable person sees the impact, not just the PDF.
Every leaf in this surface — modes, rails, regions, compliance, training — is the agent’s primary context. It reasons over expert-approved content, not the open web.
Each leaf carries a named reviewer and a review date. The agent inherits that provenance: it can only stand on content a domain-expert partner has signed off.
Regulator directives, scheme bulletins and standards bodies are cited inline. The agent surfaces the source so the human can check it — never “trust me”.
The advisor reads the pay.2nth.ai compliance and training content — including the FAIS Key-Individual leaf — all of it human-moderated.
This is the leaf where the honesty matters most. Compliance accountability is personal and statutory. An AI can brief it; it cannot bear it. These lines are absolute:
A KI attestation, a compliance approval, a suspicious-transaction filing — these are acts of an accountable human. The agent can draft and brief; it cannot sign, and its draft is not an approval.
What a directive means for your licence is a judgement call carrying personal liability. The agent points at the text; a qualified human interprets it.
No regulatory filing, no board attestation, no suspicious-transaction report is submitted by the agent. It prepares; the accountable person decides and submits.
A regulator wants the cited primary source and a human judgement, not a chat transcript. Edge models summarise and can misread nuance — always check the source before acting.
Reach for it to stay current and to triage — “what landed this week, and which of my obligations does it touch?” It turns a stack of directives and standards into a prioritised, cited briefing so the accountable person spends their scarce attention on the decisions that actually need them.
Never let it stand in for the decision. The moment the task is a sign-off, an interpretation that carries liability, or a filing, the agent’s role ends and the accountable human’s begins. The cost of being wrong here is a sanction, a licence condition, or personal liability — which is exactly why a machine is never the last step.
External framing, unchanged: “scaffolding and testing with customers who are looking to get into production.” This is a proposed agent, not a compliance system of record.