AML-KYC Casework

Financial-services AML/KYC casework with a confidential-gated learnings domain — demonstrate the workspace-KB learnings exchange in a regulated financial context.

The AML-KYC Casework pack demonstrates the org-KB learnings exchange in a regulated financial-services context. Casework teams capture investigation notes in a workspace knowledgebase; on finalization, the platform condenses them into typed AmlLearning records — red flags, controls, and severity ratings — that land in a confidential-gated, immutable org domain requiring a two-reviewer quorum before materialization.

What the pack provisions

What	Detail
`AmlLearning` object type	Properties: `title` (string), `redFlag` (text), `control` (text), `severity` (string: `low` / `medium` / `high` / `critical`)
`aml_learnings` governed domain	`confidential` classification cap, `org:mlro` role required, immutable rows (no post-promotion edits), 2-reviewer approval quorum
"AML Case Review" process flow	Final stage `approval` action with `promoteToOrgKb: true`; `knowledge.promotionTarget` → `aml_learnings` / `AmlLearning`

Installing the pack provisions both the governed aml_learnings domain and the AML Case Review process flow. Starting a template instance auto-creates a workspace KB linked to aml_learnings. The full case-investigation-to-org-corpus journey works from the download.

Download

aml-kyc-casework.scrydon-pack.tar.gz — latest.
scrydon-pack-aml-kyc-casework-1.0.0.scrydon-pack.tar.gz — version-pinned.

Install

IT admin, once: open Settings → Platform → Packs, upload aml-kyc-casework.scrydon-pack.tar.gz, and click Upload. The AmlLearning object type and the org-scoped aml_learnings domain are provisioned immediately, and the AML Case Review process flow is made available org-wide.

Workspace admin: open Marketplace in the agentic sidebar and click Add on AML Case Review to opt this workspace in (per-workspace, not per-environment).

Under Process Flows, find AML Case Review and click Start. The platform auto-creates a workspace knowledgebase linked to aml_learnings / AmlLearning — no manual KB wiring needed.

Upload investigation notes, SAR drafts, or case summaries into the workspace KB.

Learnings flow

The finalize trigger is the standard org-KB pattern — either an approval action with promoteToOrgKb: true on process completion, or the Workspace-Admin Promote learnings button.

Finalize the workspace KB (approval or manual). The platform gathers all pages, DLP-scans them, runs a single LLM pass over the whole corpus, and produces a set of discrete AmlLearning records.

Each learning's severity, redFlag, control, and title are shaped to the AmlLearning property schema resolved from the ontology. Each record carries references back to the source case pages.

The learnings are DLP-scanned and enqueued via the org-KB mutation queue.

Open Settings → Organization → Knowledgebase → Review. The aml_learnings domain requires a 2-reviewer quorum — two members with the org:mlro role must independently approve before a learning materializes.

After the quorum approves, learnings materialize into aml_learnings as immutable rows. No post-promotion edits are possible.

Immutability is enforced at the domain level. Once an AmlLearning materializes, it cannot be edited. To correct a materialized learning, revoke it (which creates an audit entry) and promote the corrected version through the full quorum review.

Governance at a glance

Policy	Setting
Classification cap	`confidential`
Required role	`org:mlro` (Money Laundering Reporting Officer)
Review gate	2-reviewer approval quorum
Immutability	Rows cannot be edited post-materialization
Re-finalize behavior	Upsert by stable learning key — quorum triggered only for new or changed content

confidential means only MLRO-role members see these rows. Members without org:mlro cannot read aml_learnings content even via Analytics or the Cortex agent — the classification cap is enforced at retrieval, not just at display.

What's interesting about it

Two-reviewer quorum. No single person can approve a financial-crime learning into the org corpus — the quorum gate requires two independent org:mlro sign-offs.
Immutable provenance. Each AmlLearning's signed provenance record is permanent and cannot be altered after materialization, satisfying audit trail requirements.
confidential classification cap. Any properties that exceed confidential — for example, nomineeDetails in a SAR — are automatically masked for members below that clearance.
References link to case pages. Each learning's references field lists the exact workspace KB pages that contributed to it, so auditors can trace a control or red flag back to the underlying case material.

Promoting content — broker client reference and both finalize paths.
Markings & governed retrieval — how classification caps are enforced at retrieval.
FOIA Precedent example — government vertical with records retention and PII redaction.