RegLens
Portfolio prototype · source-bounded
AI-POWERED AML/CFT CONTENT OPERATIONS

AI-powered regulatory change,
mapped to learning impact.

A human-in-the-loop AI workflow that compares an official AUSTRAC source with existing training, checks every proposed change, and publishes a tracked SCORM update.

Try the live comparison Explore the workflow
No open-web generation. Claims are constrained to one 4-page source.
THE WORKFLOW

One source. Four controlled steps.

Designed for a same-day content update
INTERACTIVE AI COMPARISON LAB

See how the comparison works.

Edit the current-course claims, then run the same evidence and qualifier checks used to build the report.

Operable portfolio demo
01

ExtractAI structures source and course claims

→
02

CompareClaims are matched to bounded evidence

→
03

EvaluateRules preserve qualifiers and block risk

INPUT A · LOCKED

Official AUSTRAC evidence

4 pages
p.1 · Enrolment

Enrol within 28 days of providing a designated service.

p.2 · Governance

Senior management approves; the governing body oversees; a compliance officer handles day-to-day compliance.

pp.2–3 · CDD

Initial CDD occurs before service in most cases; ongoing CDD continues through the relationship.

p.4 · Records

For most obligations, retain records for seven years.

INPUT B · EDITABLE

Current course claims

Enter one claim per line. Try changing “five” to “seven”, or add “for most obligations”.

◎

Results will appear here
Each claim receives a status, source location and reviewer explanation.

What is AI here?

In production, an approved language model would extract and normalise claims. This public demo emulates that structured AI output locally, then runs the real deterministic evidence gate—so it is interactive, reproducible and sends no course data anywhere.

01 / BOUNDED SOURCE
Verified authority View source ↗
SOURCE BOUNDARY

What the workflow is allowed to know

Only claims directly supported by this factsheet can pass evaluation. Anything else is labelled for human review.

01EnrolmentTiming & updates
02AML/CTF programRisk & governance
03Customer due diligenceInitial & ongoing
04Ongoing dutiesReports & records

Built-in limitation
The source says it is a summary and does not cover all obligations or scenarios.

02 / COURSE IMPACT REPORT

Review queue for “Tranche 2 Essentials”

Five content items mapped. Priorities reflect learner risk and source divergence.

✓ Analysis complete
Evidence map ready
5items mapped
3high priority
1claim blocked
4source sections
PriorityCourse itemDetected issueRecommended actionEvidence

Select any row to inspect the supporting evidence.

03 / EVALUATION & QA

Claim-level grounding check

The evaluator compares each draft claim with an exact source excerpt before release.

92/ 100
✓

Start date and covered sectorsSupported · page 1

PASS
✓

Enrolment within 28 daysSupported · page 1

PASS
✓

Governance responsibilitiesSupported · page 2

PASS
✓

Initial and ongoing CDD distinctionSupported · pages 2–3

PASS
!

Annual independent evaluationContradicted by bounded source · page 2

BLOCK
04 / UPDATED LEARNING
Ready for human approval

A tiny, tracked module—not an AI content dump.

The approved changes become a focused five-minute update with one scenario and three scored questions.

  • ✓ Source references inside the learning
  • ✓ SCORM 1.2 completion and score tracking
  • ✓ Graceful browser-preview fallback
Preview module ↗ Download SCORM 1.2
Tranche 2 Essentials · Update
LESSON 02 OF 03

Who owns the AML/CTF program?

Clear governance turns a policy document into day-to-day protection.

01

Senior management
Approves the program

02

Governing body
Oversees the program

03

Compliance officer
Oversees day-to-day compliance

DESIGN DECISION
“AI proposes. Evidence constrains. A human approves.”

This MVP deliberately prioritises a transparent, repeatable content workflow over technical complexity.

SOURCE EVIDENCE

Evidence title