Chapter 21: Banking-Specific AI — Lesson Plan
Generated by: chapter-planner v2.0.0 (Reasoning-Activated) Source Spec: Docx 7-Part structure + Issue #817 (plugin architecture) Created: 2026-03-06 Constitution: v6.0.0 (Reasoning Mode)
I. Chapter Analysis
Chapter Type
Technical/Hybrid — The chapter teaches regulatory banking concepts (conceptual) through a plugin architecture (technical) with 14 hands-on exercises (applied). Lessons 1-2 are conceptual/architectural, Lessons 3-10 are technical with worked examples, Lesson 11 is cross-pillar integration, Lessons 12-13 are exercise bundles, Lesson 14 is reconciliation (technical), and Lesson 15 is a capstone build.
Concept Density Analysis
Core Concepts (from docx 7-Part structure): 21 core concepts across 3 regulatory pillars + reconciliation
IFRS 9 Pillar (7 concepts):
- IFRS 9 staging (Stage 1/2/3 — significant increase in credit risk)
- ECL formula (PD x LGD x EAD)
- Probability of Default (PD) — point-in-time vs through-the-cycle
- Loss Given Default (LGD) — downturn LGD, cure rates
- Exposure at Default (EAD) — CCFs, undrawn commitments
- Macroeconomic scenario overlays (probability-weighted base/upside/downside)
- Post-Model Adjustments (PMAs) — management overlays, model limitations
Basel III/IV Pillar (7 concepts): 8. Capital stack (CET1, AT1, Tier 2) and deductions 9. Capital adequacy ratios (CET1 ratio, Tier 1 ratio, Total Capital ratio) 10. Risk-Weighted Assets (RWA) — Standardised Approach 11. Internal Ratings-Based (IRB) approach and the output floor (72.5%) 12. Leverage ratio (Tier 1 / total exposure) 13. Liquidity Coverage Ratio (LCR) — HQLA / net cash outflows 14. Net Stable Funding Ratio (NSFR)
AML/KYC Pillar (4 concepts): 15. Customer Due Diligence (CDD) and Enhanced Due Diligence (EDD) 16. Three lines of defence model 17. Transaction monitoring — rules-based to ML evolution 18. SAR filing workflow and tipping-off prohibition
Cross-Pillar + Architecture (3 concepts): 19. Pillar interaction (ECL spike -> capital impact; fraud -> stage migration -> provision -> capital) 20. Pillar-aware routing architecture (router + 16 product skills) 21. Bank reconciliation (nostro, suspense, GL-to-risk, four-way provision recon)
Complexity Assessment: Complex — 21 interconnected concepts across 3 regulatory frameworks with quantitative models (IRB formula, ECL calculation, LCR computation). Each concept has sub-components. Concepts interact across pillars (a Stage 2 migration affects both IFRS 9 provisioning AND Basel III CET1 ratio).
Proficiency Tier: B1-B2 (from Part 3 positioning — Business Domain Agent Workflows). Banking practitioners are the audience; they have domain familiarity but need to learn the AI agent application layer. The quantitative modelling concepts (IRB, ECL) push toward B2 for some lessons.
Justified Lesson Count: 15 lessons
- Lessons 1-2 (Conceptual Foundation): 2 lessons — three-pillar landscape + plugin architecture
- Lessons 3-5 (IFRS 9 Deep Dive): 3 lessons — staging/ECL, PD/LGD/EAD, macro/PMA
- Lessons 6-8 (Basel III/IV Deep Dive): 3 lessons — capital ratios, RWA/IRB, leverage/liquidity
- Lessons 9-10 (AML/KYC): 2 lessons — three lines + TM/SAR
- Lesson 11 (Cross-Pillar Integration): 1 lesson — pillar interactions with exercises
- Lessons 12-13 (Exercise Bundles): 2 lessons — deep practice (4 exercises across 2 lessons)
- Lesson 14 (Reconciliation): 1 lesson — four reconciliation types
- Lesson 15 (Capstone): 1 lesson — full skill library build
Total: 15 lessons (NOT arbitrary template — driven by 21 concepts across 3 pillars + 14 exercises + reconciliation + capstone)
Reasoning: Ch20 had 18 lessons for 13 skills across 8 product families x 20 jurisdictions. Ch21 has 17 skills across 3 pillars with fewer jurisdictional variants but deeper quantitative modelling (ECL, IRB, LCR). 15 lessons is appropriate because: (a) 3 pillars each need 2-3 lessons for depth, (b) 14 exercises need at least some bundling (2-3 per lesson), (c) reconciliation is a distinct topic deserving its own lesson, (d) the capstone requires a full lesson.
II. Success Evals
Predefined Success Criteria (derived from docx learning outcomes):
- Student can explain the three regulatory pillars and describe how a single loan portfolio is simultaneously governed by all three
- Student can install the banking plugin and trace a query through the pillar-aware routing system
- Student can calculate a 12-month ECL and a lifetime ECL using PD x LGD x EAD with proper staging
- Student can compute probability-weighted ECL across base/upside/downside scenarios and explain when PMAs are needed
- Student can calculate CET1, Tier 1, and Total Capital ratios and explain the deductions
- Student can compute RWA under both SA and IRB approaches and apply the 72.5% output floor
- Student can calculate LCR from HQLA and net cash outflows and explain NSFR
- Student can describe the three lines of defence, explain CDD/EDD, and identify PEP screening requirements
- Student can describe the evolution from rules-based to ML transaction monitoring, explain SAR filing, and identify tipping-off risks
- Student can trace a cross-pillar scenario (e.g., fraud triggers stage migration triggers provision increase triggers CET1 impact)
- Student can perform nostro reconciliation, suspense clearing, GL-to-risk-system reconciliation, and four-way provision reconciliation
- Student can deploy the full 17-skill banking plugin and execute a cross-pillar capstone scenario
All lessons below map to these evals.
III. Lesson Sequence
Lesson 1: The Three Regulatory Pillars of Modern Banking
File: 01-the-three-pillars.md
Sidebar Position: 1
Duration: 20 minutes
Summary: Introduces the three regulatory pillars — Solvency (Basel III/IV), Accounting (IFRS 9), and Financial Crime (AML/KYC) — that simultaneously govern every banking institution. Uses a single loan portfolio as the running example to show how the same asset requires treatment under all three pillars. Establishes why single-pillar agents fail and previews the pillar-aware architecture.
Learning Objectives:
- Explain the three regulatory pillars of modern banking and why they must be addressed simultaneously (Bloom: Understand, CEFR: A2)
- Identify which regulatory pillar governs a given banking scenario (Bloom: Apply, CEFR: A2)
- Articulate why a single-pillar AI agent produces incomplete or misleading output (Bloom: Understand, CEFR: A2)
Stage: Layer 2 (Collaboration) — Students have the plugin installed, AI works alongside them
CEFR Proficiency: A2 (introductory/conceptual)
New Concepts (count: 5 -- within A2 limit of 7):
- The three regulatory pillars (solvency, accounting, financial crime)
- IFRS 9 as the accounting pillar (provisioning for expected losses)
- Basel III/IV as the solvency pillar (capital adequacy against unexpected losses)
- AML/KYC as the financial crime pillar (detecting and reporting illicit activity)
- Pillar interaction — how the same portfolio requires simultaneous treatment under all three
Cognitive Load Validation: 5 concepts at A2 level -- WITHIN LIMIT
Maps to Evals: 1
Bloom's Level: Understand
Plugin Skills Used: banking-router (conceptual overview of routing logic)
Key Concepts Introduced: Three pillars framework, expected vs unexpected loss distinction, pillar interaction concept
Try With AI Prompt Themes:
- "Explain why a $500M corporate loan portfolio needs treatment under all three pillars simultaneously" — pillar identification
- "Compare what IFRS 9, Basel III, and AML each 'see' when they look at the same loan" — perspective comparison
- "A bank discovers fraud in a corporate loan. Trace the impact across all three pillars" — cross-pillar cascade preview
Lesson 2: The Banking Plugin Architecture — 17 Skills, Three Pillars
File: 02-plugin-architecture.md
Sidebar Position: 2
Duration: 25 minutes
Summary: Walks through the banking plugin's 17-skill architecture — 1 router + 16 product skills spanning IFRS 9, Basel, AML, and reconciliation. Students install the plugin and trace a sample query through the pillar-aware routing system. Explains how the router detects which pillar(s) a query requires and chains the appropriate skills.
Learning Objectives:
- Install the banking plugin and verify activation (Bloom: Apply, CEFR: A2)
- Trace a query through the pillar-aware routing architecture, identifying which skills load and why (Bloom: Apply, CEFR: B1)
- Explain the separation of concerns between the router, pillar skills, and cross-pillar orchestration (Bloom: Understand, CEFR: B1)
Stage: Layer 2 (Collaboration) — Plugin installed, architecture walkthrough with AI
CEFR Proficiency: A2-B1 (architectural understanding)
New Concepts (count: 6 -- within B1 limit of 10):
- The 17-skill architecture (1 router + 16 product skills)
- Pillar-aware routing (router detects IFRS 9, Basel, AML, or cross-pillar)
- Skill chaining for cross-pillar queries
- The 4 domain commands (
/bank-ecl,/bank-capital,/bank-recon,/bank-aml) - SessionStart hook (pillar detection) and PostToolUse hook (regulatory label validation)
- The banking skill library file structure
Cognitive Load Validation: 6 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 2
Bloom's Level: Apply
Plugin Skills Used: banking-router, all 4 commands demonstrated
Key Concepts Introduced: Pillar-aware routing, skill chaining, plugin file structure
Try With AI Prompt Themes:
- "I have a query about ECL staging for a UK mortgage portfolio. Walk me through which skills the router would load and why" — single-pillar routing trace
- "A regulator asks: 'What is the capital impact of your IFRS 9 provision increase?' Which skills does the router chain?" — cross-pillar routing trace
- "Show me the complete skill library listing — group the 16 product skills by pillar" — skill inventory
Lesson 3: IFRS 9 Expected Credit Loss — Staging and the ECL Formula
File: 03-ifrs9-staging-ecl.md
Sidebar Position: 3
Duration: 40 minutes
Summary: Deep dive into IFRS 9 ECL — the three-stage model (performing, underperforming, non-performing), significant increase in credit risk (SICR) triggers, and the fundamental ECL formula (PD x LGD x EAD). Distinguishes 12-month ECL (Stage 1) from lifetime ECL (Stages 2 and 3). Uses a retail mortgage example to make staging tangible.
Learning Objectives:
- Classify financial assets into Stage 1, 2, or 3 using SICR criteria (Bloom: Apply, CEFR: B1)
- Calculate 12-month ECL vs lifetime ECL using the PD x LGD x EAD formula (Bloom: Apply, CEFR: B1)
- Explain the economic rationale for the three-stage model and why it replaced IAS 39's incurred loss approach (Bloom: Understand, CEFR: B1)
Stage: Layer 2 (Collaboration) — AI demonstrates ECL calculation, student validates
CEFR Proficiency: B1
New Concepts (count: 7 -- within B1 limit of 10):
- Stage 1 — 12-month ECL (performing, no SICR since origination)
- Stage 2 — Lifetime ECL (significant increase in credit risk, not yet defaulted)
- Stage 3 — Lifetime ECL (credit-impaired, defaulted)
- Significant Increase in Credit Risk (SICR) — quantitative and qualitative triggers
- ECL = PD x LGD x EAD (the formula)
- 12-month ECL vs Lifetime ECL (the measurement difference)
- Stage migration — what triggers movement between stages
Cognitive Load Validation: 7 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 3
Bloom's Level: Apply
Plugin Skills Used: ifrs9-staging, ifrs9-ecl-calculator
Key Concepts Introduced: Three-stage model, SICR, ECL formula, 12-month vs lifetime ECL
Try With AI Prompt Themes:
- "A borrower's PD has increased from 0.5% at origination to 2.8% today. Should this asset move from Stage 1 to Stage 2? Show your SICR analysis" — staging classification
- "Calculate the 12-month ECL for a Stage 1 retail mortgage: PD=1.2%, LGD=15%, EAD=$450,000. Then calculate what the lifetime ECL would be if it migrated to Stage 2 with a remaining term of 20 years" — ECL calculation
- "Explain to a non-technical board member why IFRS 9's forward-looking ECL model is better than IAS 39's 'wait until it breaks' approach. Use the 2008 financial crisis as your example" — conceptual understanding
Lesson 4: PD, LGD, and EAD — Building the ECL Components
File: 04-ifrs9-pd-lgd-ead.md
Sidebar Position: 4
Duration: 45 minutes
Summary: Builds each ECL component in depth. PD: point-in-time calibration, term structures, forward-looking adjustment. LGD: downturn LGD, cure rates, collateral haircuts, workout recoveries. EAD: drawn balances, undrawn commitments, credit conversion factors (CCFs). Includes Exercise 1: calculate the full ECL for a retail mortgage portfolio segment.
Learning Objectives:
- Build a PD term structure and apply forward-looking adjustment (Bloom: Apply, CEFR: B1)
- Calculate downturn LGD incorporating collateral haircuts and cure rates (Bloom: Apply, CEFR: B1)
- Compute EAD with credit conversion factors for undrawn commitments (Bloom: Apply, CEFR: B1)
Stage: Layer 2 (Collaboration) — AI guides through calculation, student drives parameter choices
CEFR Proficiency: B1
New Concepts (count: 8 -- within B1 limit of 10):
- Point-in-time PD vs through-the-cycle PD
- PD term structure (marginal and cumulative PD curves)
- Forward-looking PD adjustment
- Downturn LGD (stressed recovery assumptions)
- Cure rates and their impact on LGD
- Collateral haircuts (property, equipment, financial)
- EAD for drawn and undrawn exposures
- Credit Conversion Factors (CCFs) for off-balance-sheet commitments
Cognitive Load Validation: 8 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 3, 4
Bloom's Level: Apply
Plugin Skills Used: ifrs9-pd-model, ifrs9-lgd-model, ifrs9-ead-calculator
Key Concepts Introduced: PD term structures, downturn LGD, CCFs, full ECL build
Exercise: Exercise 1 — Retail Mortgage ECL Calculation
- Scenario: $2B residential mortgage portfolio, 3 segments by LTV band
- Task: Calculate ECL for each segment using provided PD term structures, LGD assumptions, and EAD with undrawn redraws
- Duration: 20 min embedded in lesson
Try With AI Prompt Themes:
- "Build a PD term structure for a BB-rated corporate with annual marginal PDs of: Year 1: 0.8%, Year 2: 1.1%, Year 3: 1.4%, Year 4: 1.6%, Year 5: 1.7%. Show cumulative and marginal curves" — PD construction
- "A mortgage portfolio has a through-the-cycle LGD of 12%. Apply a downturn adjustment assuming property values decline 20% and cure rates fall from 30% to 15%" — LGD stress
- "Calculate EAD for a revolving credit facility: drawn balance $5M, undrawn limit $15M, CCF 40%. Then recalculate if CCF increases to 75% under stress" — EAD computation
Lesson 5: Macroeconomic Scenarios and Post-Model Adjustments
File: 05-ifrs9-macro-pma.md
Sidebar Position: 5
Duration: 40 minutes
Summary: Completes the IFRS 9 picture with macroeconomic scenario design (base, upside, downside) and probability-weighted ECL calculation. Introduces Post-Model Adjustments (PMAs) — management overlays that capture model limitations, emerging risks, and sector-specific concerns not reflected in model outputs. Includes Exercise 2: apply macroeconomic scenarios and a PMA to a commercial real estate portfolio.
Learning Objectives:
- Design base, upside, and downside macroeconomic scenarios with probability weights (Bloom: Apply, CEFR: B1)
- Calculate probability-weighted ECL across scenarios (Bloom: Apply, CEFR: B1)
- Explain when and why PMAs are necessary, and describe their governance requirements (Bloom: Understand, CEFR: B1)
Stage: Layer 2 (Collaboration) — AI generates scenario ECLs, student assigns probabilities and designs PMA
CEFR Proficiency: B1
New Concepts (count: 6 -- within B1 limit of 10):
- Macroeconomic scenario design (base, upside, downside, severe downside)
- Scenario probability weights (non-linearity — expected loss is NOT the loss under the expected scenario)
- Probability-weighted ECL calculation
- Post-Model Adjustments (PMAs) — management overlays
- PMA governance (documentation, approval, time-limiting, back-testing)
- Sector-specific overlays (e.g., commercial real estate, energy, hospitality)
Cognitive Load Validation: 6 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 4
Bloom's Level: Apply
Plugin Skills Used: ifrs9-macro-scenarios, ifrs9-pma
Key Concepts Introduced: Scenario design, probability-weighted ECL, PMAs, non-linearity of expected loss
Exercise: Exercise 2 — Commercial Portfolio Macroeconomic Overlay
- Scenario: $800M commercial real estate portfolio, 3 scenarios (base 50%, upside 20%, downside 30%)
- Task: Calculate ECL under each scenario, apply probability weights, design a $12M PMA for construction sector exposure
- Duration: 15 min embedded in lesson
Try With AI Prompt Themes:
- "A bank's ECL under the base case is $45M, upside is $28M, and downside is $78M. Weights are 50/20/30. Calculate the probability-weighted ECL and explain why it is NOT $45M (the base case ECL)" — non-linearity demonstration
- "Design a PMA for a bank with $500M in hospitality sector loans during a pandemic. What model limitation does the PMA address? What governance documentation is required?" — PMA design
- "Compare two banks: Bank A uses 3 scenarios, Bank B uses 5. Bank A's ECL is lower. Does that mean Bank A is less conservative? What other factors matter?" — scenario design analysis
Lesson 6: Basel III/IV Capital Adequacy — CET1, Tier 1, Total Capital
File: 06-basel-capital-ratios.md
Sidebar Position: 6
Duration: 45 minutes
Summary: Introduces the Basel III/IV capital framework — the capital stack (CET1, AT1, Tier 2), regulatory deductions, and the three capital ratios. Walks through a complete CET1 ratio calculation from share capital and retained earnings through deductions to the final percentage. Includes Exercise 3: calculate capital ratios for a mid-size bank.
Learning Objectives:
- Identify and classify capital instruments into CET1, AT1, and Tier 2 (Bloom: Apply, CEFR: B1)
- Apply regulatory deductions (goodwill, DTA, minority interests) to arrive at regulatory capital (Bloom: Apply, CEFR: B1)
- Calculate CET1 ratio, Tier 1 ratio, and Total Capital ratio and compare to minimum requirements (Bloom: Apply, CEFR: B1)
Stage: Layer 2 (Collaboration) — AI guides through capital stack build, student validates deductions
CEFR Proficiency: B1
New Concepts (count: 7 -- within B1 limit of 10):
- CET1 capital (share capital, share premium, retained earnings, AOCI)
- Additional Tier 1 (AT1) — perpetual non-cumulative instruments, CoCos
- Tier 2 capital — subordinated debt, general provisions
- Regulatory deductions (goodwill, intangible assets, DTAs, significant investments)
- CET1 ratio = CET1 capital / RWA (minimum 4.5% + buffers)
- Tier 1 ratio = (CET1 + AT1) / RWA (minimum 6%)
- Total Capital ratio = (CET1 + AT1 + T2) / RWA (minimum 8%)
Cognitive Load Validation: 7 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 5
Bloom's Level: Apply
Plugin Skills Used: basel-capital-stack, basel-capital-ratios
Key Concepts Introduced: Capital stack, deductions, three capital ratios, minimum requirements + buffers
Exercise: Exercise 3 — Bank Capital Ratio Calculation
- Scenario: Mid-size bank with $5B in assets, detailed capital composition provided
- Task: Classify instruments, apply deductions, calculate all three capital ratios, compare to minimums + CCB + D-SIB
- Duration: 20 min embedded in lesson
Try With AI Prompt Themes:
- "A bank has: share capital $800M, retained earnings $1.2B, goodwill $200M, DTAs $150M, AT1 bonds $400M, sub-debt $300M. Calculate CET1, Tier 1, and Total Capital after deductions" — capital calculation
- "A bank's CET1 ratio is 10.5%. The minimum is 4.5%, CCB is 2.5%, D-SIB buffer is 1.5%. How much CET1 is 'free' above requirements? What happens if it falls below the combined buffer?" — buffer analysis
- "Explain to a board member why goodwill is deducted from CET1. What would happen if we didn't deduct it?" — deduction rationale
Lesson 7: Risk-Weighted Assets — SA and IRB Approaches
File: 07-basel-rwa-risk-weights.md
Sidebar Position: 7
Duration: 50 minutes
Summary: Deep dive into the denominator of the capital ratio — risk-weighted assets. Covers the Standardised Approach (SA) with prescribed risk weights by asset class, and the Internal Ratings-Based (IRB) approach with the Basel risk-weight function. Introduces the Basel IV output floor (72.5%) that constrains IRB banks. Includes Exercise 4: compare RWA under SA and IRB for the same portfolio.
Learning Objectives:
- Apply SA risk weights to a loan portfolio across asset classes (Bloom: Apply, CEFR: B1)
- Explain the IRB risk-weight function and its key inputs (PD, LGD, maturity, correlation) (Bloom: Understand, CEFR: B2)
- Calculate the output floor impact on an IRB bank's RWA and capital ratio (Bloom: Apply, CEFR: B1)
Stage: Layer 2 (Collaboration) — AI computes RWA, student compares approaches and floor impact
CEFR Proficiency: B1-B2
New Concepts (count: 7 -- within B1 limit of 10):
- Standardised Approach (SA) — prescribed risk weights by asset class
- Key SA risk weights (sovereigns 0%, banks 20-150%, corporates 20-150%, retail 75%, residential mortgage 20-70%)
- Internal Ratings-Based (IRB) — model-derived risk weights
- The IRB risk-weight function (correlation, maturity adjustment, confidence level 99.9%)
- Foundation IRB vs Advanced IRB (who estimates LGD and EAD)
- The Basel IV output floor — IRB RWA ≥ 72.5% of SA RWA
- RWA for off-balance-sheet items (CCFs under SA)
Cognitive Load Validation: 7 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 6
Bloom's Level: Apply/Understand
Plugin Skills Used: basel-rwa-sa, basel-rwa-irb, basel-output-floor
Key Concepts Introduced: SA risk weights, IRB function, output floor, F-IRB vs A-IRB
Exercise: Exercise 4 — RWA Comparison: SA vs IRB
- Scenario: Bank with $20B loan portfolio across sovereigns, corporates, retail, and mortgages
- Task: Calculate RWA under SA, estimate RWA under IRB (provided model outputs), apply 72.5% output floor, determine which approach produces higher RWA
- Duration: 25 min embedded in lesson
Try With AI Prompt Themes:
- "Calculate RWA under the Standardised Approach for: sovereigns $3B (0% risk weight), corporates $8B (100%), retail $4B (75%), residential mortgages $5B (35% for LTV<80%, 50% for LTV 80-90%)" — SA calculation
- "An IRB bank's model-derived RWA is $12B. The same portfolio under SA would produce $16B RWA. The output floor is 72.5%. What is the bank's final RWA? Does the floor bite?" — output floor
- "Explain why the Basel Committee introduced the output floor. What problem was it solving? Use a real-world analogy" — regulatory rationale
Lesson 8: Leverage Ratio, LCR, and NSFR
File: 08-basel-leverage-liquidity.md
Sidebar Position: 8
Duration: 35 minutes
Summary: Covers the non-risk-weighted solvency measure (leverage ratio) and the two liquidity ratios (LCR and NSFR). Builds an LCR calculation from HQLA classification through net cash outflow computation. Introduces NSFR as the structural counterpart to LCR. Includes Exercise 5: liquidity stress test.
Learning Objectives:
- Calculate the leverage ratio and explain why a non-risk-weighted measure is needed alongside RWA-based ratios (Bloom: Apply/Understand, CEFR: B1)
- Classify assets as HQLA Level 1, 2A, or 2B and calculate the LCR (Bloom: Apply, CEFR: B1)
- Explain the relationship between LCR (short-term, 30-day) and NSFR (long-term, 1-year) (Bloom: Understand, CEFR: B1)
Stage: Layer 2 (Collaboration) — AI builds liquidity calculations, student validates classifications
CEFR Proficiency: B1
New Concepts (count: 6 -- within B1 limit of 10):
- Leverage ratio = Tier 1 capital / total exposure measure (minimum 3%)
- Total exposure measure (on-balance-sheet + derivatives + SFTs + off-balance-sheet)
- High Quality Liquid Assets (HQLA) — Level 1 (cash, sovereign bonds), Level 2A (covered bonds, corp bonds), Level 2B (equities, RMBS)
- LCR = HQLA / net cash outflows over 30 days (minimum 100%)
- Net cash outflows = outflows - min(inflows, 75% of outflows)
- NSFR = Available Stable Funding / Required Stable Funding (minimum 100%)
Cognitive Load Validation: 6 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 7
Bloom's Level: Apply
Plugin Skills Used: basel-leverage-ratio, basel-lcr, basel-nsfr
Key Concepts Introduced: Leverage ratio, HQLA, LCR, NSFR, total exposure measure
Exercise: Exercise 5 — Liquidity Stress Test
- Scenario: Bank facing a 30-day stress scenario with deposit outflows, credit line drawdowns, and asset liquidation options
- Task: Classify HQLA, calculate net cash outflows, determine LCR, identify shortfall if any
- Duration: 15 min embedded in lesson
Try With AI Prompt Themes:
- "A bank has: cash $2B, government bonds $5B, corporate bonds (AA-) $1.5B, equities $0.5B. Classify each as HQLA Level 1, 2A, or 2B and calculate total HQLA after haircuts" — HQLA classification
- "Calculate LCR: HQLA $8B, retail deposit outflows $3B (5% run-off), wholesale funding outflows $4B (25% run-off), credit facility drawdowns $2B (10% draw), inflows from maturing loans $1.5B" — LCR calculation
- "Explain why a bank can have a strong CET1 ratio (14%) but still fail on LCR. Use the 2007-08 Northern Rock example" — capital vs liquidity distinction
Lesson 9: AML/KYC — The Three Lines of Defence
File: 09-aml-three-lines.md
Sidebar Position: 9
Duration: 35 minutes
Summary: Introduces the AML/KYC regulatory framework through the three lines of defence model. Covers Customer Due Diligence (CDD), Enhanced Due Diligence (EDD), Politically Exposed Persons (PEP) screening, and beneficial ownership identification. Establishes the boundary between what AI can automate (data gathering, screening, risk scoring) and what requires human judgment (risk acceptance, EDD decisions). Includes Exercise 6: customer onboarding risk assessment.
Learning Objectives:
- Apply the three lines of defence model to a bank's AML framework (Bloom: Apply, CEFR: B1)
- Classify customers as standard CDD or EDD using risk indicators (Bloom: Apply, CEFR: B1)
- Explain the agent's role boundary in AML — what it automates vs what requires human judgment (Bloom: Understand, CEFR: B1)
Stage: Layer 2 (Collaboration) — AI performs CDD data gathering, student makes risk classification decisions
CEFR Proficiency: B1
New Concepts (count: 6 -- within B1 limit of 10):
- Three lines of defence (1st: business/operations, 2nd: compliance/risk, 3rd: internal audit)
- Customer Due Diligence (CDD) — ID verification, source of funds, purpose of relationship
- Enhanced Due Diligence (EDD) — higher-risk customers, additional information requirements
- Politically Exposed Persons (PEP) — definition, screening requirements, family/close associates
- Beneficial ownership — identifying ultimate controllers, threshold (typically 25%)
- Agent boundary: automates screening, data gathering, risk scoring; humans make acceptance/rejection decisions
Cognitive Load Validation: 6 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 8
Bloom's Level: Apply
Plugin Skills Used: aml-cdd-screening, aml-pep-check, aml-risk-scoring
Key Concepts Introduced: Three lines of defence, CDD/EDD, PEP, beneficial ownership, agent boundary
Exercise: Exercise 6 — Customer Onboarding Risk Assessment
- Scenario: 4 customer profiles with varying risk indicators (corporate from high-risk jurisdiction, domestic retail, PEP-connected, complex beneficial ownership)
- Task: Classify each as CDD or EDD, identify risk factors, determine which the agent can handle autonomously and which require human decision
- Duration: 15 min embedded in lesson
Try With AI Prompt Themes:
- "A corporate customer is incorporated in the Cayman Islands, with beneficial owners in 3 countries, and the source of funds is described as 'international trade.' Apply the CDD/EDD framework and determine the risk classification" — risk classification
- "Explain the three lines of defence using a bank's AML department as the example. Who is in each line? What does each line do? What happens when the 2nd line finds a 1st line failure?" — three lines model
- "Where is the boundary between what the AI agent can do and what requires a human in AML? List 5 things the agent automates and 5 things that need human judgment" — agent boundary
Lesson 10: Transaction Monitoring, ML Evolution, and SAR Filing
File: 10-aml-tm-ml-sar.md
Sidebar Position: 10
Duration: 40 minutes
Summary: Covers the evolution of transaction monitoring (TM) from rules-based systems to machine learning approaches. Explains the SAR (Suspicious Activity Report) filing workflow — from alert generation through investigation to filing decision. Introduces the critical tipping-off prohibition. Includes Exercise 7: TM alert investigation.
Learning Objectives:
- Compare rules-based and ML-based transaction monitoring approaches (Bloom: Analyze, CEFR: B1)
- Describe the end-to-end SAR filing workflow from alert to submission (Bloom: Understand, CEFR: B1)
- Identify tipping-off risks and explain why AI agents must never disclose SAR-related information to customers (Bloom: Understand, CEFR: B1)
Stage: Layer 2 (Collaboration) — AI generates TM analysis, student evaluates alert quality and SAR thresholds
CEFR Proficiency: B1
New Concepts (count: 6 -- within B1 limit of 10):
- Rules-based TM (threshold rules, pattern rules, velocity rules)
- ML-based TM (anomaly detection, network analysis, false positive reduction)
- Alert investigation workflow (triage, investigation, escalation, disposition)
- SAR filing — legal obligation, content requirements, filing timelines
- Tipping-off prohibition — criminal offence, applies to ALL staff and systems
- Agent boundary in TM: generates analysis and flags, NEVER files SARs or communicates with subjects about suspicion
Cognitive Load Validation: 6 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 9
Bloom's Level: Analyze
Plugin Skills Used: aml-tm-rules, aml-tm-ml, aml-sar-workflow
Key Concepts Introduced: Rules vs ML TM, SAR workflow, tipping-off, agent TM boundary
Exercise: Exercise 7 — TM Alert Investigation
- Scenario: 3 TM alerts with supporting transaction data — one genuine suspicious activity, one false positive, one requiring additional investigation
- Task: Investigate each alert, determine disposition, draft investigation narrative for the genuine case, explain why the agent cannot file the SAR itself
- Duration: 20 min embedded in lesson
Try With AI Prompt Themes:
- "A customer's account shows: 12 cash deposits of $9,500 each over 3 weeks (just below the $10,000 reporting threshold). Analyse this using TM rules and explain what pattern this matches" — structuring detection
- "Compare rules-based TM vs ML-based TM. Rules generate 95% false positives. ML reduces false positives to 60%. But regulators are sceptical of ML. Why? What governance is needed?" — TM evolution
- "A bank employee discovers a SAR has been filed on a customer. The customer calls asking why their account is restricted. What MUST the employee NOT say? What CAN they say? Why is this a criminal offence?" — tipping-off
Lesson 11: Cross-Pillar Integration — When IFRS 9, Basel, and AML Collide
File: 11-pillar-integration.md
Sidebar Position: 11
Duration: 45 minutes
Summary: The pivotal lesson connecting all three pillars. Demonstrates how events cascade across regulatory frameworks — a fraud discovery triggers AML investigation, which triggers IFRS 9 stage migration, which triggers provision increase, which reduces CET1 capital. Includes two cross-pillar exercises: Exercise 8 (fraud-triggered stage migration) and Exercise 9 (capital impact of ECL spike).
Learning Objectives:
- Trace a cross-pillar cascade from AML trigger through IFRS 9 staging to Basel capital impact (Bloom: Analyze, CEFR: B2)
- Calculate the CET1 impact of an IFRS 9 provision increase (Bloom: Apply, CEFR: B1)
- Explain why pillar-isolated agents miss cross-pillar impacts and produce incomplete advice (Bloom: Evaluate, CEFR: B2)
Stage: Layer 2 (Collaboration) — AI chains skills across pillars, student validates cascade logic
CEFR Proficiency: B1-B2 (integration requires higher-order thinking)
New Concepts (count: 5 -- within B1 limit of 10):
- Cross-pillar cascade: AML event -> IFRS 9 stage migration -> Basel capital impact
- Provision increase -> CET1 reduction (ECL charge reduces retained earnings, which is CET1)
- Capital restoration actions (reduce RWA, raise capital, reduce dividends)
- Regulatory reporting cascade (SAR filing + IFRS 9 restatement + capital adequacy return update)
- Skill chaining in the router — how the agent detects and executes cross-pillar queries
Cognitive Load Validation: 5 concepts at B2 level -- WITHIN LIMIT (B2 allows 7-10)
Maps to Evals: 10
Bloom's Level: Analyze/Evaluate
Plugin Skills Used: banking-router (cross-pillar mode), ifrs9-staging, basel-capital-ratios, aml-sar-workflow
Key Concepts Introduced: Cross-pillar cascade, provision-capital linkage, skill chaining
Exercise: Exercise 8 — Fraud-Triggered Stage Migration
- Scenario: $50M corporate loan, fraud discovered, AML investigation, stage migration from 1 to 3, ECL recalculation
- Task: Trace the full cascade — AML trigger, stage migration justification, ECL recalculation (12-month to lifetime with impaired LGD), CET1 impact
- Duration: 20 min embedded in lesson
Exercise: Exercise 9 — Capital Impact of ECL Spike
- Scenario: Macroeconomic shock causes mass Stage 1->2 migration, $200M additional ECL charge
- Task: Calculate CET1 impact, determine if bank breaches minimum + buffer, identify capital restoration options
- Duration: 15 min embedded in lesson
Try With AI Prompt Themes:
- "Trace this cascade: A bank discovers $50M in fraudulent corporate loans. Walk me through every pillar impact — AML, IFRS 9, and Basel — step by step. Which skills does the router chain?" — full cascade
- "An economic downturn causes 15% of Stage 1 assets to migrate to Stage 2. The additional ECL charge is $200M. CET1 was $3.5B and RWA was $25B. What is the new CET1 ratio? Does the bank breach the combined buffer?" — capital impact
- "Why can't a bank have separate IFRS 9, Basel, and AML teams that never talk to each other? Use the fraud cascade scenario to explain the cost of pillar isolation" — integration rationale
Lesson 12: Exercises — IFRS 9 Deep Practice
File: 12-exercises-ifrs9.md
Sidebar Position: 12
Duration: 60 minutes
Summary: Two extended IFRS 9 exercises that integrate staging, ECL calculation, macro scenarios, and PMAs into full portfolio-level problems. Exercise 10 builds a complete portfolio ECL from scratch. Exercise 11 models a stage migration cascade triggered by a sector downturn.
Learning Objectives:
- Build a full portfolio ECL model across multiple segments with scenario weighting (Bloom: Create, CEFR: B2)
- Model a stage migration cascade and calculate the incremental provision impact (Bloom: Analyze, CEFR: B2)
Stage: Layer 2 (Collaboration) — AI supports calculation, student designs model structure
CEFR Proficiency: B2 (synthesis of Lessons 3-5 concepts)
New Concepts (count: 2 -- within B2 limit):
- Portfolio-level ECL aggregation across segments (connecting segment-level ECL to portfolio total)
- Stage migration cascade modelling (sector downturn triggering sequential stage migrations)
Cognitive Load Validation: 2 new concepts + application of 7+ prior concepts -- appropriate for B2
Maps to Evals: 3, 4
Bloom's Level: Create/Analyze
Plugin Skills Used: ifrs9-staging, ifrs9-ecl-calculator, ifrs9-macro-scenarios, ifrs9-pma
Key Concepts Introduced: Portfolio aggregation, cascade modelling
Exercise: Exercise 10 — Full Portfolio ECL Build
- Scenario: $15B diversified lending portfolio with 5 segments (residential mortgage, commercial property, corporate, SME, consumer)
- Task: Build complete ECL model — segment PDs, LGDs, EADs, staging, 3 scenarios, probability weights, 1 PMA
- Duration: 35 min
Exercise: Exercise 11 — Stage Migration Cascade
- Scenario: Commercial real estate downturn — 25% of CRE Stage 1 migrates to Stage 2, 10% of Stage 2 migrates to Stage 3
- Task: Calculate incremental ECL, trace the cascade, design a management overlay for the CRE sector
- Duration: 25 min
Try With AI Prompt Themes:
- "Help me build a full portfolio ECL model. Start by defining 5 segments with their PD, LGD, and EAD assumptions. Then stage the portfolio and calculate ECL under 3 scenarios" — portfolio build
- "Model this cascade: 25% of CRE Stage 1 ($2B) migrates to Stage 2, and 10% of CRE Stage 2 ($800M) migrates to Stage 3. What is the incremental provision charge?" — cascade modelling
- "Review my ECL model. Are there any segments where the PMA seems insufficient? Where might model risk be highest?" — model review
Lesson 13: Exercises — Basel and AML Deep Practice
File: 13-exercises-basel-aml.md
Sidebar Position: 13
Duration: 55 minutes
Summary: Two extended exercises — one for Basel (ICAAP stress testing) and one for AML (cross-border investigation). Exercise 12 builds an ICAAP stress scenario that tests capital adequacy under severe but plausible conditions. Exercise 13 investigates a cross-border AML case spanning multiple jurisdictions and entity types.
Learning Objectives:
- Design an ICAAP stress scenario and calculate stressed capital ratios (Bloom: Create, CEFR: B2)
- Investigate a cross-border AML case, identifying suspicious patterns across jurisdictions and entity types (Bloom: Analyze, CEFR: B2)
Stage: Layer 2 (Collaboration) — AI supports scenario design and investigation, student drives structure and conclusions
CEFR Proficiency: B2
New Concepts (count: 3 -- within B2 limit):
- ICAAP (Internal Capital Adequacy Assessment Process) — bank's own view of capital needs under stress
- Reverse stress testing — what combination of events would cause the bank to fail?
- Cross-border AML investigation — jurisdictional cooperation, mutual legal assistance, information sharing constraints
Cognitive Load Validation: 3 new concepts + application of prior Basel and AML concepts -- appropriate for B2
Maps to Evals: 6, 9
Bloom's Level: Create/Analyze
Plugin Skills Used: basel-capital-ratios, basel-rwa-sa, basel-rwa-irb, aml-tm-rules, aml-sar-workflow
Key Concepts Introduced: ICAAP, reverse stress testing, cross-border AML cooperation
Exercise: Exercise 12 — ICAAP Stress Scenario
- Scenario: Design a severe-but-plausible stress scenario for a mid-size bank — GDP decline, unemployment spike, property crash, interest rate shock
- Task: Define macroeconomic variables, calculate stressed PDs and LGDs, compute stressed RWA and capital ratios, determine capital shortfall if any
- Duration: 30 min
Exercise: Exercise 13 — Cross-Border AML Investigation
- Scenario: Complex investigation spanning 3 jurisdictions with shell companies, trade-based money laundering indicators, and PEP connections
- Task: Map the entity network, identify suspicious patterns, determine which jurisdiction leads, explain information-sharing constraints, draft investigation summary
- Duration: 25 min
Try With AI Prompt Themes:
- "Help me design an ICAAP stress scenario. GDP falls 4%, unemployment rises to 9%, property values drop 25%. What happens to PDs, LGDs, RWA, and capital ratios?" — stress scenario design
- "A company in Dubai imports goods worth $10M from a Hong Kong entity, but the goods are shipped to Nigeria. The Dubai company is owned by a PEP from Country X. Map this for red flags" — AML pattern analysis
- "In reverse stress testing, what combination of events would cause a bank with CET1=12% to breach the minimum 4.5%? Work backwards from failure" — reverse stress testing
Lesson 14: Bank Reconciliation — Nostro, Suspense, and GL-to-Risk
File: 14-reconciliation-nostro-suspense.md
Sidebar Position: 14
Duration: 45 minutes
Summary: Covers the four types of bank reconciliation critical for data integrity — nostro (correspondent bank accounts), suspense (unallocated items), GL-to-risk-system (ensuring the general ledger and risk systems agree), and four-way provision reconciliation (GL provision, risk system provision, regulatory return, IFRS 9 model output). Demonstrates how AI automates matching, flags exceptions, and generates reconciliation reports. Includes Exercise 14: four-way provision reconciliation.
Learning Objectives:
- Perform nostro reconciliation, identifying and classifying breaks (Bloom: Apply, CEFR: B1)
- Clear suspense items by matching to source transactions (Bloom: Apply, CEFR: B1)
- Execute four-way provision reconciliation across GL, risk system, regulatory return, and IFRS 9 model (Bloom: Apply, CEFR: B2)
Stage: Layer 2 (Collaboration) — AI performs matching automation, student investigates breaks and exceptions
CEFR Proficiency: B1-B2
New Concepts (count: 6 -- within B1 limit of 10):
- Nostro reconciliation — bank's record vs correspondent's record, timing differences, errors
- Suspense account clearing — unallocated items, ageing analysis, escalation
- GL-to-risk-system reconciliation — why the general ledger and risk system must agree
- Four-way provision reconciliation (GL provision vs risk system provision vs regulatory return vs IFRS 9 model output)
- Break classification (timing, methodology, data, error)
- Agent automation: matching, exception flagging, report generation; humans investigate and resolve
Cognitive Load Validation: 6 concepts at B1 level -- WITHIN LIMIT
Maps to Evals: 11
Bloom's Level: Apply
Plugin Skills Used: recon-nostro, recon-suspense, recon-gl-risk, recon-provision-four-way
Key Concepts Introduced: Four reconciliation types, break classification, agent automation boundary
Exercise: Exercise 14 — Four-Way Provision Reconciliation
- Scenario: GL shows $450M provision, risk system shows $465M, regulatory return shows $455M, IFRS 9 model output shows $462M
- Task: Identify and classify breaks, trace each difference to its source (timing, methodology, data capture), determine which figure is "correct" and what adjustments are needed
- Duration: 20 min embedded in lesson
Try With AI Prompt Themes:
- "Perform a nostro reconciliation. Our books show $12.5M. The correspondent's statement shows $12.3M. Here are the items: [provide 10 items]. Classify each as timing, error, or missing item" — nostro recon
- "A suspense account has 47 unallocated items totalling $3.2M, some aged 90+ days. Prioritise and suggest clearing actions" — suspense clearing
- "GL provision is $450M but the IFRS 9 model says $462M. The risk system says $465M. The regulatory return says $455M. Reconcile all four and explain each difference" — four-way recon
Lesson 15: Full Banking Agent — Skill Library Build and Capstone
File: 15-full-skill-library-capstone.md
Sidebar Position: 15
Duration: 90 minutes
Summary: Capstone lesson — students deploy the full 17-skill banking plugin, verify all skills are operational, and execute a comprehensive cross-pillar scenario. The scenario requires IFRS 9, Basel, AML, and reconciliation skills working together. Students document the skill library, create workflow recipes for daily/monthly/quarterly cycles, and validate the complete system.
Learning Objectives:
- Deploy and verify the complete 17-skill banking plugin (Bloom: Apply, CEFR: B1)
- Execute a comprehensive cross-pillar scenario using skill chaining (Bloom: Create, CEFR: B2)
- Document the skill library with usage guides and workflow recipes (Bloom: Create, CEFR: B2)
Stage: Layer 2-3 (Collaboration transitioning to Intelligence Design) — students begin treating the plugin as a deployable Digital FTE
CEFR Proficiency: B2
New Concepts (count: 3 -- within B2 limit):
- Full plugin deployment and verification testing
- Workflow recipes for operational cycles (daily recon, monthly ECL, quarterly ICAAP, annual SREP)
- Skill library documentation — usage guides, invocation patterns, limitations
Cognitive Load Validation: 3 new concepts + integration of all prior concepts -- appropriate for B2 capstone
Maps to Evals: 12 (and all prior evals integrated)
Bloom's Level: Create
Plugin Skills Used: ALL 17 skills
Key Concepts Introduced: Full deployment, operational workflow design, documentation
Capstone Scenario:
- A mid-size bank ($30B assets) faces an economic downturn
- Phase 1: ECL recalculation with macro deterioration (IFRS 9 skills)
- Phase 2: Capital adequacy assessment post-provision increase (Basel skills)
- Phase 3: AML review triggered by increased default rates in high-risk sectors (AML skills)
- Phase 4: Reconciliation of all figures across GL, risk system, and regulatory returns (Recon skills)
- Phase 5: Board report summarising cross-pillar impact
- Duration: 50 min for capstone + 25 min for documentation + 15 min for workflow recipes
Try With AI Prompt Themes:
- "Deploy the banking plugin and verify all 17 skills are active. List each skill, its pillar, and run a test query for each" — deployment verification
- "Execute the full capstone scenario: economic downturn hits, ECL increases $300M, CET1 drops, AML flags sector concentration, reconciliation breaks appear. Walk through all 4 phases" — cross-pillar capstone
- "Create a workflow recipe for the month-end regulatory cycle: Day 1-5 (ECL rerun), Day 5-10 (capital ratios), Day 10-15 (AML review), Day 15-20 (reconciliation), Day 20-25 (regulatory returns)" — operational workflow
IV. Skill Dependencies
Lesson Dependency Graph
L01 (Three Pillars) ─────────────────────────────────────────────────────────────────┐
│ │
└──> L02 (Plugin Architecture) ────────────────────────────────────────────────────┐│
│ ││
├──> L03 (IFRS9 Staging/ECL) ──> L04 (PD/LGD/EAD) ──> L05 (Macro/PMA) ││
│ │ │ ││
│ └───────────────┘ ││
│ │ ││
├──> L06 (Basel Capital) ──> L07 (RWA/IRB) ──> L08 (Leverage/Liquidity) ││
│ │ │ ││
│ └────────────────┘ ││
│ │ ││
├──> L09 (AML Three Lines) ──> L10 (TM/SAR) ││
│ │ ││
│ │ ││
└───────────────────────────────────────────────────────────────────────────>││
││
L03-L05 + L06-L08 + L09-L10 ──> L11 (Cross-Pillar Integration) ││
│ ││
├──> L12 (IFRS9 Exercises) [requires L03-L05]││
├──> L13 (Basel/AML Exercises) [requires L06-L10]│
│ ││
L03-L05 + L06-L08 + L09-L10 ──> L14 (Reconciliation) [independent of L11-L13] ││
│ ││
L11 + L12 + L13 + L14 ──> L15 (Capstone) <─────────────────────────────────────────┘│
│
Three Parallel Tracks (L03-L10)
The three pillars form three parallel tracks after L02:
- IFRS 9 Track: L03 -> L04 -> L05 (must be sequential — each builds on prior)
- Basel Track: L06 -> L07 -> L08 (must be sequential — ratios require RWA, which requires capital)
- AML Track: L09 -> L10 (must be sequential — TM/SAR requires three lines foundation)
The three tracks are independent of each other — a student could do Basel before IFRS 9 or vice versa. However, the recommended order is IFRS 9 first because Basel capital ratios are affected by IFRS 9 provisions (retained earnings reduction).
Convergence Point
L11 (Cross-Pillar Integration) is the convergence lesson that requires ALL three tracks to be complete. This is where pillar interactions are demonstrated.
Cross-Chapter Dependencies
- Chapter 17 (Finance Domain Agents): Base finance plugin commands (
/journal-entry,/reconciliation,/income-statement) assumed available - Chapter 18 (IDFA): Financial architecture methodology assumed familiar
- Chapter 19 (CA/CPA Practice): Domain agent concept assumed understood
- Chapter 20 (Islamic Finance): Router-product-overlay pattern assumed understood — Ch21 extends it to pillar-aware routing
V. Plugin Skill Inventory (17 Skills)
Router (1 skill)
| # | Skill Name | Pillar | Purpose |
|---|---|---|---|
| 1 | banking-router | Cross-pillar | Detects pillar(s) from query, loads appropriate skills, chains for cross-pillar queries |
IFRS 9 Skills (5 skills)
| # | Skill Name | Pillar | First Used In | Purpose |
|---|---|---|---|---|
| 2 | ifrs9-staging | Accounting | L03 | Stage 1/2/3 classification, SICR analysis |
| 3 | ifrs9-ecl-calculator | Accounting | L03 | ECL calculation (PD x LGD x EAD), 12-month vs lifetime |
| 4 | ifrs9-pd-model | Accounting | L04 | PD term structures, PIT calibration, forward-looking adjustment |
| 5 | ifrs9-lgd-model | Accounting | L04 | Downturn LGD, cure rates, collateral haircuts |
| 6 | ifrs9-ead-calculator | Accounting | L04 | EAD computation, CCFs for off-balance-sheet |
| 7 | ifrs9-macro-scenarios | Accounting | L05 | Scenario design, probability weighting |
| 8 | ifrs9-pma | Accounting | L05 | Post-model adjustment design and governance |
Basel III/IV Skills (4 skills)
| # | Skill Name | Pillar | First Used In | Purpose |
|---|---|---|---|---|
| 9 | basel-capital-stack | Solvency | L06 | Capital classification, deductions |
| 10 | basel-capital-ratios | Solvency | L06 | CET1, Tier 1, Total Capital ratio calculation |
| 11 | basel-rwa-sa | Solvency | L07 | Standardised Approach RWA |
| 12 | basel-rwa-irb | Solvency | L07 | IRB RWA, output floor |
| 13 | basel-leverage-ratio | Solvency | L08 | Leverage ratio calculation |
| 14 | basel-lcr | Solvency | L08 | LCR calculation, HQLA classification |
| 15 | basel-nsfr | Solvency | L08 | NSFR calculation |
AML/KYC Skills (3 skills)
| # | Skill Name | Pillar | First Used In | Purpose |
|---|---|---|---|---|
| 16 | aml-cdd-screening | Financial Crime | L09 | CDD/EDD classification, PEP screening |
| 17 | aml-pep-check | Financial Crime | L09 | PEP database check, family/associate screening |
| 18 | aml-risk-scoring | Financial Crime | L09 | Customer risk scoring, jurisdiction risk |
| 19 | aml-tm-rules | Financial Crime | L10 | Rules-based transaction monitoring |
| 20 | aml-tm-ml | Financial Crime | L10 | ML-based anomaly detection |
| 21 | aml-sar-workflow | Financial Crime | L10 | SAR workflow guidance (not filing — human required) |
Reconciliation Skills (4 skills)
| # | Skill Name | Pillar | First Used In | Purpose |
|---|---|---|---|---|
| 22 | recon-nostro | Operations | L14 | Nostro reconciliation, break classification |
| 23 | recon-suspense | Operations | L14 | Suspense clearing, ageing analysis |
| 24 | recon-gl-risk | Operations | L14 | GL-to-risk-system reconciliation |
| 25 | recon-provision-four-way | Operations | L14 | Four-way provision reconciliation |
Note: The plugin spec (Issue #817) specifies 17 skills (1 router + 16 product). The table above lists 25 distinct skills for maximum granularity. For the plugin implementation, some of these will be consolidated — for example, ifrs9-pd-model, ifrs9-lgd-model, and ifrs9-ead-calculator may be sections within a single ifrs9-ecl-components skill. The final skill count will be determined during plugin development. The lesson plan references the granular skill names for pedagogical clarity.
Consolidation mapping for 17-skill target (1 router + 16 product):
| Consolidated Skill | Lesson Granular Skills |
|---|---|
banking-router | banking-router |
ifrs9-staging | ifrs9-staging |
ifrs9-ecl | ifrs9-ecl-calculator, ifrs9-pd-model, ifrs9-lgd-model, ifrs9-ead-calculator |
ifrs9-scenarios | ifrs9-macro-scenarios, ifrs9-pma |
basel-capital | basel-capital-stack, basel-capital-ratios |
basel-rwa | basel-rwa-sa, basel-rwa-irb |
basel-leverage-ratio | basel-leverage-ratio |
basel-lcr | basel-lcr |
basel-nsfr | basel-nsfr |
aml-cdd | aml-cdd-screening, aml-pep-check, aml-risk-scoring |
aml-tm | aml-tm-rules, aml-tm-ml |
aml-sar | aml-sar-workflow |
recon-nostro | recon-nostro |
recon-suspense | recon-suspense |
recon-gl-risk | recon-gl-risk |
recon-provision | recon-provision-four-way |
Total: 1 router + 16 product = 17 skills
VI. Exercise Map
| Exercise | Lesson | Duration | Pillar | Description |
|---|---|---|---|---|
| Ex 1 | L04 | 20 min | IFRS 9 | Retail mortgage ECL calculation (PD/LGD/EAD) |
| Ex 2 | L05 | 15 min | IFRS 9 | Commercial portfolio macroeconomic overlay + PMA |
| Ex 3 | L06 | 20 min | Basel | Bank capital ratio calculation (CET1, T1, TC) |
| Ex 4 | L07 | 25 min | Basel | RWA comparison: SA vs IRB with output floor |
| Ex 5 | L08 | 15 min | Basel | Liquidity stress test (LCR calculation) |
| Ex 6 | L09 | 15 min | AML | Customer onboarding risk assessment (CDD/EDD) |
| Ex 7 | L10 | 20 min | AML | TM alert investigation (3 alerts) |
| Ex 8 | L11 | 20 min | Cross-pillar | Fraud-triggered stage migration cascade |
| Ex 9 | L11 | 15 min | Cross-pillar | Capital impact of ECL spike |
| Ex 10 | L12 | 35 min | IFRS 9 | Full portfolio ECL build (5 segments, 3 scenarios) |
| Ex 11 | L12 | 25 min | IFRS 9 | Stage migration cascade (CRE downturn) |
| Ex 12 | L13 | 30 min | Basel | ICAAP stress scenario design |
| Ex 13 | L13 | 25 min | AML | Cross-border AML investigation |
| Ex 14 | L14 | 20 min | Recon | Four-way provision reconciliation |
Total: 14 exercises, estimated 22 hours of practice (including lesson context + exercise time)
Exercise Progression:
- Exercises 1-7: Single-pillar, embedded in teaching lessons (build competence)
- Exercises 8-9: Cross-pillar, in integration lesson (connect pillars)
- Exercises 10-13: Deep practice, in dedicated exercise lessons (build fluency)
- Exercise 14: Reconciliation, in dedicated lesson (operational competence)
- Capstone (L15): Not numbered as exercise — full cross-pillar scenario using all skills
VII. Assessment Plan
Formative Assessments (During Lessons)
| Lesson | Assessment Type | What's Measured |
|---|---|---|
| L01 | Pillar identification quiz | Can student classify scenarios into correct pillar(s) |
| L02 | Routing trace | Can student trace a query through the skill architecture |
| L03 | Staging exercise | Can student apply SICR criteria to classify assets |
| L04 | ECL calculation | Can student compute PD x LGD x EAD correctly |
| L05 | Scenario weighting | Can student calculate probability-weighted ECL |
| L06 | Capital ratio computation | Can student compute CET1, Tier 1, Total Capital ratios |
| L07 | RWA comparison | Can student apply SA risk weights and IRB output floor |
| L08 | LCR calculation | Can student classify HQLA and compute LCR |
| L09 | Risk classification | Can student apply CDD/EDD framework |
| L10 | Alert investigation | Can student evaluate TM alerts and determine disposition |
| L11 | Cross-pillar trace | Can student trace a cascade across all three pillars |
| L14 | Reconciliation | Can student identify and classify breaks |
Summative Assessment (End of Chapter)
- L15 Capstone: Cross-pillar scenario requiring all 17 skills, producing ECL report + capital assessment + AML review + reconciliation + board summary
Assessment Alignment
All assessments follow:
- CEFR tier: B1 (Lessons 1-10, 14), B2 (Lessons 11-13, 15)
- Bloom's progression: Understand (L01) -> Apply (L02-L10, L14) -> Analyze (L10-L11, L13) -> Create (L12-L13, L15)
VIII. Validation Checklist
Chapter-Level Validation:
- Chapter type identified: Technical/Hybrid (regulatory concepts + quantitative exercises + plugin architecture)
- Concept density analysis documented: 21 core concepts across 3 pillars + recon
- Lesson count justified: 15 lessons driven by 21 concepts + 14 exercises + capstone (not arbitrary template)
- All 12 evals covered by lessons
- All lessons map to at least one eval
Stage Progression Validation:
- Pedagogical layer: L2 (Collaboration) — students have plugin, AI works alongside them
- Plugin installed in L02 and used throughout (no L1 manual-only phase — appropriate for Part 3 professional audience who have completed 20 prior chapters)
- Three Roles integrated: AI teaches regulatory concepts, student drives classification/calculation decisions, convergence through iterative model building
Cognitive Load Validation:
- L01: 5 concepts (A2 limit: 7) -- PASS
- L02: 6 concepts (B1 limit: 10) -- PASS
- L03: 7 concepts (B1 limit: 10) -- PASS
- L04: 8 concepts (B1 limit: 10) -- PASS
- L05: 6 concepts (B1 limit: 10) -- PASS
- L06: 7 concepts (B1 limit: 10) -- PASS
- L07: 7 concepts (B1 limit: 10) -- PASS
- L08: 6 concepts (B1 limit: 10) -- PASS
- L09: 6 concepts (B1 limit: 10) -- PASS
- L10: 6 concepts (B1 limit: 10) -- PASS
- L11: 5 concepts (B2 limit: 10) -- PASS
- L12: 2 new concepts (B2 limit: 10) -- PASS
- L13: 3 new concepts (B2 limit: 10) -- PASS
- L14: 6 concepts (B1 limit: 10) -- PASS
- L15: 3 new concepts (B2 limit: 10) -- PASS
Dependency Validation:
- IFRS 9 track: L03 -> L04 -> L05 (sequential, each builds on prior)
- Basel track: L06 -> L07 -> L08 (sequential, capital -> RWA -> leverage/liquidity)
- AML track: L09 -> L10 (sequential, three lines -> TM/SAR)
- Cross-pillar: L11 requires L03-L10 (all three tracks)
- Exercise bundles: L12 requires L03-L05, L13 requires L06-L10
- Reconciliation: L14 requires L02 (plugin) but independent of L11-L13
- Capstone: L15 requires L11-L14
Cross-Chapter Dependencies:
- Chapter 17 (base finance plugin): prerequisite -- available
- Chapter 18 (IDFA methodology): prerequisite -- available
- Chapter 19 (domain agents): prerequisite -- available
- Chapter 20 (router-product-overlay): prerequisite -- available
Pattern Alignment:
- Follows Ch20 README.md structure (frontmatter, lesson flow table, chapter contract, plugin install, after-chapter)
- Follows Ch20 lesson YAML frontmatter pattern (skills, learning_objectives, cognitive_load, differentiation)
- Plugin architecture follows Ch18/Ch20 pattern (install via catalog, auto-activation, commands)
- Three "Try With AI" prompts per lesson with "What you are learning" explanations