GCC APPLIED AI CURRICULUM — FINANCIALS TRACK
A practical, audit-first curriculum built specifically for GCC Financial Analysts, Controllers, and FP&A leaders. 100% No-Code / Zero-Coding Required: Master Microsoft 365 Copilot, Excel Power Query, Power BI Smart Narratives, and Power Automate review workflows to compress monthly close commentary, variance triage, and dashboard reporting with strict GAAP/IFRS numeric integrity.
Establishes the core grounding methodology that prevents hallucinated financial metrics. Teaches analysts how to construct prompt templates with explicit citation tags and strict human review gates.
You are a financial analysis assistant. RULES: 1. ONLY reference numbers present in the VERIFIED_TABLE below. 2. DO NOT perform calculations. Use only pre-computed metrics. 3. Every claim must cite the exact Table Row ID. [VERIFIED_TABLE] Row 1: Gross Margin FY25 = 42.4% (+180 bps YoY) Row 2: Operating Cash Flow = ₹142 crore (94% EBITDA conversion) [/VERIFIED_TABLE]
Trainees are paired in Route A (Finance Pro) + Route B (Data/Tech Analyst) pairs. Finance leads the domain/audit correctness; Data leads pipeline automation.
| Day & Session | Time | Module / Activity | Dual-Track Learner Responsibility |
|---|---|---|---|
| Day 1 • Morning | 09:00 – 11:30 | M1: AI-Assisted Financial Analysis (DuPont) | Finance sets ratio definitions; Data builds prompt bindings. |
| Day 1 • Mid-Day | 11:30 – 14:00 | M2: Automated Reporting & GL Narratives | Finance defines variance materiality; Data sets JSON schema. |
| Day 1 • Afternoon | 14:30 – 17:30 | M3: FP&A Forecasting & Driver Scenarios | Finance specifies business drivers; Data tests LLM sensitivity. |
| Day 1 • Evening | 17:30 – 19:00 | M4: Spreadsheet AI & Copilot Formulas | Hands-on natural-language formula build & verification spot-checks. |
| Day 2 • Morning | 09:00 – 12:00 | M5: Python for Finance & pandas Pipelines | Data codes trial-balance parser; Finance verifies reconciliation ties. |
| Day 2 • Mid-Day | 12:30 – 15:00 | M6: LLM-Powered Variance Root-Cause Triage | Building automated exception commentary with sign-convention checks. |
| Day 2 • Afternoon | 15:00 – 17:00 | M7: Automated CFO Dashboards | Feeding real-time GL variance commentary into executive dashboards. |
| Day 2 • Capstone | 17:00 – 19:00 | M8: Live Capstone Demo & Defense | Live run of automated reporting pipeline judged against 100-pt rubric. |
Designed specifically for corporate GCC Controllers, FP&A Managers, and Shared Services teams who work with Microsoft 365, Excel Power Query, Power BI and visual workflow engines like Make.com and n8n.
Drag-and-drop ingestion of 12 subsidiary CSVs, automated unpivot, and debit-credit tie-out checks in 4 minutes with zero code.
Prompting Copilot in Excel for complex nested XLOOKUP/LAMBDA formulas with mandatory 3-step manual verification.
Drag-and-drop workflow nodes connecting Google Drive/SharePoint $\rightarrow$ LLM Context Envelope $\rightarrow$ Interactive Slack/Teams approval cards.
Embedding dynamic AI commentary blocks directly beside executive CFO dashboard visuals that refresh automatically.
Learners configure visual workflows across the 4 major enterprise no-code platforms without touching code:
[Trigger: New Invoice in NetSuite/Email] ➔ [Zapier Formatter: Extract Totals] ➔ [AI by Zapier: Draft Variance Memo] ➔ [Zapier Approval: Slack Gate].
[Watch Folder: SharePoint/Drive] ➔ [CSV Parser] ➔ [OpenAI/Claude Module] ➔ [Router: High vs Low Variance] ➔ [Teams Approval Card].
[Webhook: Close Event] ➔ [Filter Node: Variance > ₹50 lakh] ➔ [LangChain AI Agent Node] ➔ [Human Wait for Webhook Gate].
[Source Ingestion: 10-Ks & SOX Manuals] ➔ [Instant NotebookLM Grounded Q&A] ➔ [Gemini in Google Sheets] formula generation with inline citations.
Pre-formatted corporate financial workbooks with embedded DuPont ratios, variance triage conditional formatting, and Copilot prompt binding tabs:
labs/excel_templates/lab1_dupont_ratio_model.xlsx — Includes 3-year P&L/BS actuals and Copilot narrative prompt box.
labs/excel_templates/lab3_driver_forecast_model.xlsx — Implements ₹50 lakh/5% materiality flags and F/U sign conventions.
labs/excel_templates/lab4_copilot_formula_verifier.xlsx — 3-step spot-check sheet (TRUE, FALSE, Boundary edge cases).
100% browser-based execution across Microsoft 365 Copilot, Google Workspace & NotebookLM, Make.com, Zapier, and n8n Cloud. No local admin rights, no command-line tools, and no software installations needed on locked-down enterprise laptops.
Test the real-time transaction risk scoring rules taught in Module 5 and the FDE Lab series.
class FinTechRiskEngine:
def evaluate_transaction(self, tx: Dict[str, Any]) -> Dict[str, Any]:
risk_score = 0.10
if tx["amount"] > (tx["historical_daily_avg"] * 4):
risk_score += 0.45
if tx["location"] == "INTERNATIONAL":
risk_score += 0.30
decision = "DECLINE_AND_ALERT" if risk_score >= 0.75 else "APPROVE"
return {"risk_score": risk_score, "decision": decision}
Direct, production-grade open data portals, macroeconomic feeds, regulatory repositories, and Hugging Face datasets utilized across course labs and capstone builds.
Authoritative macroeconomic indicators, interest rates, sovereign debt tables, and consumer price indexes.
Audited corporate annual reports (10-K), quarterly statements (10-Q), and machine-readable XBRL financial lines.
Pre-trained financial domain models (FinBERT, FinGPT) and curated benchmark evaluation datasets.
Sovereign risk ratings, FX historical exchange rates, and credit card fraud benchmark repositories.
All external datasets ingested into corporate pipelines must undergo deterministic schema validation and PII sanitization (masking PAN, SSN, and banking credentials) before being passed into model context envelopes.
Every capstone is judged by the Course Director panel on a live running pipeline. Minimum pass: 60 pts. Controls minimum: 13/25.
Measured (not estimated) manual baseline hours, frequency, and error rate before AI automation.
Every AI claim is explicitly bound to a source GL line or verified ratio. Demonstrates reproducible pipeline.
Enforces "Golden Rule", named human review gate, SOX audit log, and prompt injection defense (Min 13/25).
Documented close compression metrics (e.g. 4 hrs to 20 mins) with realistic enterprise scalability.
Clear presentation, defense against live edge-case scenarios, and dual-track pair synergy.
"Transforms informal AI usage into an auditable, controller-approved operating discipline with strict human review gates."
"Ensures that every AI commentary ties out to verified General Ledger lines — built for external audit acceptance, not toy demos."
"Directly closes the onshore productivity challenge by equipping delivery teams with working Python and spreadsheet automations."