Chapter 7 Quiz: Claude Code for Finance Professionals
Test your understanding of applying Claude Code as a Financial Reasoning Engine. This quiz covers the concepts, workflows, and governance principles introduced throughout the chapter.
Pass Threshold: 80% (12/15 correct)
Answer Key
Question 1: Financial Reasoning Engine
Answer: B - It interprets intent and reasons contextually rather than executing predefined rules
Explanation: The core distinction is reasoning versus rules. Traditional tools execute IF-THEN logic without understanding meaning. Claude interprets context, considers alternatives, and explains its reasoning—enabling it to handle ambiguous cases that break rigid rules.
Lesson Reference: Lesson 1 - Claude Code General Agent for Finance
Question 2: OODA Loop Orient Phase
Answer: C - Claude identifies patterns and relevant context from prior transactions
Explanation: Orient is the pattern recognition and context-building phase. When analyzing a transaction, Claude looks at vendor history, amount patterns, and similar past transactions to build understanding before making a classification decision.
Lesson Reference: Lesson 1 - Claude Code General Agent for Finance
Question 3: CLAUDE.md Contents
Answer: B - Chart of accounts, materiality thresholds, accounting policies, and governance rules
Explanation: CLAUDE.md provides persistent organizational context. It should contain the structural and policy information Claude needs to make consistent, organization-appropriate recommendations—not transaction data or sensitive employee information.
Lesson Reference: Lesson 2 - Finance Workspace Setup with CLAUDE.md
Question 4: Finance Skill Anatomy
Answer: C - Auto-posting capability (automatic ledger modifications)
Explanation: The four components are Persona, Logic, Context, and Safety. Auto-posting capability should never be part of a finance skill because it violates the fundamental governance principle: AI proposes, human approves, system records.
Lesson Reference: Lesson 3 - Creating Finance Skills
Question 5: Variance Analyzer Thresholds
Answer: B - Because specific thresholds make the skill actionable rather than vague
Explanation: Concrete thresholds enable consistent, repeatable behavior. "Significant variance" is subjective; "> 5% AND > $10,000" is objective. Organizations can adjust thresholds to match their materiality definitions while maintaining consistency.
Lesson Reference: Lesson 3 - Creating Finance Skills
Question 6: Three-Layer Architecture
Answer: B - Execution layer: deterministic computation and formula calculation
Explanation: Google Sheets serves as the Execution layer in the three-layer architecture. Claude provides Intelligence (reasoning), Sheets provides Execution (reliable calculation), and Humans provide Governance (approval and oversight).
Lesson Reference: Lesson 4 - Connecting to Google Sheets via MCP
Question 7: CLAUDE.md for Formulas
Answer: B - The fiscal year dates and department structure
Explanation: Claude references CLAUDE.md for organization-specific context. If your fiscal year runs July-June, Claude should use those dates in SUMIFS formulas, not calendar year dates. This context makes outputs appropriate for your organization.
Lesson Reference: Lesson 4 - Connecting to Google Sheets via MCP
Question 8: Intent-Driven Modeling Benefit
Answer: B - It translates natural language objectives into structured financial logic with documented assumptions
Explanation: Intent-driven modeling lets you describe financial scenarios in business terms ("lose our largest customer in Q2") and have Claude translate that into structured projections with explicit, documented assumptions. This reduces technical barriers while maintaining auditability.
Lesson Reference: Lesson 5 - Intent-Driven Financial Modeling
Question 9: Assumption Validation
Answer: B - Because assumptions should be explicit, traceable, and challengeable before they affect projections
Explanation: Claude surfaces assumptions so you can validate them against your business knowledge. You might know the customer loss will be gradual, not sudden. Explicit assumptions create audit-ready documentation and ensure models reflect reality.
Lesson Reference: Lesson 5 - Intent-Driven Financial Modeling
Question 10: AI Classification Governance (Scenario)
Answer: B - Review each classification, verify the reasoning, approve or correct, then post
Explanation: The AI-assisted, human-governed model requires human review before any ledger changes. Review Claude's reasoning for each transaction, approve correct classifications, correct errors, then post. This maintains accountability and audit trail integrity.
Lesson Reference: Lesson 6 - AI-Native Accounting Workflows
Question 11: Bank Reconciliation Action (Scenario)
Answer: B - Claude explains the difference; you record the wire fee journal entry and note the outstanding check
Explanation: Claude's role is to identify and explain the discrepancy. Your role is to take action: create the journal entry for the wire fee and document the outstanding check as a reconciling item. Claude reasons and explains; you execute and approve.
Lesson Reference: Lesson 6 - AI-Native Accounting Workflows
Question 12: Finance Subagent Roles
Answer: B - Each mirrors a specific finance team function with isolated context and focused expertise
Explanation: Finance subagents parallel how finance teams work: specialized roles with focused expertise. Isolated context means each subagent isn't distracted by unrelated information, producing more focused, higher-quality outputs.
Lesson Reference: Lesson 8 - Finance Subagents
Question 13: Multi-Agent vs Single Conversation
Answer: C - Multi-agent adds value when work involves distinct specialties and structured handoffs; single conversation wins for iterative back-and-forth
Explanation: Choose multi-agent for complex workflows with distinct phases (modeling, validation, narrative) and structured handoffs. Choose single conversation for exploratory analysis where context builds iteratively. Match the approach to the workflow structure.
Lesson Reference: Lesson 8 - Finance Subagents
Question 14: Autonomous Posting (True/False)
Answer: C - False - ledger entries are legal records requiring human authorization before posting
Explanation: Ledger entries are legal records with audit, tax, and investor implications. Human authorization before posting is an architectural requirement for accountability, not a limitation of current AI technology. The governance model is: AI proposes, human approves, system records.
Lesson Reference: Lesson 9 - Governance, Compliance, and Safety
Question 15: AI and Safety (True/False)
Answer: A - True - AI surfaces anomalies and documents reasoning that pure automation would miss
Explanation: The governance paradox shows that properly governed AI increases safety. AI detects patterns humans miss, documents reasoning for audit trails, maintains consistency, and surfaces anomalies. Governance transforms AI from potential risk into safety enhancement.
Lesson Reference: Lesson 9 - Governance, Compliance, and Safety
Cognitive Distribution
| Level | Questions | Percentage |
|---|---|---|
| Recall/Remember | Q2, Q3, Q4 | 20% |
| Understand/Apply | Q1, Q5, Q6, Q7, Q8, Q10, Q12 | 47% |
| Analyze/Evaluate | Q9, Q11, Q13, Q14, Q15 | 33% |
Question Types:
- Multiple Choice: 10 questions
- Scenario-Based: 3 questions (Q10, Q11, Q13)
- True/False with Explanation: 2 questions (Q14, Q15)