Career+: Build Controls Around AI-Assisted Finance Work

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-finance-ai-control-checklist-creators

1. A finance team wants to use AI to help analyze quarterly revenue data. What is the primary reason they should keep the source data in approved company systems rather than pasting it into consumer chat tools?

   1. Approved systems have better formatting features for charts
   2. Consumer chat tools may not handle large datasets efficiently
   3. Consumer chat tools cannot perform mathematical calculations on revenue figures
   4. Approved systems maintain data security and prevent sensitive financial information from leaking to external platforms

2. What information should be stored in a prompt/output log for material financial analysis assisted by AI?

   1. The exact prompt used, the complete output, and metadata about who ran the analysis and when
   2. Just a summary of what the AI was asked to do, since full transcripts are too long
   3. Only the final output generated by the AI to save storage space
   4. The AI model's version number and any error messages that occurred during processing

3. Which type of financial document requires mandatory human review before being considered final, according to the controls framework?

   1. Informal emails discussing lunch expenses
   2. Personal notes calculating monthly spending for budgeting practice
   3. Board materials, forecasts, and budgets prepared with AI assistance
   4. Internal brainstorming notes about potential cost savings

4. Why should AI-generated text in working papers be labeled as AI-assisted when it will undergo review?

   1. To ensure reviewers know which parts of the analysis originated from AI and may need extra scrutiny for accuracy
   2. To indicate that the work is incomplete and needs more editing
   3. To satisfy legal requirements that all AI use must be disclosed in every document
   4. To make the document appear more sophisticated to senior management

5. What does the control practice of separating drafting from approval involve?

   1. Using one software tool for drafting and a different tool for final approval
   2. Drafting documents in the morning and approving them in the evening to allow time for review
   3. Having one person create initial work and a different person review and approve the final version
   4. Ensuring that AI tools are never used during the drafting phase of a document

6. A small internal team is analyzing marketing spend using AI for their own planning. Why might this require fewer controls than preparing board materials?

   1. AI is more accurate for small projects than large ones
   2. Small projects never involve sensitive data
   3. Internal teams are not required to follow any financial controls
   4. The risk of incorrect analysis affecting organizational decisions is lower for internal planning than for board reporting

7. What is the main purpose of maintaining an approved tool list for AI in finance?

   1. To limit the number of AI tools available to finance teams
   2. To guarantee that approved tools always produce correct analysis
   3. To prevent sensitive financial data from being entered into consumer-grade or unvetted AI tools
   4. To ensure all AI tools produce the same formatting in documents

8. What specific risk does a formula lock control address in AI-assisted spreadsheet work?

   1. AI converting text numbers to actual numbers incorrectly
   2. AI creating viruses that corrupt spreadsheet files
   3. AI silently changing spreadsheet formulas or logic without the user's awareness
   4. AI accidentally deleting the entire spreadsheet

9. In the context of working papers, what does labeling AI-generated text accomplish that would not happen without the label?

   1. It signals to reviewers that they should pay special attention to those sections during their review
   2. It prevents the AI from generating text in future documents
   3. It automatically corrects any mistakes the AI made
   4. It legally protects the company if the AI makes an error

10. What makes an AI-assisted financial analysis difficult or impossible to review after the fact?

    1. The analysis was completed using complex mathematical formulas
    2. The final report was shared with more than five people
    3. The prompts and outputs were not saved, leaving no record of what the AI was asked or what it produced
    4. The analysis used data from more than one fiscal quarter

11. A finance analyst uses AI to draft budget projections, then has their manager review and initial the document before final submission. Which control principle is being demonstrated?

    1. Formula lock
    2. Separation of drafting from approval
    3. Prompt/output logging
    4. Source data isolation

12. Which implementation represents a lightweight approach to maintaining a prompt/output log for AI-assisted analysis?

    1. Requiring IT to install specialized logging software on all finance computers
    2. Saving the chat transcript directly in the workpaper file or adding a note about the AI interaction
    3. Setting up an automated database with version control and audit trails
    4. Creating a separate physical logbook where analysts handwrite each prompt

13. A protected spreadsheet with version history enabled represents what type of control implementation?

    1. Approved tool list
    2. Formula lock with diff review
    3. Prompt/output log
    4. Human review requirement

14. Why is reviewer initials and timestamp considered an approval step control?

    1. It provides evidence that a human actually reviewed and accepted the work before it became final
    2. It makes the document look more professional
    3. It automatically corrects any errors found during review
    4. It allows the AI to learn from the reviewer's feedback

15. When would the control burden for AI-assisted finance work appropriately be lighter?

    1. When the analysis will be used for external regulatory filings
    2. When the analysis is for internal brainstorming that won't influence organizational decisions
    3. When the analysis involves publicly available data
    4. When the finance team is under time pressure to complete the work quickly