Cognitive Sovereignty Self-Audit for Chief Financial Officers

1. When Bloomberg Terminal AI or Anaplan generates a revenue forecast with a confidence interval, what do you do before presenting it to your board?

Present the AI output directly. The confidence interval is the stress test. Review the confidence interval but accept the underlying assumptions without challenging them. Manually interrogate 3 to 4 key assumptions that drove the forecast, then adjust if needed. Build a separate scenario from first principles using different assumptions, then compare it to the AI forecast.

2. Your FP&A team is building next year's financial model. How much of the scenario logic comes from ChatGPT or Copilot starting points versus original team analysis?

Most scenarios start with AI suggestions. The team refines what the AI proposes. About half the scenarios use AI starting points. The other half come from the team. The team builds scenarios from their own logic, then uses AI to sense-check the maths. The team develops all scenario logic independently. AI tools are used only for calculation verification.

3. A Tableau AI dashboard shows your cash position trending red three months out. What happens next?

You trust the visualisation and brief the board on the risk without checking the data feed. You review the chart but rely on the AI's interpretation of what is driving the trend. You verify the underlying data and manually check whether seasonal or one-off items explain the trend. You examine the data, remodel the drivers yourself, and build your own narrative before using the dashboard insight.

4. During M&A due diligence, your team uses AI tools to flag forecast anomalies and risk areas. How do you use these flags?

The AI flags become your due diligence agenda. You investigate what the AI suggests. You use AI flags as one input but also conduct some independent risk assessment. You develop your own diligence questions first, then use AI to help execute them. You build your full diligence plan from industry knowledge and deal specifics, then use AI only to accelerate data gathering.

5. Your board expects a narrative explaining profit variance this quarter. Do you write it yourself or rely on Copilot to generate a first draft?

Copilot writes the narrative. You edit for tone and send it. Copilot generates a draft. You add your own insight but keep most of the AI structure. You write the narrative yourself, then use Copilot to check for clarity and completeness. You write and own the narrative entirely. Copilot is not involved in variance story development.

6. When reviewing audit findings or internal control assessments flagged by AI risk tools, how do you engage?

You accept the AI's risk rating and severity classification as stated. You review the AI assessment and adjust only obvious errors. You assess each flagged risk independently, then compare your judgement to the AI rating. You evaluate all material risks using your own control knowledge, then use AI findings to check for gaps you may have missed.

7. Your team notices that board narratives this year look increasingly similar in structure and language. What is your response?

No response. Consistency in reporting format is efficient. You mention it but continue using the same AI templates because they work. You conduct a review of last four quarters of board reports and identify where narratives have become generic. You rebuild your board narrative process to require original insight statements before any drafting begins, then measure variance in narrative structure quarter on quarter.

8. A major assumption in your three year plan is challenged in the board meeting. Can your FP&A team defend it from first principles, or would they need to trace it back to the original AI model output?

The team would reference the AI model. The assumption came from there. The team could explain the assumption but with some gaps in the reasoning. The team can defend the assumption but acknowledges it partly came from an AI starting point. The team built the assumption independently and can explain every component of the logic.

Cognitive Sovereignty Self-Audit for Chief Financial Officers

Cognitive Sovereignty: How To Think For Yourself When AI Thinks For You