ai-ledger

PythonMuse Trust but Verify Checklist

Audit-friendly AI workflow checklist for accounting and finance teams.

---

How to Use This Checklist

Use this before, during, and after working with AI on any accounting task:

• reconciliation
• payroll testing
• margin analysis
• audit support

If you can check every box, your workflow is controlled and reviewable.

Print it, paste it into Notion, attach it to a workpaper — use it however works for your team.

See Article 13: AI in Accounting Isn’t Just About Efficiency – It’s About Control for the full context.

---

Section 1 – Before Using AI: Data Control

• I am not using raw client or company data
• Sensitive fields (names, IDs, account numbers) are masked
• Dollar amounts are generalized or adjusted if needed
• I am only sharing the minimum data required
• I understand what data I am sending and where it is going

If any box is unchecked – stop here.

Masking techniques are covered in Article 06: How to Use AI in Accounting Without Sending the Wrong Data.

---

Section 2 – Define the Task: Do Not Let AI Guess

• I clearly described the task (reconcile, analyze variance, etc.)
• I asked AI to propose a plan before processing data
• I reviewed and approved the plan
• I confirmed assumptions and definitions

AI should not be deciding the approach on its own. This plan-first pattern is described in Article 11: From One-Time Analysis to Repeatable Workflows.

---

Section 3 – Processing: Controlled Execution

• Data provided matches the approved plan
• No additional or unintended data was included
• Steps performed by AI are clear and logical
• I can explain what AI is doing

If you cannot explain it – do not trust it.

---

Section 4 – Output Review: Trust but Verify

• Results tie back to source data
• Calculations make sense independently of AI
• Any unusual items have been investigated
• Output is reviewed before sharing or posting

Clean output does not mean correct output.

---

Section 5 – Documentation: Make It Reproducible

• Inputs are saved (masked version)
• Outputs are saved in a defined location
• Steps are documented (plan.md or equivalent)
• Results can be recreated if needed
• Someone else could follow this process

If it cannot be reproduced – it is not audit-ready.

The plan.md + status_update.md documentation pattern is described in Article 08: Why Claude “Forgets”. A starter template is available at examples/ai-project-memory.

---

Section 6 – Final Check: Control Mindset

• I would be comfortable explaining this to an auditor
• I understand how the result was generated
• I did not rely solely on AI output
• The process follows trust but verify

---

Quick Reference

Section	Focus	Gate
1. Before Using AI	Data Control	Stop if data is not masked
2. Define the Task	Plan Before Processing	AI proposes, you approve
3. Processing	Controlled Execution	You can explain every step
4. Output Review	Trust but Verify	Results tie to source
5. Documentation	Make It Reproducible	Someone else could repeat this
6. Final Check	Control Mindset	You could explain it to an auditor

---

PythonMuse Reminder

AI in accounting should be:

• Controlled
• Repeatable
• Reviewable
• Explainable

Not:

• Ad hoc
• Opaque
• Dependent on a single chat session

---

From PythonMuse – Practical Python, AI, and automation for accounting and finance teams.

This site is open source. Improve this page.