Audit Trail, PBC Lists, and Walkthrough Prep with AI
Audit season has its own rhythm. Auditors request documents, controllers chase them, walkthroughs are scheduled, and the same questions get asked every year. AI does not change what auditors ask for, but it dramatically compresses the time you spend assembling the answers. This lesson is about turning audit prep from a six-week scramble into a steady two-week deliverable.
What You'll Learn
- How to draft a defensible audit trail memo in 5 minutes instead of an hour
- A prompt for generating PBC (Prepared By Client) lists from scratch
- The walkthrough rehearsal pattern that gets you ready for any process question
- Where AI helps with sample selection narratives and where it must not be used
What Auditors Actually Need
Strip away the formality and your audit team needs four things:
- Evidence that the control was performed. Documents, sign-offs, screenshots, system logs.
- A narrative explaining the process. Walkthroughs of how the control works in practice.
- Population data and sample support. Lists of transactions and the supporting evidence for the sample.
- Memos on judgemental areas. Revenue recognition decisions, accrual estimates, impairment assessments.
AI is excellent at the second and fourth — narratives and memos. It is useful for the first if you have a structured input. It must not be used for sample selection itself; that needs to be defensible and reproducible.
The Audit Trail Memo Prompt
For any control that an auditor will test, you typically need a short memo explaining what the control is, who performs it, and how it operates. Most finance teams write 50 to 200 of these. AI converts each from a 30-minute task to a 5-minute task.
Master prompt:
"Act as a finance controller drafting an audit trail memo for a control. Audience: external auditors. Tone: factual, audit-defensible, no marketing language.
Control name: [e.g., Monthly journal entry review] Control objective: [e.g., Ensure all journal entries above $10,000 posted in the month are reviewed and approved before close] Frequency: [e.g., Monthly] Preparer: [role, not name] Reviewer: [role, not name] Inputs: [e.g., Journal entry listing from GL, exported on day 3 of close] Process: [step-by-step what the reviewer does] Evidence retained: [e.g., Sign-off in close checklist, retained in SharePoint folder X]
Output a 200-word audit-defensible memo with sections: Control description, Operation, Evidence, and Reliance considerations. Avoid speculation about effectiveness — state only what happens."
The output reads exactly like an audit memo from a Big Four manuscript. Save the prompt; rerun for every control you need to document.
Generating a PBC List from Scratch
The PBC list — "Prepared By Client" — is the working list of documents the auditor needs. Mid-tier and Big Four firms have templates, but most small and mid-size finance teams build their own. AI is great at producing a comprehensive PBC list calibrated to your business.
Prompt:
"Generate a comprehensive PBC list for a financial statement audit of a [size: e.g., $50m revenue] [industry: e.g., SaaS] company under [framework: e.g., US GAAP / UK GAAP / IFRS]. Organise by audit area: Cash and equivalents, Accounts receivable, Inventory, Property and equipment, Other assets, Accounts payable, Accruals, Debt, Equity, Revenue, Cost of sales, Operating expenses, Payroll, Tax, Related parties, Subsequent events. For each area, list the documents and reports typically requested. Format as a markdown table with columns: Area, Document, Format expected, Period. Be specific — say 'GL detail for cash accounts' not 'cash records'."
The result is a 80 to 120 line PBC list that you customise to your business. Use it as the starting point for your auditor kickoff meeting.
Walkthrough Rehearsal
The audit walkthrough is the conversation where the auditor asks you to describe how a process works. The questions are predictable. The answers are where you stumble if you have not rehearsed.
Prompt:
"Act as an external audit senior. I will describe a finance process. Ask me the questions you would ask in a walkthrough to test whether controls operate effectively. After each of my answers, ask the natural follow-up. Continue until you have established: (a) what initiates the process, (b) who performs each step, (c) what the control objectives are, (d) what evidence is retained, (e) what happens if exceptions are identified, (f) how segregation of duties is enforced, (g) how the control was operating during the year, (h) any changes in the period.
The process I want to walk through is: [e.g., monthly bank reconciliation for the operating account]. I will respond as the control owner."
Now rehearse the conversation. AI plays the auditor, you play the controller. By the end of 15 minutes you will have surfaced the gaps in your explanation — usually two or three points where you said "we always do it" but cannot point to evidence. Fix those before the real walkthrough.
This is the single highest-ROI use of AI for first-time audit clients. Practise three processes this way and your kickoff goes from stressful to routine.
Drafting Memos on Judgemental Areas
For revenue recognition decisions, accrual estimates, impairment assessments, and going concern memos, AI drafts the structure and the language. You provide the facts and the judgement.
Prompt for a revenue recognition memo:
"Draft a revenue recognition memo for the following contract under [ASC 606 / IFRS 15]. Facts: [contract description, performance obligations, transaction price, allocation method, point of recognition, any variable consideration]. Walk through the five steps explicitly. State the conclusion. End with a section on judgements and assumptions. Tone: audit-defensible, no marketing language. Include explicit references to standard sections where relevant."
The result is a 600 to 800 word memo with the five-step structure. You verify the facts, sharpen the judgement section, and sign off. Often this would have been a 3-hour writing task; AI compresses it to 45 minutes.
Where AI Must Not Be Used
Three hard limits in audit work.
Limit 1 — Sample selection. Audit samples must be selected deterministically with documented rationale. Do not ask AI to "pick a sample". Use statistical software or your audit firm's sampling tool.
Limit 2 — Conclusions on materiality and risk. AI does not know your firm's materiality calculation or your specific risk environment. State the conclusion yourself; AI drafts the explanation.
Limit 3 — Anything related to fraud risk assessment. This is human judgement supported by tools. AI is not a fraud detector and should not be presented as one.
Beyond these limits, AI is now a normal part of audit prep. Many audit firms themselves use it for drafting their reports. Your auditor will increasingly expect to see it used responsibly on your side.
A Compressed Audit Prep Calendar
Here is what a typical audit prep cycle looks like with AI integrated.
Week 1: Generate the PBC list with AI. Walk through it with the auditor. Assign owners.
Week 2: Pull source documents. AI drafts memos for each judgemental area.
Week 3: AI drafts audit trail memos for each control to be tested. Walkthrough rehearsals begin.
Week 4: Walkthroughs with auditor. Sample selection (manual). Sample support gathered.
Week 5: Auditor fieldwork. Day-of questions answered with AI-assisted drafting.
Week 6: Audit clearance.
What used to be 8 to 10 weeks for first-time audit clients is now 5 to 6 weeks. The reason is not AI doing your audit; it is AI compressing the writing and rehearsal portions of the prep work.
Key Takeaways
- AI drafts audit trail memos in 5 minutes — save one prompt and reuse for every control
- Generate your PBC list from scratch with AI calibrated to your industry and framework
- Rehearse walkthroughs with AI playing the auditor — surfaces gaps before the real session
- For judgemental memos, AI drafts the structure; you provide the facts and judgement
- Hard limits: AI must not select samples, conclude on materiality, or run fraud risk assessment