Audit season used to mean three things for me: long hours, repetitive documentation, and the creeping feeling that I was spending my most expensive hours on my least valuable tasks.
Walkthroughs. Sampling. Ticking and tying. Drafting workpaper descriptions that say the same thing every year with slightly different numbers. Formatting analytical procedure memos. Writing management letter points.
All of it necessary. All of it professional. And about 60% of it is structured, repetitive work that AI handles well.
Here’s how I use AI in audit preparation. Not to replace the professional judgment that auditing requires — but to make the preparation fast enough that I actually have time for that judgment.
Let me be direct about scope. AI does not perform audit procedures. It does not evaluate evidence. It does not form opinions. These are professional responsibilities that require a licensed auditor and cannot be delegated to technology.
What AI does:
The pattern is always the same: AI does the first draft, you do the thinking. This distinction matters for quality control and for professional standards compliance.
Walkthroughs are essential. Documenting them is tedious. You observe the process, take notes, and then spend an hour turning those notes into a proper narrative with controls identified, risks assessed, and conclusions drawn.
The prompt:
“Convert my walkthrough notes into a professional audit walkthrough memo for [PROCESS — e.g., Revenue Recognition, Accounts Payable, Payroll].
Format:
1. Process description (step-by-step narrative)
2. Key controls identified at each step
3. Control type (preventive/detective, manual/automated)
4. Segregation of duties assessment
5. Potential risk points
6. Conclusion on process design effectiveness
Client type: [INDUSTRY, SIZE]
My walkthrough notes: [PASTE RAW NOTES]”
You get a structured memo in 60 seconds. Spend 15 minutes reviewing: Does the narrative accurately reflect what you observed? Are the controls correctly identified? Are the risk points real? Adjust and finalize.
Critical: The walkthrough memo documents YOUR observations. AI structures them but cannot add observations you didn’t make. If your notes are thin, the memo will be thin. The quality of AI output is bounded by the quality of your notes.
Analytical procedures require comparing current-year data against expectations and investigating significant differences. The comparison and investigation framework is highly structured — perfect for AI.
The prompt:
“Set up analytical procedures for [AUDIT AREA — e.g., Revenue, Operating Expenses, Payroll] for [CLIENT TYPE] for the year ended [DATE].
Current year data: [PASTE TRIAL BALANCE OR ACCOUNT BALANCES]
Prior year data: [PASTE COMPARISON]
Budget data (if available): [PASTE]
For each significant account:
1. Current vs prior year ($ and % change)
2. Current vs budget (if provided)
3. Expectation: what would I expect based on known business changes?
4. Flag accounts with changes exceeding [THRESHOLD — e.g., 10% or $50,000]
5. Suggested investigation points for flagged items
6. Possible explanations to corroborate with management
Known business changes this year: [LIST — e.g., ‘acquired a competitor in Q2’, ‘lost a major customer’, ‘opened new location’]
Format as a table with a narrative summary.”
How to use it: AI generates the comparison table and flags significant items. You review the flags against what you know about the client. The “known business changes” context is critical — it helps AI generate relevant possible explanations instead of generic ones.
You still perform the actual procedure: corroborate explanations with management, obtain supporting evidence, and document your conclusion. AI did the setup; you do the audit.
Every workpaper needs a header: purpose, source of data, procedures performed, results, conclusion. You write essentially the same description with minor variations hundreds of times per audit season.
The prompt:
“Write a professional audit workpaper description for:
- Workpaper type: [RECONCILIATION / LEAD SCHEDULE / TEST OF DETAILS / SUBSTANTIVE ANALYTICAL]
- Account: [NAME AND NUMBER]
- Period: [DATE RANGE]
- What I did: [DESCRIBE IN YOUR OWN WORDS]
- What I found: [RESULTS]
- Conclusion: [YOUR CONCLUSION]
Format: Purpose, Source of Data, Procedures Performed, Results, Conclusion. Professional audit language.”
Run this for each workpaper as you complete it. AI drafts the description in 15 seconds. You review in 60 seconds. Move on.
Over a full engagement with 30-40 workpapers, this saves 30-45 minutes of pure documentation time.
When you need to select a sample for testing, AI helps with the framework — determining sample size, selection methodology, and documentation.
“I need to design a sampling plan for testing [WHAT — e.g., ‘accounts payable disbursements for proper authorization’].
Population: [SIZE AND DESCRIPTION]
Testing objective: [WHAT YOU’RE TESTING FOR]
Risk assessment: [LOW / MODERATE / HIGH]
Tolerable error rate: [PERCENTAGE]
Provide:
1. Recommended sample size with rationale
2. Selection method (random, systematic, haphazard) with rationale
3. Attributes to test for each item
4. Documentation template for exceptions
5. Evaluation framework (how many exceptions before the control fails)
Follow AICPA sampling guidance.”
Important: Verify the sample size against your firm’s methodology. Different firms use different sampling approaches and thresholds. AI gives you a reasonable framework; your firm’s audit manual gives you the specific standard.
Writing management letter points is one of the most time-consuming deliverables in an audit. Each point needs: condition, criteria, cause, effect, and recommendation. The structure is rigid and repetitive.
“Draft a management letter point for the following finding:
What I found: [DESCRIBE — e.g., ‘3 out of 25 disbursements tested lacked proper approval documentation’]
What should have happened: [CRITERIA]
Why it matters: [RISK OR IMPACT]
Format using the 5 Cs:
1. Condition (what I found)
2. Criteria (what should be)
3. Cause (why it happened, if known)
4. Consequence (the risk or impact)
5. Corrective action (recommendation)
Tone: professional, constructive, not accusatory. The goal is improvement, not blame.”
For engagements with 5-8 management letter points, this saves roughly an hour of writing time.
For larger datasets, AI can spot patterns and anomalies that support your analytical procedures.
“Analyze this journal entry data and flag anything unusual:
Flag:
1. Entries posted on weekends or holidays
2. Round-number entries over [THRESHOLD]
3. Entries posted by the same person who approved them
4. Entries just below authorization thresholds
5. Unusual account combinations
6. Entries with vague descriptions (‘adjustment’, ‘correction’, ‘miscellaneous’)
7. Duplicate amounts on the same date
[PASTE JOURNAL ENTRY DATA — anonymized]”
This doesn’t replace journal entry testing. It gives you a risk-based starting point for selecting entries to investigate. The flags become your targeted sample instead of random selection.
The boundaries are firm:
Audit data is sensitive. Apply the standard anonymization protocol with extra care:
Your firm likely has specific policies about technology use in audit. Check them before implementing any AI workflow. If your firm doesn’t have an AI policy yet, that’s a conversation worth starting.
For a typical small company audit:
Total: 8-10 hours per engagement. That’s a full day. On a 40-hour audit, that’s 20-25% of total time recovered for higher-value work.
The hours go back into the parts of auditing that actually require your expertise: evaluating evidence, exercising judgment, having difficult conversations with management, and forming your opinion.
That’s the trade-off AI enables. Less time documenting. More time thinking. Same professional standards. Better audit quality because you’re not exhausted from formatting workpapers at 11 PM.
The free PDF includes workpaper documentation prompts, analytical procedure templates, and a verification checklist designed for audit work. Download it here.
Get 50 copy-paste prompts delivered to your inbox — free.
Get the Free PDF →