Reconciliation and Error Detection

🔄 Quick Recall: In the last lesson, you learned to generate financial reports and narrative summaries with AI. But reports are only trustworthy if the underlying data is correct. This lesson ensures your data is right before it reaches any report.

Why Reconciliation Eats Your Time

Reconciliation is the accounting equivalent of proofreading — essential, tedious, and easy to rush through when you’re behind. You’re comparing two sets of records, line by line, looking for things that don’t match.

Bank reconciliation alone takes the average accountant 30-60 minutes per account per month. Multiply that by 5 clients with 3 accounts each, and you’ve burned a full day just verifying that the numbers agree.

AI can’t replace your judgment on why something doesn’t match. But it can instantly find the items that need your attention instead of making you scan every line.

Bank Reconciliation with AI

Here’s the core reconciliation prompt. You provide both datasets and AI finds the gaps:

Reconcile these two datasets and identify all discrepancies:

BANK STATEMENT (First National, January 2026):
Date | Description | Amount | Balance
01/02 | Opening Balance | — | $24,567.89
01/03 | DEP - CLIENT PAYMENT | +$8,500.00 | $33,067.89
01/05 | CHECK #1042 | -$2,400.00 | $30,667.89
01/07 | ACH - GUSTO PAYROLL | -$12,250.00 | $18,417.89
01/10 | DEP - CLIENT PAYMENT | +$5,000.00 | $23,417.89
01/12 | WIRE OUT - VENDOR | -$3,200.00 | $20,217.89
01/15 | CHECK #1043 | -$875.00 | $19,342.89
01/22 | DEP - CLIENT PAYMENT | +$12,000.00 | $31,342.89
01/31 | BANK FEE | -$45.00 | $31,297.89

GENERAL LEDGER (Cash Account 1000, January 2026):
Date | Description | Debit | Credit | Balance
01/01 | Beginning Balance | — | — | $24,567.89
01/03 | Invoice #1087 - Client A | $8,500 | — | $33,067.89
01/05 | Check #1042 - Office Rent | — | $2,400 | $30,667.89
01/07 | Payroll - Jan 1-15 | — | $12,250 | $18,417.89
01/10 | Invoice #1092 - Client B | $5,500 | — | $23,917.89
01/12 | Vendor - Web Design Co | — | $3,200 | $20,717.89
01/15 | Check #1043 - Insurance | — | $875 | $19,842.89
01/20 | Invoice #1095 - Client C | $12,000 | — | $31,842.89

IDENTIFY:
1. Transactions in the bank but not in the GL
2. Transactions in the GL but not yet cleared at the bank
3. Amount discrepancies between matching transactions
4. Running balance differences and where they start
5. Reconciling items (timing differences vs. errors)

AI will catch: The $500 discrepancy on the January 10 deposit ($5,000 bank vs. $5,500 GL), the January 22 vs. January 20 date mismatch on Client C’s payment, and the missing bank fee in the GL. These are exactly the items you’d spend twenty minutes scanning for manually.

✅ Quick Check: Why does AI need both the bank statement and general ledger to perform a reconciliation?

It needs both sides of the comparison. The bank shows what actually moved through the account; the GL shows what your books recorded. Discrepancies only become visible when you put them side by side.

Detecting Common Errors

Certain errors show up constantly. Build a prompt that specifically hunts for them:

Analyze this transaction list for common bookkeeping errors:

[Paste your transaction list]

CHECK FOR:
1. DUPLICATE ENTRIES — Same vendor, same amount, same date or within 3 days
2. TRANSPOSITION ERRORS — Amounts where digits might be swapped (e.g., $1,350 vs $1,530 — difference divisible by 9)
3. ROUND NUMBER ANOMALIES — Exact round numbers ($1,000, $5,000) that should have cents
4. MISSING SEQUENCES — Gaps in check numbers or invoice numbers
5. SIGN ERRORS — Debits recorded as credits or vice versa
6. PERIOD CUTOFF — Transactions dated in the wrong month

For each issue found, explain what it is and suggest the likely correction.

The transposition error check is especially powerful. If you’ve ever spent an hour hunting for why your trial balance is off by $180 — a difference divisible by 9 — you know the frustration. AI catches these in seconds.

Unusual Transaction Detection

Beyond matching errors, AI can flag transactions that break patterns:

Review these 6 months of expense transactions and identify anything unusual:

[Paste 6 months of categorized transactions]

FLAG:
1. Transactions significantly larger than the category average (more than 2x typical)
2. New vendors not seen in previous months
3. Round-number payments to individuals (potential unreported compensation)
4. Expenses that spiked more than 30% vs. the prior month with no obvious reason
5. Weekend or holiday transactions in categories that normally occur on business days
6. Vendors receiving payments at increasing frequency

For each flag, rate the concern level (Low/Medium/High) and suggest what to investigate.

This isn’t fraud detection — that requires specialized tools and professional skepticism. But it catches the anomalies that deserve a second look. Maybe that $4,000 “office supplies” charge was a legitimate bulk purchase. Maybe it wasn’t. AI surfaces it; you decide.

Accounts Receivable Reconciliation

A/R reconciliation is about making sure what clients owe matches what your books say. Use AI to analyze aging:

Analyze this accounts receivable aging report and identify concerns:

A/R AGING - January 31, 2026:
Client | Current | 1-30 | 31-60 | 61-90 | 90+ | Total
Client A | $8,500 | $0 | $0 | $0 | $0 | $8,500
Client B | $5,200 | $3,800 | $0 | $0 | $0 | $9,000
Client C | $0 | $0 | $4,200 | $6,100 | $2,800 | $13,100
Client D | $12,000 | $0 | $0 | $0 | $0 | $12,000
Client E | $3,500 | $2,200 | $1,800 | $0 | $0 | $7,500

PROVIDE:
1. Total A/R and aging distribution percentages
2. Clients requiring immediate collection action
3. Recommended allowance for doubtful accounts based on aging
4. Draft collection priority list with suggested actions per client
5. Cash flow impact if 90+ day balances become uncollectible

AI calculates the aging percentages, identifies Client C as a collection priority, and drafts an action plan. You’d reach the same conclusions, but AI organizes it in minutes instead of thirty minutes of analysis and writing.

Building a Reconciliation Checklist

For each client, create a standardized reconciliation checklist AI can work through:

Create a month-end reconciliation checklist for a small business with these accounts:

- 3 bank accounts (operating, savings, payroll)
- 1 credit card
- Accounts receivable
- Accounts payable
- Payroll tax liabilities

For each account, list:
1. What to reconcile against (bank statement, subledger, tax deposit records)
2. Common discrepancies to check for
3. The specific data I need to prepare
4. A prompt template I can use each month to run the reconciliation

Format as a step-by-step workflow I can follow every month-end.

Save this checklist. It becomes your firm’s standardized reconciliation procedure — consistent across clients, easy to delegate, and faster to execute with AI assistance.

Exercise: Reconcile a Sample Account

Practice with this scenario:

Your client’s checkbook register shows an ending balance of $15,430.22. The bank statement shows $16,205.22. You know:

• Check #2048 for $1,200 hasn’t cleared
• A $425 deposit was made on January 31 but appears on the February bank statement
• The bank charged a $50 service fee your client didn’t record

Ask AI to perform the reconciliation, identify the adjusted book balance, the adjusted bank balance, and explain each reconciling item. Then verify the result yourself — do the adjusted balances agree?

Key Takeaways

• Provide AI with both sides of any reconciliation — bank data and book data — to identify discrepancies automatically
• AI catches mechanical errors (duplicates, transpositions, missing items) faster than manual scanning
• Unusual transaction detection is an alert system, not a verdict — always investigate flagged items with source documents
• Build standardized reconciliation checklists with embedded prompt templates for consistency across clients
• A/R aging analysis with AI produces both the numbers and the collection action plan in one step
• The goal isn’t AI doing your reconciliation — it’s AI finding the items that need your attention

Up Next: In the next lesson, we’ll tackle tax research and compliance — using AI to research tax rules, identify deductions, and check regulatory requirements.