Reconcile Transactions Across Accounts

Introduction

Month-end reconciliation is the same story every cycle: export three systems, paste into a spreadsheet, and hunt for the one row that does not match. Dume Cowork can automate the heavy lifting — normalizing fields, matching transactions, and producing a clean exception list so your team spends time on judgment, not row scanning.

How It Works

Dume Cowork reads files from your machine (CSV, XLSX) or pulls exports you’ve saved from banking and payment tools. It then:

1. Normalizes columns — dates, currencies, fees, and reference IDs
2. Defines match keys — e.g. date + amount + last 4 of reference
3. Scores matches — exact vs fuzzy (for timing differences across time zones)
4. Surfaces exceptions — missing, duplicate, partial, or amount-off entries
5. Outputs artifacts — reconciled table + grouped exceptions + optional summary for leadership

Step 1 — Gather exports for the same period

Use the same date range and timezone assumptions across sources. If your bank uses settlement dates and Stripe uses UTC, say so in your prompt — Dume can apply offset rules.

Step 2 — Describe your matching rules

Example: “Match bank line items to ledger by external_id when present; otherwise match by amount within ±$0.01 and date within 2 business days.”

Step 3 — Run reconciliation and review exceptions

Start with the exception report — usually 5–20% of rows drive 80% of the work.

Step 4 — Document recurring patterns

If the same mismatch appears every month (fees, FX, timing), capture a rule Dume can reuse next close.

What Dume Reconciles

• Bank vs general ledger — deposits, withdrawals, fees
• Processor payouts vs invoices — gross, net, and fee lines
• Internal transfers — between accounts that appear twice
• Duplicates — same payment recorded in two systems
• Categorization — suggest GL codes for uncategorized rows (optional)

Example Prompt

Reconcile transactions for March 2025.

Inputs:
1) bank_statement_march.csv (columns: date, description, amount, balance)
2) ledger_march.csv (columns: date, memo, amount, ref_id)
3) stripe_payouts_march.csv (columns: arrival_date, amount, payout_id)

Rules:
- Primary match: ref_id when present in memo
- Else: match amount (absolute) + date within 2 calendar days
- Bank fees: allow ledger line "Stripe fees" to match net difference on payout batches

Deliverables:
A) reconciled_master.csv with columns: source, matched_group_id, status
B) exceptions.md grouped by: MISSING_IN_LEDGER, MISSING_IN_BANK, AMOUNT_MISMATCH, DUPLICATE
C) short executive summary: total matched %, top 3 root causes

Example Exception Summary (Shape)

Reconciliation — March 2025

Matched: 1,842 / 1,901 (96.9%)

Top exceptions:
• 38 rows — Stripe payout arrived April 1 (timezone / settlement lag)
• 12 rows — Bank fees not booked in ledger
• 9 rows — Duplicate Stripe charge IDs (needs manual merge)

Customization

• "Treat amounts in cents internally; display in dollars"
• "Ignore rows under $1.00 for matching"
• "Always flag FX transactions for manual review"
• "Output a pivot by vendor for the CFO summary"

Tips for Best Results

• Keep raw exports — avoid pre-editing columns so Dume can infer mappings
• Name files with period and source (bank_2025-03.csv)
• If you use multiple entities, separate files per entity before reconciling
• Re-run the same prompt next month with updated files — rules carry over in your session notes

Limitations

Dume helps with structure, matching logic, and reporting. Final sign-off on financial statements and tax treatment remains with your finance team and advisors. Always validate material exceptions before publishing numbers.