ChatGPT Prompts Every Bookkeeper Should Save (Updated 2026)
Published: May 6, 2026 | Category: AI Tools | By Qualora Career Advisors
Most "ChatGPT prompts for bookkeepers" articles are listicles of 50 prompts that nobody actually uses. The prompts are too generic, too long, missing the verification step, and don't address the data-handling discipline that working bookkeepers need.
This article is the practical version: 12 prompts that working bookkeepers actually use, organized by daily-workflow category, with the sanitization rules baked in and the verification checklist for each.
The prompts work in ChatGPT (Plus, Team, or Enterprise) and Claude (Pro or Work) equivalently — pick whichever you have access to. Microsoft Copilot for M365 also handles most of these patterns within the M365 environment. The prompt language is functionally identical across these tools.
A note on data handling before we start: never paste a client's actual business name, EIN, full account numbers, full SSNs, or other identifying data into a public LLM. The prompts below all use generic placeholders. Replace
[business type][amount]Category 1: Categorization Reasoning
Prompt 1: Account A vs Account B distinction
When to use: A transaction where two categories both arguably fit, and you want a structured way to think through the distinction.
I'm trying to decide whether a transaction belongs in [ACCOUNT A] or [ACCOUNT B]. The transaction is: - Vendor: [GENERIC VENDOR DESCRIPTION] - Amount: $[AMOUNT] - Context: [BRIEF, e.g., "single-location restaurant"] What's the substantive difference between [ACCOUNT A] and [ACCOUNT B] for tax and reporting purposes? When would each one be the right answer? What information about the transaction would push the decision one way or the other?
Verification: AI's tax-treatment summary may be 6-18 months stale on rule specifics (especially for IRC §274 meals deductibility, which has shifted multiple times). Verify against current IRS publications before relying on tax treatment in client communications.
Prompt 2: Transaction split across multiple accounts
When to use: A single transaction (Costco run with mixed items, vendor invoice bundling services and reimbursable expenses) that should be split across categories.
I have a single transaction that needs to be split across multiple accounts. Here are the details: - Vendor: [GENERIC VENDOR] - Total amount: $[AMOUNT] - Date: [DATE] - Line items: [PASTE LINE ITEMS] The chart of accounts options that might apply: [LIST] Walk me through how to think about the split. What's the right allocation logic? What documentation should I keep on the transaction memo?
Verification: Verify the splits sum exactly to the original transaction amount (AI math errors on splits are common). Confirm each split category exists in the client's actual chart of accounts.
Prompt 3: Opaque vendor identification
When to use: A bank-feed transaction with an unclear vendor string (
POS DEBIT 04/12 *XX5839 SQ *MERCHANT C 4123876I have a bank-feed transaction with an unclear vendor identifier. Here's what I have: - Vendor string: [PASTE FROM BANK FEED] - Amount: $[AMOUNT] - Date: [DATE] - Account it hit: [DEBIT / CREDIT] Client context: [BRIEF, e.g., "solo consulting practice"] Help me think through what kinds of transactions typically show up with this vendor string format. What information could the client provide that would identify the actual vendor?
Verification: AI's pattern recognition on vendor strings is approximately right but bank-feed format conventions vary. Identify the merchant through the client or the bank, not through AI's guess.
Category 2: Client Communication
Prompt 4: The receipt-chase email
When to use: You need to email a client requesting documentation for a list of transactions.
I'm a bookkeeper writing to a client to request receipts for these transactions: [PASTE LIST: each line "Date | Vendor | Amount"] The client is [BRIEF DESCRIPTION, e.g., "a friendly small-business owner who's busy and sometimes forgets to upload receipts promptly"]. Draft an email that: - Opens warmly without being saccharine - Lists transactions in a scannable format - Tells them how to upload (via [QuickBooks app / Dext / email reply]) - Gives a deadline ([DATE]) without sounding like a dunning notice - Briefly notes why this matters - Closes professionally with an offer to help Tone: collegial. About 150-200 words.
Verification: Confirm every transaction in the list is real and the amounts match. Adjust the opening line for relationship-specific context (a personal note about something the client mentioned).
Prompt 5: The reclassification explanation
When to use: You reclassified a transaction and need to tell the client what changed and why.
I reclassified a transaction. Facts: - Original categorization: [ORIGINAL] - New categorization: [NEW] - Transaction: [VENDOR, AMOUNT, DATE] - Reason: [SPECIFIC REASON] - Tax/reporting impact: [SHORT DESCRIPTION] Draft a brief client email that: - Acknowledges the change without making it sound like a correction - Explains the reasoning in plain language - Notes the tax/reporting implication factually, not alarmingly - Reassures that this is normal cleanup, not a problem - Invites a conversation if they want more detail Tone: matter-of-fact, confident, not preachy. About 150 words.
Verification: Confirm the tax-impact statement is accurate for the client's specific entity type. AI sometimes generalizes tax impact in ways that aren't quite right.
Prompt 6: The monthly summary email
When to use: The recurring monthly email that delivers the close package.
I'm sending the monthly client email for [GENERIC CLIENT, e.g., "a single-location restaurant"]. Verified facts from the close: - Revenue: $[AMOUNT] ([UP/DOWN] [%] vs trailing average) - Key expense variances: - [CATEGORY]: $[AMOUNT] ([%] variance; cause: [SPECIFIC CAUSE]) - [CATEGORY]: $[AMOUNT] ([%] variance; cause: [SPECIFIC CAUSE]) - Net income: $[AMOUNT] - Action items needed from client: [LIST] - Heads-up for next month: [LIST] Draft a 200-300 word email with: brief greeting + one-line summary, the story of the month, action items, sign-off. Tone: friendly professional, plain language. Every figure must come from the verified facts above; do not generate or compute new figures.
Verification: Critical — every dollar amount and percentage in the draft must match the platform's reports exactly. AI cannot do reliable arithmetic; pull figures from the close and use them as fixed inputs.
Category 3: Compliance and Audit-Defense
Prompt 7: Substantiation framework walk-through
When to use: A transaction in a substantiation-required category (Meals, Travel, Vehicle, Capital Expenditures, Charitable Contributions).
I'm reviewing a transaction in a substantiation-required category. Walk me through the IRS substantiation framework for this category and the specific documentation that would make this transaction defensible. Transaction details: - Category: [CATEGORY] - Amount: $[AMOUNT] - Date: [DATE] - Documentation we have: [LIST] - Client context: [BRIEF] Walk through: 1. What does IRC / Treasury Regulations require for substantiation? 2. What's the minimum documentation that satisfies the rule? 3. What we have — is it enough? What's missing? 4. If documentation is missing, what should I ask the client for? 5. What memo on the transaction would document the audit-trail reasoning? Important: I will verify rule citations and current treatment before relying on this. Flag anything where the rule has changed recently.
Verification: Tax thresholds and substantiation rules update. Verify against current IRS publications before relying. Especially check meals deductibility percentages (have shifted between 50% and 100% in recent legislative cycles).
Prompt 8: 1099 vs W-2 worker classification
When to use: A vendor or worker payment where the classification is unclear.
I'm thinking through worker classification for someone the client is paying. Facts: - Worker description: [GENERIC, e.g., "delivery driver"] - How they're paid: [STRUCTURE] - Annual amount: $[AMOUNT] - Work structure: [SCHEDULE, EQUIPMENT, EXCLUSIVITY, INTEGRATION] - Client preference: [CLIENT'S CURRENT POSITION] Walk through the worker-classification framework. What factors determine 1099 contractor vs W-2 employee? Which factors point which way in this case? What's the risk of getting it wrong? Important: I will verify against current IRS and state guidance. Flag any state-level rules that may differ from federal.
Verification: Worker classification has both federal (IRS common-law factors) and state (often stricter ABC tests in CA, NY, MA, NJ) rules. Verify state-specific rules separately.
Prompt 9: Audit-trail memo drafting
When to use: You made a non-routine categorization decision and need to document the reasoning.
I made a non-routine categorization decision and need to write the audit-trail memo. Facts: - Transaction: [VENDOR, AMOUNT, DATE] - Final categorization: [CATEGORY] - Decision factors: [WHY THIS CATEGORY] - Information source: [RECEIPT / CLIENT EMAIL / etc.] - Date of decision: [DATE] Draft a transaction memo that: - States what was categorized and where - Documents the reasoning concisely - Cites the source of information - Notes any uncertainty or open questions - Reads as if I'd defend this in an audit three years from now Tone: factual, professional, no flourish. 2-4 sentences typically.
Verification: The memo accurately reflects what happened. Don't invent substantiation. Keep it concise — defensibility lives in specificity, not length.
Category 4: Reporting and Narrative
Prompt 10: Variance commentary
When to use: The 2-3 paragraph monthly explanation of what changed and why.
I'm drafting the monthly variance commentary. Verified facts from the close (use as fixed inputs; do not compute new figures): Revenue: $[AMOUNT] ([UP/DOWN] [%] vs trailing 3-month) Key expense variances (verified, with documented causes): - [CATEGORY]: $[AMOUNT] ([%]; cause: [SPECIFIC]) - [CATEGORY]: $[AMOUNT] ([%]; cause: [SPECIFIC]) Net income: $[AMOUNT] Watch items for next period: [LIST] Client context: [BRIEF] Draft 2-3 paragraphs that: 1. Identify the dominant story of the month 2. Explain supporting variances without burying the lead 3. Note watch items for next period briefly 4. Close with a clear-eyed read Tone: confident, factual, plain English. About 250 words.
Verification: Every dollar amount and percentage must match the platform's reports. AI cannot generate reliable arithmetic.
Prompt 11: Balance-sheet translation for non-financial clients
When to use: A client wants a plain-English explanation of their balance sheet.
A client wants a plain-English explanation of their balance sheet: Assets: - Cash on hand: $[AMOUNT] - Accounts receivable: $[AMOUNT] - Inventory: $[AMOUNT] - Equipment (net of depreciation): $[AMOUNT] Liabilities: - Accounts payable: $[AMOUNT] - Credit cards: $[AMOUNT] - Equipment loan: $[AMOUNT] Owner's equity: $[AMOUNT] Client context: [BRIEF, e.g., "small restaurant owner who finds the balance sheet confusing"] Specific question: [WHAT THEY ASKED] Draft a translation that walks through what each major section means, highlights anything notable about the current state, answers their question, and closes with a "bottom line" assessment. Tone: explanatory but not condescending. Avoid jargon. About 350 words.
Verification: Asset totals + liability totals = assets total (basic accounting identity). Stay descriptive, don't drift into financial advice ("you should refinance this loan" is out of bookkeeper scope).
Prompt 12: Year-end handoff to tax preparer
When to use: The annual handoff package to the client's tax preparer.
I'm drafting the year-end handoff for [GENERIC CLIENT]'s tax preparer. Books are closed and reconciled. Summary facts: Revenue: $[AMOUNT] Net income: $[AMOUNT] Notable items the tax preparer needs to know: [LIST EACH WITH: what, where in books, bookkeeper categorization decision, tax-treatment flag] Open items / handoff questions: [ITEMS WHERE TAX PREPARER NEEDS TO MAKE THE CALL] Documentation attached: [LIST] Client context: [ENTITY STRUCTURE] Draft a professional handoff memo (peer-to-peer tone) that: 1. Opens with summary financials 2. Itemizes notable items in materiality order 3. Lists open questions clearly 4. Confirms documentation attached 5. Closes with bookkeeper contact for follow-up About 400-600 words.
Verification: Every figure must match the year-end close. The "what I'll do" recommendations should match firm scope. Items in materiality order, not chronological.
The Sanitization Discipline (Before You Paste)
Every prompt above uses placeholders for client-identifying data. The discipline is non-negotiable when using public AI tools (free ChatGPT, Claude.ai). Replace:
- Client business name → generic descriptor ("a small pizza restaurant", "a solo consulting practice")
- Owner name → "the owner" or first name only
- EIN → "[CLIENT EIN — REDACTED]" or skip
- Account numbers → last 4 only, or skip
- Vendor names that identify the relationship → generic vendor types ("a wholesale food vendor", "a graphic designer")
- SSNs, full credit card numbers, full bank account numbers → never paste these into any AI tier
The reasoning AI provides on a sanitized prompt is just as good as on a fully-identified prompt. The identifying data adds nothing to the categorization or drafting work; it only adds confidentiality risk.
For enterprise AI tiers (ChatGPT Team / Enterprise, Claude for Work, Microsoft Copilot for M365 Enterprise), the data-handling agreements make some of these restrictions less stringent — but the conservative posture (anonymize anyway) is still the right default.
The full sanitization framework — including what changes for healthcare clients (HIPAA), what changes for litigation-active clients, what changes for government-contracting clients — is covered in the AI for Bookkeepers course Section 3.
The Verification Step (Before You Send)
For every AI-drafted output, the verification step matters:
For categorization prompts:
- Verify tax-treatment claims against current authoritative sources (IRS publications, FASB pronouncements)
- Confirm split totals sum exactly to the original amount
- Confirm categories exist in the client's actual chart of accounts
For client communication prompts:
- Verify every figure matches the platform's reports
- Adjust the opening line for relationship-specific context
- Cut tone-drift toward dunning, condescending, or apologetic
- Confirm action items are real (transactions actually need receipts; conversations are actually pending)
For compliance prompts:
- Verify the rule citations and current treatment
- Flag any rule that has changed in the last 12-18 months for fresh verification
- Confirm scope (some questions need a CPA / EA / tax attorney, not a bookkeeper)
For reporting prompts:
- Verify every dollar amount and percentage against the platform
- Confirm the narrative matches the actual documented cause from the close
- Confirm the closing read (good, mixed, concerning) is the bookkeeper's call
The verification step is what separates the bookkeeper using AI well from the bookkeeper sending AI-drafted content unreviewed. The first practice is durable. The second practice is the one that loses clients when something goes wrong.
What These Prompts Don't Do
A few things to be clear about. The prompts above do not:
- Replace your judgment. Every output is a draft or a framework. The decisions belong to the bookkeeper.
- Generate financial figures. The prompts use figures as fixed inputs (pulled from the platform's reports); they do not ask AI to compute or estimate numbers. AI cannot do reliable arithmetic.
- Send anything. Every drafted email is reviewed and sent by the bookkeeper. AI never sends.
- Replace professional credentials. Bookkeeper-level work has clear scope boundaries. Tax advice, legal advice, audit response, and similar regulated services require credentials AI cannot provide.
The prompts compress drafting time and add reasoning structure. The accountability stays with the bookkeeper.
Where to Go Deeper
These 12 prompts cover the daily-workflow patterns. The full prompt library — 19 templates across 4 categories — is in the AI for Bookkeepers course.
The course also covers:
- The 5 core AI workflows for bookkeepers (categorization, reconciliation, receipt capture, month-end close, client communication)
- The complete sanitization and audit-trail framework
- The categories where AI is unreliable (cash, crypto, foreign currency)
- Client disclosure — when and how to tell clients you use AI
- A capstone walk-through of a complete month-end close for a fictional client
The course is built for working bookkeepers and aspiring bookkeepers who want to use AI deliberately rather than reactively. By the end, you have the workflows, the prompts, the discipline, and the integration that makes AI a productivity tool rather than a competitive threat.
If you're looking for the longer career-context read, the companion article Will AI Replace Bookkeepers? addresses the defensive question directly.
Save These Prompts
Bookmark this page. Save the 12 prompts to your firm's prompt library or your own working document. Use them as starting points; adapt the placeholders for your specific clients and chart of accounts.
The first month using these prompts will feel a little awkward — you're learning where the verification time pays off, where the tone needs adjustment, where AI's defaults match your voice and where they don't. The second month, the prompts feel native and the time savings compounds.
Six months in, the bookkeeper using prompts like these well is consistently ahead of the bookkeeper who's still typing every email from scratch. That's the leverage AI offers when used deliberately. The 12 prompts above are the starting point.