The prompts below are written for accountants and bookkeepers using ChatGPT or a similar chat assistant. Each one has placeholders in square brackets, and none of them needs a client’s name, tax ID or account number to work. The output is a draft. The accountant who uses it is responsible for checking every figure and every rule it mentions.
Settings were checked on November 9, 2026.
Before the first prompt: three settings
Don’t paste this in
Tax return information, SSNs, EINs, bank details or anything else that identifies a client. Return preparers who disclose return information without an exception or signed consent risk penalties under Section 7216. On personal ChatGPT Free, Plus and Pro accounts, OpenAI’s data controls page says data sharing for model training is on by default. Even with it switched off, a conversation that gets a thumbs up or thumbs down may be used for training.
- Account type. OpenAI says it does not train on ChatGPT Business, Enterprise or Edu workspace content by default. A firm-managed workspace is the safer home for anything work-related.
- Training toggle. On a personal account, Settings, then Data controls, then “Improve the model for everyone” controls training use.
- Temporary Chat. These chats aren’t used for training and don’t appear in history, but OpenAI says they may be kept for up to 30 days for safety purposes.
Client communication
1. Plain-English explanation of a tax change
Explain [TAX RULE OR CHANGE] to a small-business owner with no accounting background.
Use under 150 words, short sentences and one concrete example with round numbers.
End with one question they should ask their accountant about their own situation.
Do not give advice specific to any person.2. Missing-documents reminder
Write a polite second reminder email to a client who hasn't sent [DOCUMENTS] for [ENGAGEMENT].
Deadline: [DATE]. Consequence of missing it: [e.g. extension needed].
Tone: friendly but clear. Under 120 words. Include a bulleted list of the missing items.3. Fee increase letter
Draft a short letter telling existing clients that fees for [SERVICE] will change from [OLD] to [NEW] from [DATE].
Reasons to mention: [REASONS].
Keep it factual, avoid apologizing more than once, and offer a call for questions.Month-end and year-end close
4. Close checklist from scratch
Create a month-end close checklist for a business using [SOFTWARE] with [NUMBER] bank accounts, [NUMBER] credit cards, payroll through [PROVIDER] and sales tax in [STATES].
Columns: step, owner role, business day due, evidence for the reviewer.5. Variance commentary draft
Here are account balances for two periods with the entity name removed:
[ACCOUNT | PRIOR | CURRENT]
For changes above [THRESHOLD], write one neutral sentence describing the movement and list possible causes to confirm. Do not present any cause as fact.6. Accrual review questions
I'm reviewing year-end accruals for a [INDUSTRY] company. List the accrual areas commonly missed in this industry, and for each one the question I should ask management and the document that would support the answer.Excel and data
7. Formula builder
Columns: [HEADERS AND LETTERS]. I need a formula in [CELL] that [GOAL].
Explain each part in one line and give two test rows with expected results.8. Data cleanup plan
I exported a general ledger detail report with these problems: [e.g. merged cells, dates stored as text, amounts with trailing minus signs].
Give me step-by-step instructions to clean it in Excel (Power Query or formulas), in the order I should do them.9. Reconciliation exception grouping
Unreconciled items after a bank reconciliation (no account numbers):
[DATE | DESCRIPTION | AMOUNT | BOOK OR BANK]
Group by likely type (timing, missing entry, duplicate, unknown) and write a one-line next step for each group.Research and learning
10. Research question map
Question: [RESEARCH QUESTION].
List the issues to consider, the code sections or standards that may apply (label them "to verify"), and the facts I need from the client. Do not quote rule text.Anything labeled “to verify” means exactly that. Chat assistants can name the wrong section or describe one that doesn’t exist.
11. Explain a standard to a junior
Explain [STANDARD OR CONCEPT, e.g. revenue recognition for multi-year software contracts] to a first-year staff accountant.
Use a worked example with round numbers and a journal entry for each step.
Then list three common mistakes and how a reviewer would spot them.12. Quiz for training
Write 8 multiple-choice questions to test a staff accountant on [TOPIC].
Give the answer and a one-sentence explanation for each.
Mark any question where the answer depends on recent rule changes so I can check it against the source.Making the prompts work better
A few habits improve results across all twelve:
| Habit | Why it helps |
|---|---|
| State the audience (client, partner, junior) | Sets vocabulary and length |
| Give a word limit | Stops padding |
| Ask for questions back | Exposes what the tool is guessing |
| Ask it to label uncertain points | Makes review faster |
| Keep identifiers out from the start | Easier than cleaning up afterward |
Turning the list into a firm prompt library
Firms that use these regularly tend to keep them in one shared document rather than in individual chat histories. A simple structure works:
- Name and task. For example, “Variance commentary, monthly clients.”
- Approved tool and account. Which workspace the prompt is cleared for.
- Allowed inputs. What can be pasted (anonymized balances, procedures) and what can’t.
- Reviewer step. Who checks the output and what they look for.
- Last edited. Prompts drift as staff tweak them, so a date and an owner help.
Keeping the library short is better than keeping it complete. Ten prompts that staff actually use, each with its data rule written next to it, does more for consistency than a long list nobody opens.
For a wider look at where AI fits in a practice, including Section 7216 consent, the FTC Safeguards Rule and AICPA confidentiality guidance, see the full workflow guide for accountants on this site.