ChatGPT for Accountants: 9 Practical Prompts to Boost Your Productivity

You already know what ChatGPT is. The question that matters is whether it can help you complete your work faster without introducing errors or compromising client confidentiality. The short answer: yes, if you know how to ask.

The difference between a vague, unhelpful response and a genuinely useful one comes down to how you structure your prompt. Most accountants who try ChatGPT once and dismiss it simply ask it the wrong question. This article gives you nine prompts you can copy, adapt, and use today, along with the principles behind why they work, so you can start building your own.

Why Prompt Structure Matters More Than You Think

Typing "help me with month-end" into ChatGPT will get you a generic overview of what month-end involves. That is not useful. What works is giving the model a role, context, and a specific output format. This technique, known in the AI world as prompt engineering, is rapidly becoming a core skill for finance professionals.

Think of it like briefing a capable but context-blind junior. You would not hand them a trial balance and say, "Do something with this." You would tell them who they are working for, what the deliverable looks like, and what constraints apply.

Every prompt in this article follows a simple formula:

Role + Context + Task + Output format + Constraints

Once you internalise this pattern, you will find yourself writing better prompts instinctively. Let's get into the ten.

1. Draft a Client Engagement Letter

Writing engagement letters is necessary but repetitive. ChatGPT handles the first draft well, freeing you to focus on the specifics.

Prompt:

You are a senior accountant at a South African audit firm. Draft a professional engagement letter for a new compilation engagement under ISRS 4410 (Revised) for a private company (Pty) Ltd. The letter should cover the scope of services, management's responsibilities, our responsibilities, limitations of the engagement, fees (leave as a placeholder), and a signature block. Use a professional but approachable tone. Format the output as a Word-ready document with clear headings.

Pro Tip:

Upload ISRS 4410 (Revised) to the prompt to give it the standard to read.

Why it works:

Specifying the standard (ISRS 4410), the entity type, and the South African context means you get a draft that needs minor tailoring rather than a complete rewrite.

Important:

Always review the output against the relevant IRBA guidance and your firm's own templates. AI gives you speed; your professional judgement gives you accuracy.

2. Explain a Complex Accounting Standard in Plain English

When you need to brief a client, a director, or even a trainee on what a standard means in practice, ChatGPT can translate the technical language into something accessible.

Prompt:

You are a training expert. Explain IFRS 16 (Leases) to a non-financial business owner who currently leases office space and two delivery vehicles. Focus on how the standard affects their balance sheet and what they should expect from their accountant at year-end. Avoid jargon. Keep it under 400 words.

Why it works:

Specifying the audience (non-financial), providing concrete context (office and vehicles), and setting a word limit prevent the model from producing an academic essay no one asked for.

3. Build a Month-End Close Checklist

Every firm has its own month-end routine, but ChatGPT can generate a comprehensive starting point that you refine over time.

Prompt:

You are a process and workflow documentation expert. Create a month-end close checklist for a South African SME with an annual turnover under R20 million. The business is a trading entity that carries inventory, operates a fleet of delivery vehicles, and employs 15 people. Include bank reconciliations, VAT reconciliation to the next SARS submission, payroll reconciliation, fixed asset register review, inventory count procedures, and any provisions or accruals typically required. Group items by category and indicate which items are preparer responsibilities and which are reviewer responsibilities.

Why it works:

The SA-specific detail (VAT, SARS, Rand threshold) grounds the output in your jurisdiction. Adding the preparer/reviewer split makes the checklist immediately usable in a team environment.

4. Analyse Financial Statement Ratios

Rather than manually calculating and interpreting ratios, paste a set of financial statements into ChatGPT and let it perform the initial analysis.

Prompt: You are a financial analyst. I will provide a summarised income statement and balance sheet for a South African retail company. Calculate the following ratios: current ratio, quick ratio, debt-to-equity, gross profit margin, net profit margin, return on equity, and inventory turnover. Present the results in a table. Below the table, provide a brief narrative (200 words max) highlighting any areas of concern or strength. Flag anything that would warrant further investigation during an audit.

Then paste your (anonymised) data below the prompt.

Why it works:

Specifying the exact ratios avoids an overwhelming data dump. Asking for a narrative plus a table gives you something you could drop into a working paper or management letter with minimal editing.

Data caution:

Never paste identifiable client information into any public AI tool. Anonymise all data first. Under South Africa's Protection of Personal Information Act (POPIA), you are the responsible party for any personal data you process, including data you feed into third-party AI platforms.

5. Summarise a Long Regulatory Document

Tax Acts, SARS interpretations, IRBA circulars: the reading list never ends. ChatGPT can condense lengthy documents into actionable summaries.

Prompt: I am going to paste the text of a SARS binding general ruling. Summarise it in the following format: (1) What is the ruling about, in one sentence? (2) Who does it affect? (3) What is the effective date? (4) What are the key practical implications for a tax practitioner advising SME clients? (5) Are there any transitional provisions or deadlines to note? Keep the summary under 500 words.

Then paste the ruling text.

Why it works: The structured format ensures consistent output that is easy to scan. Asking specifically about SME implications filters out information relevant only to large corporates or specific industries.

Verification note: Always cross-check summaries against the original SARS publication. Large language models can misinterpret technical provisions, particularly around effective dates and thresholds.

6. Write a Professional Email to a Difficult Client

We have all had the client who is three months late with their records and still expects the annual financial statements by Friday. ChatGPT helps you draft a firm, professional response without the passive aggression that creeps in at 10 pm.

Prompt: Draft a professional but firm email to a client who has repeatedly missed deadlines for providing supporting documents needed to complete their annual financial statements. This is the third reminder. The tone should be empathetic but clear about the consequences: specifically, that continued delays may result in late filing with CIPC and potential penalties. Include a specific deadline (two weeks from today) and a short list of the outstanding items (leave as placeholders for me to fill in). Keep the email under 250 words.

Why it works: Setting the tone ("empathetic but clear"), providing the scenario context, and specifying the consequences means you get a usable draft rather than something either too aggressive or too soft.

7. Prepare Interview Questions for a Trainee Accountant

Hiring trainees? ChatGPT can generate situationally relevant interview questions far more efficiently than starting from scratch.

Prompt: I am a partner at a small audit firm in South Africa (14 people, mainly audit and tax work). We are interviewing SAICA trainee accountant candidates who have completed their CTA. Generate 10 interview questions that test: (a) technical competence in auditing basics and financial reporting, (b) problem-solving under pressure, (c) communication skills, and (d) cultural fit for a small, technology-forward firm. For each question, include a brief note on what a strong answer would cover. Avoid generic HR questions like "Where do you see yourself in five years?"

Why it works: The firm size, SAICA pathway, and values context (technology-forward) ensure the questions are tailored rather than pulled from a generic HR database.

8. Generate a VAT Reconciliation Template Explanation

VAT reconciliation is a recurring pain point, especially when discrepancies exist between the accounting system and the VAT 201 return. Use ChatGPT to create training material or a client-ready explanation.

Prompt:

Write a clear, step-by-step explanation of how to perform a VAT reconciliation in South Africa, aimed at a bookkeeper who is competent with accounting software but has never performed a formal VAT reconciliation before. Cover: reconciling output VAT per the income statement to the VAT 201, reconciling input VAT per the expense accounts to the VAT 201, common causes of differences (timing, exempt supplies, zero-rated items, capital goods), and how to document the reconciliation for audit purposes. Use practical examples with Rand amounts.

Why it works:

Pitching the explanation at the right level (competent bookkeeper, not complete beginner) avoids both condescension and jargon overload. Asking for Rand-denominated examples makes the explanation immediately relatable.

9. Create a Risk Assessment Brainstorm for Audit Planning

During audit planning, the engagement team needs to identify risks of material misstatement. ChatGPT is a useful brainstorming partner here, provided you treat its output as a starting point for discussion, not a substitute for professional scepticism.

Prompt:

You are an experienced auditor. I am planning an audit of a mid-sized South African construction company (annual revenue of R150 million, 80 employees; projects range from R5 million to R40 million). The company recognises revenue over time using the percentage-of-completion method under IFRS 15. Brainstorm the top 10 risks of material misstatement I should consider in my audit plan. For each risk, briefly explain why it is a risk for this specific industry and entity, and suggest one audit procedure that would address it. Consider both inherent risks and control risks.

Why it works:

Industry-specific detail (construction, percentage-of-completion, project size) produces far more relevant risks than a generic request. Asking for both the risk and a corresponding procedure gives you a starting point to include in your planning documentation.

How to Use AI Responsibly in Your Practice

These prompts can save hours every week. But speed without safeguards is a liability. Keep these principles front of mind.

Never input identifiable client data into public AI tools. ChatGPT's standard terms allow OpenAI to use inputs for model training unless you specifically opt out or use an enterprise plan. Under POPIA, transferring personal information to a third party (especially one processing data outside South Africa) requires appropriate safeguards. Anonymise everything or use an enterprise deployment with data processing agreements in place.
Always verify the output. Large language models produce confident-sounding text regardless of whether the content is correct. This is especially dangerous with tax thresholds, filing deadlines, and regulatory references. Cross-check every factual claim against primary sources like SARS, IRBA, and CIPC.
Use AI as a first draft, not a final product. The prompts above are designed to give you a 70-80% complete output that you refine with your professional expertise. The value is in skipping the blank-page problem, not in outsourcing your judgment.
Stay current on professional guidance. SAICA has signalled that it anticipates rapid growth in AI adoption within the profession and is actively working on frameworks for responsible use. Keep an eye on developments from both SAICA and IRBA as formal guidance evolves.

Where to From Here

The accountants who will thrive over the next few years are not the ones who avoid AI. They are the ones who learn to use it as a force multiplier for their existing expertise. These ten prompts are a starting point. Once you get comfortable with the role-context-task-format-constraints formula, you will start writing prompts tailored to your own workflows, your own clients, and your own practice.

The real productivity unlock is not any single prompt. It is developing the habit of asking: "Could AI give me a useful first draft of this?" Often, the answer is yes.

Want to learn more about how BVA is integrating AI into its audit and advisory practice? Contact us to discuss how technology can work harder for your business.

Disclaimer: This article is for general informational purposes only and does not constitute professional audit, tax, or legal advice. Always verify AI-generated outputs against authoritative sources and apply your own professional judgement. For guidance specific to your circumstances, please consult a qualified professional.