Prompt Engineering for Auditors & Accountants

Prompt engineering is a key consideration for accounting professionals to get the most from their use of AI. Crafting the right prompts for each task can seem daunting, especially when having experienced less than ideal responses from AI Large Language Models in the past. However, there are considerations both within and beyond prompts themselves that can assist auditors and accountants in accomplishing more with ChatGPT and other generative AI tools.

Before the Prompt: Custom Instructions in ChatGPT

Before diving into prompt engineering more thoroughly, it's worth exploring a lesser-known feature of OpenAI's platform that isn't often considered in discussions around prompt engineering. Within the user's profile at the bottom of the ChatGPT homepage is an option to customize ChatGPT to the user.

This feature can be incredibly valuable for users before ever considering a formal prompt. Users will see suggestions that OpenAI has included for each of the inputs—both about you as a user and how you prefer it to respond. As expected with ChatGPT, these fields are open-ended and can be experimented with for optimal results.

Setting Up Custom Instructions

For accounting use cases, consider letting ChatGPT know that you're an auditor/accountant and are looking for concise and objective answers. Here's an example of how custom instructions can dramatically improve responses:

Example Scenario: A client asks about lease accounting under ASC 842.

Without Custom Instructions:

• Response: 550+ words
• Includes irrelevant information like IFRS considerations
• Too lengthy for client communication

With Custom Instructions:

• Response: ~400 words
• Focused on US GAAP
• Concise and client-appropriate

The custom instructions used in this example included basic background information about being a US-based accounting professional who needs concise, practical responses for client communications.

Core Prompt Engineering Principles

1. Rethinking Prompt Specificity

The common concern is: "If I spend lots of time writing a very specific prompt, I might as well have looked up and written out the answer myself."

Solution: Think about specificity in terms of context, not length.

Instead of: "Provide my client with an overview of lease accounting"

Try: "Provide my client steps to account for a finance lease of an office building under ASC 842"

Without adding much effort to writing the prompt, the focus shifts to information most relevant to the client.

2. Providing Essential Context

Context is crucial for accurate responses. Key contextual elements to include:

• Geographic Location: Mentioning "US" steers ChatGPT toward US GAAP
• Industry: Many accounting standards are industry-specific
• Entity Type: Public vs. private company considerations
• Specific Standards: Reference ASC sections when relevant

Example of Missing Context: Without specifying "US," ChatGPT might discuss IFRS convergence and other irrelevant international considerations.

3. Using Proper Accounting Terminology

Using correct accounting terminology and citing specific standards ensures the AI can accurately interpret requests and provide relevant information.

Good Example: "Can you provide a journal entry example for recording a revenue transaction that meets the criteria of ASC 606, focusing on the allocation of transaction price to performance obligations for a software service contract?"

Advanced Tip: You can ask ChatGPT to cite specific sections of the Accounting Standards Codification (ASC) for verification purposes.

4. Focus on Objectives and Outcomes

When stating objectives clearly and eliciting objective answers from AI, the outputs will often be more actionable in an accounting context.

Microsoft Copilot Advantage: Copilot's interface provides toggleable inputs for:

• Tone (professional, casual, etc.)
• Format (bullet points, paragraphs, etc.)
• Length (brief, detailed, etc.)

These considerations should be built into ChatGPT prompts when the interface doesn't provide them automatically.

5. Iteration and Refinement

Think of iterating with LLMs as comparable to training new team members. While LLMs have vast knowledge bases, they need guidance to arrive at desired outcomes.

Best Practices for Iteration:

• Use ChatGPT's memory of previous prompts in the conversation
• Build on responses to refine accuracy
• Create Custom GPTs to retain successful prompt styles
• Be patient with the refinement process

Practical Applications

Document Analysis

ChatGPT and other LLMs can analyze uploaded documents. This capability is worth exploring for:

• Contract review
• Financial statement analysis
• Compliance documentation review

Audit Planning

Use LLMs to help with:

• Risk assessment procedures
• Audit program development
• Client communication drafting

Looking Forward

Generative AI's ability to transform how auditors and accountants approach their work is only beginning to be explored. As AI evolves beyond simple text inputs and outputs into more substantial tasks involving:

• Document analysis and processing
• Complex audit procedures
• Bespoke AI agents for specific accounting tasks

It will be important to keep these prompt engineering principles in mind to continue generating optimal results from accounting professionals' use of AI.

Key Takeaways

1. Set up custom instructions before crafting individual prompts
2. Focus on context over length when being specific
3. Use proper accounting terminology and cite relevant standards
4. Be clear about objectives and desired outcomes
5. Iterate and refine your prompts for better results
6. Explore document analysis capabilities for more advanced use cases

The investment in learning prompt engineering will pay dividends as AI becomes increasingly integrated into accounting and auditing workflows.