By Randy Kardas, CPA, CITP, CGMA, MBA

Like many of you, I’ve been experimenting with tools like ChatGPT almost daily, eager to find genuinely impactful ways to put this technology to work. Yet, when I was recently asked to highlight a few artificial intelligence (AI) use cases that truly move the needle for accountants, I struggled. Nothing has fully lived up to the early AI hype. I couldn’t just flip a switch and have AI flawlessly reconcile general ledger accounts or produce a full set of month-end financials. This only seemed to reinforce my notion that AI will not replace accountants; however, I do believe that those who effectively leverage AI will surpass those who don’t.

I think we’ve started to climb down from the peak of AI’s inflated expectations and landed in a more practical place. Right now, AI can streamline tasks like data entry, classification, anomaly detection, pattern analysis, auto-reconciliation, systems integration, sentiment analysis, and even predictive analytics. For forward-thinking firms, these gains will be tangible. It’s likely AI will soon engage in more advanced financial analysis and, in a few years, perhaps approach real-time autonomous accrual accounting.

Early access to generative AI tools like ChatGPT confirmed that AI could articulate complex concepts with greater clarity and efficiency than I could on my own. Yet, for a long while, I didn’t feel that transformative “light bulb” moment. AI felt helpful but limited, like gaining an extra hand without knowing exactly how to use it. That changed when I faced two very specific core accounting tasks: modeling an accounting workflow and preparing journal entries. It struck me that today’s generative AI, with the right prompts, can represent accounting processes through familiar tools like T accounts, and then convert that analysis into ready-to-import journal entry templates. The process is surprisingly straightforward. With a carefully crafted prompt, you can streamline process visualization and jumpstart your journal entry preparation.

What follows will show you how to leverage ChatGPT to automate key aspects of your accounting workflows. You’ll see how to guide the AI to produce not just polished narrative explanations, but also visual workflow representations and properly formatted templates for journal entries.

T Account Analysis

Traditionally, we may have created T accounts for each journal entry, but visually representing the entire workflow helps us more easily understand accounting events over time. By laying out the events vertically (over time) and the accounts horizontally, we get a comprehensive view. Here is an example.

Instructing the AI Model

To get an AI model to replicate and manipulate a spreadsheet based on a given set of accounting events, general ledger accounts, and time periods, we need to provide clear, structured instructions. By giving the model a precise purpose, outlining restrictions, and detailing the desired output, we ensure results align closely with our accounting needs.

Purpose – Clearly articulate what you want the AI to achieve. For example, your goal might be to have the AI generate a journal entry import template (complete with all necessary debit and credit accounts and amounts) based on a series of specified accounting transactions.

Restrictions – Define any boundaries or constraints the AI must respect. Perhaps certain accounts should never appear in the output, or maybe there are formatting requirements, such as dates must follow a specific format or certain columns must always be included.

Desired Output – Specify precisely what the final product should look like. This might be a CSV file with columns for date, account, debit, credit, and description, or a spreadsheet with a particular structure and styling. The more detail provided, the more likely the AI is to produce a consistent, ready-to-use result.

Spreadsheet Template – As mentioned above, a baseline sample of the final output should show exactly how you want the finished spreadsheet to appear. This sample should include prompt wording or instructions baked into the sheet itself. By doing this, the AI can learn from a clear example and replicate it for future accounting scenarios.

This may sound like a lot of setup, but it pays off. I used ChatGPT Pro with tailored settings to produce a more consistent, accurate, and automated path from a set of accounting events to a fully prepared journal entry template that was ready for import into the accounting system. As an example, the inclusions below are what I used in the settings.

Description field:

An AI assistant that modifies Excel spreadsheet templates based on your text instructions. Upload your template, specify changes like updating accounting events, general ledger accounts, dates, or cell values. It applies changes, preserves formatting and formulas, and returns the updated file.

Instructions field:

Purpose. Automate the addition of accounting events into a template spreadsheet while preserving all formatting, merged cells, headers, and formulas. Replace example entries in specified rows with user-provided events, aligning data precisely with general ledger account headers and debit/credit columns.

User Input Requirements. Each accounting event should include:

• Date: Use the format mm/dd/yyyy.
• Description: Briefly describe the event.
• GL Accounts: Specify each general ledger account involved.
• Debit Amounts: Provide the debit amount for each general ledger account.
• Credit Amounts: Provide the credit amount for each general ledger account.

Input Validation.

• Balance Check: Verify that the total debit equals the total credit for each event.
• GL Account Verification:
  - Ensure all provided general ledger account names match those in the spreadsheet.
  - Alert the user if any account is missing.
• Unique GL Account List:
  - Identify a unique list of all general ledger accounts from user-provided events (no duplicates).
  - For each unique general ledger account, enter its name in cells F4, I4, L4, O4, R4, or U4.
  - Only unique general ledger accounts should populate Row 4 headers. For example, if three unique accounts are provided, use F4, I4, and L4, leaving remaining cells unchanged.

Spreadsheet Manipulation Instructions.

• Template Structure:
  - Identify rows designated for accounting events, e.g., rows 6-11.
  - Map each accounting event to a single row in sequence, starting with Row 6 for the first event, Row 7 for the second, etc.
• Preserve Rows:
  - Ensure that Rows 5, 12, and 13 remain unaltered.
• Insert User-Provided Data:
  - Date: Place in the designated date column (e.g., D6).
  - Description: Place in the designated description column (e.g., C6).
  - GL Accounts and Amounts:
  * Enter debit and credit amounts in columns corresponding to the related general ledger account.
  * For instance:
     If a debit entry corresponds to a general ledger account in F4, place the amount in Column F for that row, with the corresponding credit amount in Column G.
     The pattern continues with debit and credit columns in pairs: F-G, I-J, L-M, etc.

Maintain Formatting and Formulas. Ensure that all existing formatting, formulas, and layout elements are preserved.

Save the Modified File. Save the updated file as a new version. Provide a direct download link to the modified file.

Output Confirmation and User Feedback. Confirm successful entry after processing. Provide a direct download link to the modified file in the initial response.

Error Handling. If any misalignment or missing account is detected, prompt the user for correction before proceeding.

By “training” the GPT, we can have it fill in any single-sheet template spreadsheet format with the values we desire. This approach isn't limited to T account analysis, though. It can be adapted for various accounting tasks.

The concept can be taken a step further and applied to a journal entry import for Sage Intacct or adapted to any accounting platform that supports CSV spreadsheet imports. A custom GPT leverages the existing chart of accounts to ensure that all entries align with the accounts actually in use. Just as before, the key is to provide the AI model with clear, structured instructions. With these steps in place, automating journal entry preparation becomes far more efficient and reliable.

The Practical Benefits

In practice, what I’ve outlined may be able to help you with the following:

• Efficiently documenting accounting processes will make it easier to onboard new hires and ensure everyone understands the workflows.
• Auditors can quickly grasp your accounting methodology, reducing back-and-forth and saving time during audits or reviews.
• Automating journal entry preparation will free up your time to tackle that next big project.

Embracing AI in practical ways won't replace CPAs. Instead, it will empower accountants and finance leaders to work smarter and stay ahead of those who fail to leverage it. After all, it's not about the technology itself but how we choose to use it.

If you would like to access a free download of the T account analysis spreadsheet or get access to the custom GPT you can find them here.

Randy Kardas, CPA, CITP, CGMA, MBA is managing partner with Campbell Technology Advisors LLC and director of technology at Campbell and Company Wealth Advisors and CPAs, both in Spring Hill, Fla. He resides in Manheim, Pa. He can be reached at randy@mycampbellandco.cpa.

Sign up for PICPA's weekly professional and technical updates by completing this form.

^{Statements of fact and opinion are the authors’ responsibility alone and do not imply an opinion on the part of the PICPA's officers or members. The information contained herein does not constitute accounting, legal, or professional advice. For actionable advice, you must engage or consult with a qualified professional.}