
The power of AI tools in accounting
AI tools don’t just work for you; they work with you. The true value of AI in accounting isn’t in automating one off tasks; it’s making all your tools work together to provide real-time, data-driven advisory at scale.
Accounting firms who take the intentional approach of adopting AI tools can avoid the traps of fragmented apps and instead build a unified AI-driven system to make their practice more valuable to clients.
The art of the AI prompt
Here is the truth about prompt engineering using AI tools specifically for accountants: these tools are designed to work with the way you ask questions, give instructions, and set the context. This directly impacts the quality of responses that you’ll receive
A well-crafted prompt can:
-
Track your clients’ financial trends and suggest actions; leading to more personalised service without the extra work;
-
Save hours of manual work navigating the ATO and Fair Work websites to find relevant compliance rules, allowing accountants to ask simple questions instead of searching;
-
Pull real-time insights across your firm databases using natural language, unlocking hidden insights without running reports manually;
-
Draft responses and handle common client queries, leading to faster client service and happier clients.
This leads to the following outcomes for your business:
- Time saving: AI eliminating manual workflows;
- Smarter decision-making: insights at your fingertips means you’ll be improving your advisory services;
- Error reduction: AI tools ensure accuracy in your tax, compliance and reporting;
- Futureproofing: AI adoption will make your firm more competitive.
On the other side of the coin, if you’re prompt that is vague or poorly composed can lead to confused, underwhelming answers, factual errors and the need to rework or dig deeper to get the output you need.
A major cause of bad responses in AI tools is because we don’t know how to ask for what we want. AI tools are never going to tell you that your question is bad, in fact they will just guess what the right answer is.
Issues in prompt engineering arise when we ask bad questions, because AI tools haven’t been taught not to provide answers that could be considered prejudiced. Therefore, the best first step is to work hard to ensure that the questions you ask are clear and precise.
For a broader look at the impact of artificial intelligence for the accounting industry, check out our guide on AI in accounting.
What makes a good AI prompt for accountants?
Think of a prompt as you would a job brief for a new member of your team. The clearer and more specific you can be in your prompt, the better the outcome for you and your firm.
Here are a few golden rules for prompt engineering in accounting:
Be specific and provide content
The prompt should always provide as much context as possible. This lets the AI know exactly what the purpose of your query is and can help generate a targeted response.
On the other hand, a vague prompt can produce generic or off-topic responses because of the lack of precise instructions that would have otherwise helped avoid ambiguity.
Instead of: "Summarise (insert client name)’s revenue"
Try: "What is (insert client name)’s revenue trend for the last five years?"
Specify the output format
You can save yourself serious time if you provide precise instructions to the AI tool on the format you’ll need to use immediately, such as tables, lists, bullet points, step-by-step instructions or emails.
Asking the tool to provide a format to guide its logic.
Example: "Create a list of clients with an overdue ATO debt."
Break it all down
AI tools perform much better if complex tasks are broken down into smaller, digestible steps. This also allows you to course-correct if you need to.
Instead of: “Analyse my client data and generate insights.”
Try the following:
-
“Which of my clients have overdue lodgments?”
-
“Which of these clients are I forms?”
-
“Generate an email template to follow up with these clients.”
Fine tune when you need to
Remember that prompts can evolve. If the result you’ve generated isn't quite what you requested, take a step back, and iterate and refine your request.
Look over the language you have used and the details you have provided. Even a good prompt can always be improved!
Example:
Prompt A: “Show clients with rental deductions over $5,000.”
Prompt B: “List clients who claimed more than $5,000 in interest from rental properties in FY25, formatted as a bullet list.”
Prompt B is likely to be more accurate and useful. Don’t settle for your first attempt if the result is underwhelming.
See Access Evo prompts in action
Here’s an example to show you how powerful Access Evo prompts can generate precise, efficient results for your practice:
In this scenario, Dave is an accountant focusing on delivering services to his clients every day. He asks Copilot which of his clients have rental properties, using the following prompt:
“Which clients have rental properties?”
This query generates a response showing the clients who have rental properties, pulling from real-time practice data, and showing a formatted list containing the client’s reference number and name.
From here, Dave iterates his prompt, providing additional context to refine the client information he needs:
“Which of these are claiming interest greater than $5,000? Include property address in the results.”
This query specifically filters down for those clients claiming interest amounts greater than $5,00. Dave then decides to action an additional prompt:
“Show me a list of clients who have claimed over $5,000 in deductions returning name and deductions formatted as currency.”
From there, Dave can drill down into an individual client. Notice that he is iterating and refining the prompt, and providing instructions on the formatting that he requires from Copilot:
“What is the breakdown of James Taylor’s deductions shown as a list and formatted as currency?”
Why are these prompts so effective?
Using this scenario as an example, there are several reasons why Dave’s Copilot prompts are so effective:
-
Clear, natural language: Dave’s prompts use simple, conversational language that clearly expresses his intent. This makes it easy for Access Evo to interpret and act on his request without ambiguity.
-
Iterative prompting: Rather than trying to ask a single complex question, Dave iteratively builds on his previous query which allows him to refine his results step-by-step, add new conditions and extract more insights.
-
Contextual awareness: Dave’s follow-up prompts assume Copilot remembers the context of previous queries, showing that it is capable of contextual continuity, making interactions feel more conversational and intelligent.
-
Actionable filtering: Instead of just using Copilot to view data, Dave uses it to filter and act on it, making it a powerful decision-making tool rather than just a reporting assistant.
-
Formatting instructions: Dave didn’t just ask for data, he specified how he wanted it presented (for example “shown as a list” or “formatted as currency”).
How to empower your team to prompt better
Training your team to think like prompt engineers doesn’t require technical expertise, just a shift in mindset. After all, there is no instruction manual for AI accounting tools, or one single right way to use them.
However, there are several things that you can do in your practice to prevent people experimenting in silos and ensure that everyone is aligned:
-
Running internal training or lunch and learn sessions on how to use AI tools correctly and prompt effectively;
-
If you’re doing similar tasks regularly, such as writing client summaries or conducting compliance checks, you could consider creating a shared library of reusable prompt templates to save considerable time in your practice;
-
Encouraging staff to document and share their best prompts so that the practice can reuse them. Over time, your practice could develop a “prompt playbook” tailored to your clients and services.
If your firm is thinking about the bigger picture, you might be thinking about how to articulate an AI policy that’s fit for purpose. Watch ChangeGPS's AI policy webinar on-demand to receive a comprehensive AI policy template that has been especially crafted for Australian accounting firms.
The future belongs to prompt-savvy accountants
AI assistants like Access Evo are powerful—but only if you know how to steer them.
By developing strong prompting habits now, your firm can unlock serious value: from faster compliance jobs to elevated client experiences.
The art of prompting is the new accounting superpower. As with any skill, the more you practice, the more powerful it becomes.