Six tips for accounting professionals to excel in a world with ChatGPT
Accountants have an opportunity to ‘level up’ and become proactive and influential advisors to businesses and clients through the selective and smart use of AI
By Professors Paul Andon, Professor Fahim Khondaker (UNSW Business School) and ChatGPT*
We urgently need to move to productive conversations around ChatGPT and similar 'intelligent' technologies. AI is here to stay, and with the Microsoft 365 Copilot announcement hot off the press, ChatGPT's next-generation AI will soon be right there on our desktop, ready to support us with every Word document or Excel worksheet we create. The everyday presence of AI holds the potential to revolutionise the role of technology in knowledge work, and supercharge the productivity and 'human potential' of all accounting professionals.
The future prosperity of the accounting profession relies on our capacity to leverage rapid advancements in AI. We encourage our accounting peers to 'join the AI revolution’. To this end, we offer six tips below to help our professional colleagues embark on their AI learning journey.
Accountants have an opportunity to ‘level up’ our capabilities as proactive and influential advisors to the businesses and clients we serve. AI allows us to go beyond our knowledge of ‘the numbers’ to actively enable and influence key business decisions. We sell ourselves as ‘trusted business advisors’ – now more than ever, and with the assistance of AI, we will be able to walk that talk. The future of our profession depends on it!
Our 6 tips for accounting professionals
1. Give it a go: If you haven’t done so yet, give it a go now. The best way to learn how ChatGPT works is by experimenting with it. There is no need for a user manual or help function as its plain English conversation interface makes it so easy to use. By experimenting – asking it questions, giving it tasks, and observing its responses – you can discover where and how ChatGPT can improve your efficiency and effectiveness at work. The only caveat to this is to keep privacy in mind, do not include confidential information in ChatGPT. It will store and likely make use of any information you feed into it.
2. Make ChatGPT your assistant: We encourage you to think expansively about what ChatGPT can do for you. For example, we often 'Google it' to find initial information on a given topic or content inspiration for a specific task. Soon, we will ask Chat GPTfirst – we may never need to write a first draft again.
The AI does not simply provide you with a dump of links. With appropriate prompting it can compile, organise, synthesise, script, style, and edit relevant information it has access to, and much more. Rather than a mere 'answer machine', think of ChatGPT more as an assistant, writing coach, idea generator, or script/memo advisor you've always needed. For this reason, it is quite apt that Microsoft has chosen the name Copilot for its new AI model.
3. Ask the right questions: The responses you receive from ChatGPT will only be as good as the questions you ask it. The good news is you do not have to get it perfect the first time, you can ask as many follow-up questions as you like to refine the answers (as you would in a normal conversation or interview).
Asking the right questions of data or clients is a professional skill and it is exciting to imagine the types of questions our human brains will think of when answers are so readily available at our fingertips. We will have more time to focus on the purpose, meaning and implications of the answers across whole organisations or topics we are exploring. ChatGPT has also provided some basic tips for prompts:
- Be clear and concise
- Provide context
- Be specific
- Avoid vague or open-ended questions
- Use examples
- Use proper grammar and punctuation
- Review and refine
4. Evaluate responses: It is essential to critically evaluate all generated responses and not assume them to be correct. It is easy to be fooled by the confident tone of ChatGPT responses (often referred to as ’botsplaining'). We must avoid mistaking confidence for competence, just as we do with our interactions with work colleagues, clients, and other stakeholders.
This should not be too challenging for accounting professionals, given our attention to detail and professional scepticism honed through years of training and experience. We simply need to adapt these skills to our use of AI, ensuring that we apply the same level of scrutiny to ChatGPT's responses as we would elsewhere. This will enable us to use ChatGPT credibly, and with confidence.
5. Understand limits and risks: There are key limits and risks associated with using ChatGPT, though these are constantly being improved upon. Essentially, ChatGPT is a 'large language generation model' that learns patterns between words and sentences from extensive training on a vast textual database of books, articles, websites, and the like. So while ChatGPT answers look very 'human-like' and knowledgeable, they actually only reflect the model's assessment of the most probable sequence of words that should follow the prompt based on its training.
Consequently, answers simply reflect ChatGPT’s probabilistic assessment of what the answer should be. Answers do not reflect a true understanding of what is produced, and can often miss real-world sensibilities or specificities of context in its responses. Additionally, it struggles with abstract reasoning and computational logic (it knows that one plus one equals two because it has read it somewhere), and it cannot comment on anything outside of its training dataset. ChatGPT also cannot express empathy or emotion, and it can also make mistakes as a result of biases or gaps in its available data.
6. Reference the use of AI: We must always acknowledge how it has been used in our work. Appropriate acknowledgment of the AI’s contribution will promote transparency in tools used, demonstration of how AI is used to produce high-quality work, demonstration of the ‘value-add’ the accountant brings to work enhanced by AI, and accountability for the part that AI plays where issues or concerns arise in work produced.
Professor Paul Andon is the Head of the School of Accounting, Auditing and Taxation at the UNSW Business School. Professor Fahim Khondaker, is a Professor of Practice at the UNSW Business School and is a Partner in Data Analytics and Insights at BDO, a global professional services firm.
We acknowledge the use of ChatGPT for helping us draft and refine this article. Specifically, it was used to spark ideas, draft, edit, clarify and organise our thinking of various parts of the article. However, the primary value-adding ideas presented in the article are ours, developed from our combined experiences.