Preface/Background
Before we start to discuss ChatGPT’s foreseeable impact within the field of accounting, let us first understand what defines ChatGPT as a whole as well as what it entails for the field of AI and all other interrelated industries. First off, let us break down the name of the AI model itself, “ChatGPT”. It’s supposed suffix, GPT stands for Generative Pre-trained Transformer. A Generative Pre-Trained Transformer is a type of artificial intelligence model used for natural language processing. It uses machine learning techniques to understand, interpret, apply, and generate human-like text.
ChatGPT was trained on a set of datasets known as WebText, and as the name suggests, the dataset is a corpus of text scraped from the web which contains 8 million webpages. Another dataset that the model was trained on is the Common Crawl dataset, of which is open-source and composes the bulk of the training dataset. The model was also trained on various other text sources including books, articles, a mixture of licensed data, data from human trainers (questions and prompts) and other written content. This training process involved multiple iterations and improvements to achieve the best possible performance.
As of currently, ChatGPT has been used to generate both instructional and creative texts such as short stories, songs, press releases, technical manuals, and other forms of text. Additionally, the AI model can also mimic pastiches of particular authors.
It was only until recently that when the latest version of GPT-3 (the version prior to the now current ChatGPT (GPT-3.5)) launched that individuals had the opportunity to interact with ChatGPT directly, ask it questions, and receive comprehensive and practical responses.
ChatGPT has had significant impacts across various industries, including:
- Customer service: Companies are leveraging ChatGPT to automate responses to common inquiries.
- Education: ChatGPT is being used to create intelligent tutoring systems capable of providing personalized assistance to students.
- Content creation: Journalists, copywriters, and content creators are using ChatGPT to generate creative ideas, draft articles, and even write poetry.
- Businesses: All kinds of professionals are using the chatbot to automate tasks like drafting emails or writing code.
- Healthcare: Providers and staff can leverage the chatbot for use cases like clinical decision support, medical recordkeeping, analysing and interpreting medical literature, and disease surveillance.
- Entertainment: ChatGPT can be used to generate video game storylines and movie scripts, write dialogue, and improve gaming.
Microsoft is planning to integrate ChatGPT functionality into its productivity tools, including Word, Excel, and Outlook, in the near future.
Brief summary of applications of ChatGPT in the field of accounting
Adopting ChatGPT in accounting can simplify complex tasks, streamline processes, and enhance overall efficiency. With its ability to understand context, analyse data, and also generate valuable insights.
ChatGPT can provide more efficiency in tasks such as financial reporting or invoice categorisation which can improve the quality and reliability of financial information. However, users must keep in mind not to enter any Personally Identifiable Information written or printed on invoices or bank notes.
The AI model can also process and analyse vast volumes of data quickly, extracting valuable insights and information which might be overlooked when done so manually. The model can be used to detect trends in financial data such as how a business should distribute their expenses to achieve a more optimal level of business performance – in terms of human resources (hiring, training, consulting, and professional development), physical and technological resources (rental, resources, manufacturing, and production) and events and promotions (team building, advertising, and marketing).
As evident within its name, the term “Chat” suggests that the primary use and intention of the model is to provide a user-friendly mode of interaction. In a more professional sense, it can be applied in a customer-service environment to provide timely, informed responses. The model can be used to provide real-time, high-quality customer service. It can respond to queries accurately and promptly, enhancing customer experience and satisfaction, which allows humans to focus on more complex customer interactions that aren’t entirely technical.
The model can also be used to mitigate risk. It can detect anomalies in large datasets, which means it can be used to detect financial fraud, manipulation, falsification or alteration of records or documents, misappropriation of assets, suppression, or omission of transactions from records, and misapplication of accounting policies knowingly. Generative Pretrained Transformers can provide an extra layer of security against fraudulent activities. The model’s ability to mitigate risk can help protect and prevent business from suffering potential financial losses and lawsuits.
The automation and delegation of routine tasks to ChatGPT will result in a reduction in human-errors and ultimately lead to saving costs. The savings can be directed and reallocated to more strategic business areas and processes, which allows business to maximise their monetary resources to more productive processes such as investments, business pitches and deals, and manufacturing and production to increase the company profitability both on local and foreign market listings.
More-detailed breakdown of how ChatGPT can be applied in accounting
1. Invoices and expenses categorising
Business receive and issue stacks of invoices daily. It might be possible to record every single one given that a company has sufficient manpower to record every single one accurately – which is unlikely. And even if they manage to, it is possible that duplicate entries might occur. Not to mention, manual entry and categorisation is quite time-consuming and may introduce errors to records. As ChatGPT has been trained on a diverse corpus of text, which can help businesses automate this process. Equipped with machine learning capabilities, it can analyse and categorise invoices and expenses based on predefined categories and even learn and improve over time, increasing efficiency and accuracy.
2. Financial projection/consulting
One of the key responsibilities of accountants and financial professionals is to provide accurate financial projections. These are essential for strategic planning, securing funding, and evaluating the financial health of a business.
3. Financial risk assessment
Business are prone to various financial risks, such as market volatility, competition, and overlooked or unexpected expenses. ChatGPT can help pinpoint and identify such potential risks and provide recommendations on how to mitigate them.
Below is a step-by-step guide on how to perform financial risk assessment using ChatGPT:
- Collect financial data
The first step is to collect and compile all your relevant financial data, including your income statement, balance sheet, and cash flow statement. - Analyse historical data
Once you have compiled your financial data, you can analyse it to identify trends and patterns. For example, you may notice that your profit margins are declining or that your cash flow is becoming increasingly strained. - Identify risks and mitigation strategies
You can use ChatGPT to identify potential financial risks associated with business expansion, such as increased operating expenses or reduced demand for your products. ChatGPT can also provide guidance on how to mitigate these risks, such as by developing a contingency plan or adjusting your pricing strategy. - Assess financial health
You can use ChatGPT to assess your overall financial health, including your liquidity, solvency, and profitability. This assessment can help identify any potential weaknesses or areas in your business that require improvement. - Review and analyse results
Once ChatGPT has generated a financial risk assessment, you can review and analyse the results. You can use this information to make informed decisions about how and whether to proceed with a business expansion and what steps to take to minimise financial risks. - Monitor and adjust as required
It is crucial to monitor your financial performance and adjust your strategies as required. If your financial performance deviates significantly from your actual projected performance, you may need to adjust your strategies or seek professional assistance to mitigate these risks.
4. Financial report generation
Generating accurate and timely financial reports is a crucial aspect in accounting, which often requires extensive time and effort. ChatGPT can streamline this process by automatically generating basic financial reports, such as income statements, balance sheets, and cash flow statements, based on input data. This gives accountants more time to analyse the reports generated and provide strategic advice to businesses. ChatGPT can help by analysing a business’s financial data and generate cash flow forecasts for the coming months or years depending on how cohesive and reliable the data provided is. ChatGPT has an extension called ChatGPT File Uploader Extended that provides templates for Cash flow, Profit & Loss reports and Balance Sheet templates. Again, users have to keep in mind to not enter any sensitive or confidential information into their templates.
Below is a step-by-step guide on how to perform cash flow forecasting using ChatGPT:
- Collect financial data
The first step is to collect and compile all your relevant financial data, including your sales, expenses, and accounts receivable and payable. - Analyse historical data
Once you have compiled your financial data, you can analyse it to identify trends and patterns. - Input the data into ChatGPT
You can input your financial data into ChatGPT and prompt it to generate a cash flow forecast for the next six months and also identify any potential risks or opportunities that you should be aware of. - Review and analyse results
Once ChatGPT has generated a cash flow forecast, you can review and analyse the results. You can use this information to make informed decisions about inventory management, purchasing, and other expenses. For example, if ChatGPT predicts that you will have a cash surplus in the next few months, you may want to invest in new inventory or marketing initiatives. - Monitor and adjust as required
It is crucial to monitor your actual cash flow and adjust your forecast as required. If your actual cash flow deviates significantly from your forecast, you may need to adjust your inventory management or other expenses to stay within your budget.
5. Auditing Documents
Auditors have to sift through vast amounts of financial data, which can be tedious and labour-intensive. ChatGPT can be used to scan through and organise data, effectively making the auditing process more manageable. It can help to cross-check entries, identify anomalies, and organise the required documentation for the audit.
6. Customer interaction/service
As professionals, apart from providing technical services such as producing financial reports and organising finances, accountants also have the responsibility of informing their clients regarding financial regulation changes, upcoming deadlines, and other related relevant financial matters. Apart from producing financial templates, ChatGPT can also produce email and message templates based on the title and context of the message.
Advantages
Faster analysis
Easily accessible machine learning resources
Ability to analyse unstructured data
Disadvantages
Limited scope
Dependence on quality of natural language input
Limited customisation
Lack of industry-specific knowledge
- Complex Decision Making
Apart from producing and compiling financial reports and analysing them, accountants must make complex financial decisions. These decisions include providing strategic business and financial advice, and applying professional judgment; which are some of the tasks that AI has yet to handle effectively. For instance, an AI model can detect financial anomalies, but an accountant is required to trace and investigate the source of those anomalies and how the transactions occurred.
-
Complex Decision MakingSeveral aspects of an accountants day-to-day tasks are heavily involved in interpersonal relations such as discussing financial matters with clients, resolving conflicts, or presenting financial information to non-financial stakeholders. AI is far from replicating human emotional intelligence as well as the nuances of human communication such as body language, tone, and verbal cues.
- Ethics and Professionalism
Accountants must adhere to ethical standards and professional guidelines. While AI can be programmed to adhere to certain rules, it has yet to comprehend ethics and professionalism in a business scenario.
-
Adapting to Change
The financial landscape is ever-changing and dynamic, and the same is true for the regulations that inform it. The business models and financial tools used in the industry are always evolving and emerging – and to keep up with this ever-changing landscape, AI models have to be trained along with it.
Setbacks of ChatGPT
As with all AI models, the results that they produce are prone to the nature of the data that they are trained on – more specifically, the parameters and the biases that they introduce. Users have entered prompts into the model using the words “black”, “women”, and “holocaust” and needless to say, the AI model produces some offensive results.
Not to mention, due to the way AI models are developed, as behind nearly all AI models is what is known as a “Hidden Layer”. We know little to nothing in regards to what occurs behind the scenes within the hidden layer of an AI model – All we know is that an AI model is made up of neurons (as the neurons within our brains are) and those neurons make connections (or associations) with neighbouring neurons as they learn, integrate, and apply that information over time. As the hidden layer learns and applies what it learns along the way, there are instances where it will produce unexpected and unexplainable results.
Results aside, ChatGPT has also been proven to purposely generate less than satisfactory responses when prompted to – During one of its testing settings, it was instructed to provide toxic responses and did so without hesitation – in fact, the responses were significantly more toxic.
Fortunately, the world of accounting need not be concerned with such matters as the terms used in the industry are largely technical; albeit it’s always helpful to be aware that even ChatGPT itself isn’t the epitome of all AI models as much as it appears to be, due to its popularity.
To quote Sam Altman, co-founder of OpenAI along with Elon Musk, expressed the following: “The GPT-3 hype is way too much. It’s impressive (thanks for the nice compliments!) but it still has serious weaknesses and sometimes makes very silly mistakes. AI is going to change the world, but GPT-3 is just a very early glimpse. We have a lot still to figure out.”
Future prospects of ChatGPT
As ChatGPT gains even more popularity over the decades to come, we can expect large corporations to start using a more fleshed-out proprietary version of ChatGPT for commercial and business purposes; possibly to facilitate already existing business workflows to make them more efficient or to obtain up-to-date information related to their industry.
Although I am not a specialist within the field of AI itself, I suppose the appropriate question to ask is “Can AI models unlearn knowledge and information that it has been trained on?”. Because even we, as humans ourselves, find it challenging to unlearn what we have been taught – and unlearning knowledge and information takes considerably more effort in comparison to learning, integrating, and applying new information. It is evident that the process of unlearning something is accompanied by resistance, as it is challenging to forget knowledge that we have believed and held true within our knowledge base and memory for so long. I suppose what I am proposing is that we are placing mighty high expectations for an AI model to be perfect when we ourselves are far from it.
References:
https://www.popautomation.com/post/how-will-chat-gpt-impact-accounting
https://www.pcguide.com/apps/chat-gpt-plus/
https://towardsdatascience.com/how-chatgpt-works-the-models-behind-the-bot-1ce5fca96286
https://medium.com/@dlaytonj2/chatgpt-show-me-the-data-sources-11e9433d57e8
https://brixx.com/financial-forecasting-and-planning-with-chatgpt/
https://brixx.com/how-to-use-chatgpt-to-write-a-business-plan/
https://www.linkedin.com/pulse/chatgpt-future-accounting-ca-mohit-loonkar/
https://www.foraccountants.com.au/why-accountants-should-pay-attention-to-chatgpt/
https://pdf.wondershare.com/chatgpt/chatgpt-for-accounting.html
Authors:
Katherine Lee Chin Fei Maxine Yang Fu Ting
