• Thu. Apr 18th, 2024

The Evolution Of AI Chatbots For Finance And Accounting

CFO/Director at Eventus & author of Deep Finance, Glenn has spent the past two decades helping startups prepare for funding or acquisition.

The digital transformation of the accounting and finance department that has occurred over the last three decades is merely the beginning. There are at least three key areas where digitization has and will have the greatest impact: democratization of data, and automation of manual and machine learning.

At the end of 2023, these key components have rapidly merged through the evolution of large language models (LLMs) like ChatGPT and others. LLMs and the neural network computing that powers the computation of hundreds of billions of variables have ushered in a period of experimentation. Given where we are now and where we are going, I wanted to briefly explain how we got here and what’s likely to be coming next.

AI Chatbot Pre-History

Chatbot technology or natural language processing harks back to 1966 and the creation of ELIZA. MIT professor Joseph Weizenbaum created the first machine/human interaction based on a natural language series of questions and responses. Other experiments followed with PARRY and SHRDLU by machine learning pioneers psychiatrist Kenneth Colby and Stanford professor Terry Winograd (who would become the advisor to Google founder Larry Page), respectively. The 1990’s begets Richard Wallace’s ALICE, which in many ways, was the true parent to the modern AI chatbot like ChatGPT.

The Rise Of LLMs

Large language models are the massive data structures upon which AI and natural language processing is built. Early language models relied on statistical approaches, utilizing n-grams and other probabilistic techniques to understand and generate human-like text. In the 2010s, the rise of neural networks, especially deep learning, significantly impacted natural language processing. Word embeddings and recurrent neural networks (RNNs) allowed for better language representation.

OpenAI introduced Generative Pre-trained Transformer 2 (GPT-2) in February 2019. GPT-2 demonstrated the power of natural language generative AI, but at the time OpenAI withheld the full model. OpenAI’s concern about misuse of the model led to the delay. OpenAI released the third iteration, GPT-3, in 2020 with 175 billion parameters, enabling it to perform a wide array of language-related tasks, from translation to code generation. Today, GPT-3.5 and subsequent models have been applied across diverse domains, including content creation, chatbot development, programming assistance and more.

Use Of Chatbots In Accounting And Finance

Accounting and finance in the corporate setting are complex and filled with rules, regulations and standards. It requires a level of detail and accuracy that has thwarted much of the effective use of chatbots until recently. Pre-2010, commercially available bots were rules-based decision trees that were not adept or flexible enough to handle the minutia inherent in this part of the business, so we’ll jump ahead.

From the late 2010s to today, the rise of artificial intelligence and natural language processing has brought an array of chatbots to accounting and finance. They have started to understand and respond to more complex queries, making them more useful for tasks like account reconciliation and financial analysis.

Lessons For The Accounting Professional

From using Amazon’s Lex language to create a rule-based finance chatbot, through a couple of iterations of Python applications that could perform FP&A at scale, and work with an automated audit software, I’ve experimented with how these new technologies can and will transform the accounting and finance profession.

Here are a few lessons I’ve learned.

Lesson 1: Experimenting Is (More Or Less) Free

All the work I have done in creating my AI tools for accounting and finance has been done in my spare time and with freely available tools. There is a minimum cost to access certain plug-ins or APIs and you need to have a modicum of technical understanding—but I am by no means either an AI expert or a computer scientist.

The truly amazing environment for AI exploration is at anyone’s fingertips. Many professionals may not have the deep interest that I have in getting their hands dirty, so just as important is keeping up to date with what’s happening in the generative AI space. It’s moving so rapidly—for example, I created a tool for one of my projects that, days later, OpenAI made natively available through ChatGPT.

Lesson 2: Less And Less Limitations

My experiments have shown me just how expansive the tools that generative AI can be. From financial modeling to audit prep to equity analysis—you name it—if there is a data set available for it (whether your own in-house or a publicly available one) you can train an LLM to understand and analyze the data. By understanding, I don’t mean in the human sense.

I believe that human beings will continue to be needed to truly comprehend what the data ultimately means when it comes to making decisions—but the power of AI (hopefully) should bring with it better decisions made faster or deeper analysis that identifies problems sooner. If there is a sticky problem you are wrestling with, think about how AI could help solve it.

Lesson 3: Generative AI Is Here To Stay

If you think your job is safe from the AI revolution as an accountant or finance professional, you may be right in the short term. In the long term, however, many white-collar jobs are in jeopardy from AI, but this doesn’t suggest we should fight against the introduction of this technology into our workflows. Any more than fire can be returned to Prometheus, generative AI is out there and flourishing, already entrenched in our technological lives.

The recent boardroom controversy at OpenAI and the attention it received highlights this truth. There will likely be an ongoing debate about various aspects of AI, from its ethical considerations to the creation of misinformation to the fear of AIs enslaving mankind—but it has a place in today’s economy and work world and its role is only going to increase.

What’s the bottom line for us as accountants? Ignore AI at your own peril.


Forbes Finance Council is an invitation-only organization for executives in successful accounting, financial planning and wealth management firms. Do I qualify?


link

By admin

Leave a Reply

Your email address will not be published. Required fields are marked *