AI Revolution in Accounting

The world renowned venture capitalist firm Andreessen Horowitz recently published an article on how Generative AI will change accounting. They present a multitude of jobs and functions currently performed by accountants which are ripe for disruption by the generative AI revolution.

The launch of ChatGPT at the end of 2022 seemed like a breakthrough moment – like seeing a computer or the internet for the first time (for the millennials and older generations out there). Most professionals in B2B services quickly realized the incredible potential and existential threat generative AI represented to our industries. Everyone started making predictions on how long we had before we had to reinvent ourselves. Yet, almost two years later, the AI revolution has so far fallen short of driving meaningful productivity gains in accounting and FP&A functions.

Make no mistake, our view is that this remains a question of when and how, not if the revolution will happen. Here, we share our thoughts on three essential factors which currently hold back the potential for real productivity gains from generative AI in accounting: segmentation of applications and data, changes in business processes and the necessary evolution and training for accounting and finance professionals.

Segmentation of applications and data

The Andreessen Horowitz article presents a number of tasks and functions performed by accountants: data collection, reconciliations, research, financial reporting, tax filings among others. For good or bad, these inevitably form part of the day-to-day lives of accountants, finance professionals and business owners worldwide. To perform all of these tasks, accountants need to receive information from multiple sources (invoices, statements, emails, instructions, queries, data extracts and more). Currently for most SMEs worldwide, all of these are handled by different applications: emails, Slack, WhatsApp groups, Quickbooks, Xero, RAMP, Dext… to cite just a few among the multitude of applications which form part of their usual toolkit. Integrations between some of these applications are possible but they are limited and usually focus on one specific task or process.

Large Language Models (like ChatGPT) were trained on huge amounts of data available on the internet, other digital sources and through interactions with human trainers. This makes them very good at writing, creating images, videos and so on. The lack of global data sources to determine how businesses analyze specific transactions and make decisions is one key limiting factor to achieve similar results in accounting.

Accounting standards, tax laws, online articles and opinions can be an efficient source of information but ultimately each business is unique. Automating a significant portion of accounting functions will require the development of AI models that can adapt to the uniqueness of each business, the applications it uses, its procedures, its tax management and more.

Improvements in business processes

At CloudCFO, we work with hundreds of SMEs worldwide, with the mission of empowering them to make timely and financially informed decisions about their business. Most of the companies we start working with come to us with concerns related to their inability to obtain timely and accurate financial reports, plan cash flow and rely on financial data to make decisions for their business.

These challenges are not due to technological barriers. Software as a Service accounting solutions (such as Xero and Quickbooks Online) have been available for decades. They are accessible, easy to implement and affordable.

Based on our experience, the source of these major hurdles lies in the lack of efficient business processes which consider accounting procedures, financial controls and utilize the full capabilities of existing software. Technology without clear operating procedures will not result in optimal gains in business value and efficiency. The same can be expected for AI solutions. To successfully adopt and derive efficiency from generative AI in accounting, companies and accountants will need to rely on efficient business processes. In turn, they will also need to adapt these processes to leverage the full potential of AI.

As recent history has shown, meaningful changes in business processes take time. The personal computer revolution started in the 1980s but real productivity gains in businesses only materialized in the mid-1990s as companies needed to adapt their entire operating model and processes to fully make use of the new technologies.

Reinventing the role of the accountant and professional adaptation

The idea of fully automated accounting and tax processes certainly sounds like a relief for a majority of business owners. But it can be frightening at the same time. Would you board a fully automated plane, without a human pilot? Similarly would you let AI complete and file your financial statements and tax returns without a human professional control?

Hence, it is fair to assume that the role of an accountant will not disappear but it will certainly evolve in a drastic way. This represents the third hurdle to substantial productivity gains from AI: changing a profession which has not evolved much in decades.

Accountants are trained and certified based on understanding and application of principles and standards (IFRS, GAAP…). Even today, new technologies, their implementation and optimization barely forms any part of most degrees or certifications. These are acquired through practical experience and hands-on training.

For AI to drive real efficiency gains in accounting, the role of the accountant needs to fundamentally evolve to focus on technology integration, analytical and advisory functions. This adaptation will take time and require a major upgrade in training and educational programs.

The accounting AI revolution is certainly well on its way. But to some extent, its march has been slower than expected. AI powered applications already provide potential for automation and productivity improvements but their impact remains limited.

In their recent book, Power And Prediction: The Disruptive Economics of Artificial Intelligence, economists Ajay Agrawal, Joshua Gans and Avi Goldfarb make a compelling analogy between the introduction of electricity and AI. Initial gains in production capacity and productivity from the commercialization of electricity were fairly limited for decades because the initial adoption of electricity focused on improving existing tasks or creating new applications for electrical devices. It took close to 40 years and the adoption of the assembly line, a new end-to-end production process in the early 1900’s for significant gains to materialize.

The AI revolution in accounting seems to follow a similar path. To meet its full potential, the profession will need to rethink and reshape itself end-to-end: its applications and software, its processes and most importantly the focus of its expertise.