What CPAs should know about machine learning vs. deep learning

Artificial intelligence (AI) is already changing the nature of our lives. When you ask Siri for a restaurant recommendation or when you tell Alexa to play your favorite song, you’re interacting with AI.

AI has also been making its way into the accounting world. Finance departments, for example, have found ways to automate what was once laborious data entry using the technology. Among the activities automated are vendor bill data entry, general ledger coding, and expense reporting, to name just a few.

While these changes are helping to increase efficiency, they’re just the beginning of the innovations headed our way. To help you prepare for those, this article looks at two subsets of AI, machine learning (ML) and deep learning, that could have a huge impact on the way accounting professionals do their jobs. Keep in mind that to take full advantage of AI, you must move to cloud technology, as discussed in my Sept. 17 article “The Required Step Before AI and Blockchain.”

As machine learning and deep learning have become buzzwords throughout the tech and business worlds, commentators have rushed to make sweeping generalizations about robots replacing human jobs within accounting. In this race to make the grandest statement possible, a lot of nuance gets lost and the narrative of accountants losing their jobs gains an audience. The truth is that while these technologies will certainly alter our careers, nobody yet knows exactly what the transformation will look like. In fact, many people don’t even fully understand the terms that are being used so that they can make their own determination of how this will affect them and their careers.

Defining AI, ML, and deep learning

You may remember the old adage from your high school geometry class that “every square is a rectangle, but every rectangle is not a square.” AI, machine learning, and deep learning have a similar relationship. All deep learning is machine learning, and all machine learning is artificial intelligence, but not vice versa. AI is the largest umbrella, followed by machine learning and finally deep learning.

Let’s start at the top. AI refers to the ability of machines to mimic human intelligence. Software developers facilitate this by taking knowledge of how a human performs a set of tasks and then writing code that empowers a machine to perform that set of tasks on its own. AI includes things such as workforce planning, understanding and translating language, recognizing images and sounds, and even applying knowledge and problem-solving skills to complete tasks. For example, a software program could mimic an accountant who uses knowledge of the tax code to run your tax information through a set of static rules and gives the amount of taxes you owe as a result.

Machine learning takes AI to the next level. Beyond the coding to mimic human behavior with AI, the software begins to learn the result of the set of tasks and how a human may respond, in order to speed up the task for the next time. For instance, after you watch a movie or a TV show on Netflix, you get recommendation for other movies and shows to watch based on what you’ve already watched. This is an example of machine learning creating a better experience by automatically providing you with relevant content options, potentially saving you the time of having to do a manual search.

Taking things one step further, deep learning is based on the structure and function of our brains, the interconnection of the many neurons. Within deep learning, artificial neural networks (ANNs) are algorithms that mimic the biological structure of the brain. In ANNs there are different neurons that have discrete layers and connections to other neurons. Each layer picks out a specific feature to learn. It is that layering that gives DL its name; depth is created by using multiple layers as opposed to a single layer. An example of deep learning is Google Translate, which can automatically translate images with text in real time to a language of your choice. You hold your camera phone over an image or text and your phone runs a deep learning network to read the image and translate into the language you speak.

An example of using all of this together is a driverless car. With AI, you put a destination into your GPS, and the car executes your orders. Machine learning could prompt the car to ask whether you want to go a certain place based on the date and time you turn on the engine. A deep learning car could automatically stop for coffee at your favorite cafe on the way to work, just by knowing that you often stop there.

What this means for accounting

While AI adoption for data entry may be a viable option very soon, machine learning and deep learning applications are still a ways off.

Because machine learning relies on huge amounts of data to provide accurate results — e.g., the more content you choose on Netflix, the better its predictions on what else you would like — the biggest accounting firms are leading the way when it comes to developing machine learning applications. That’s because the biggest firms not only have the most resources to invest in research and development, their huge client bases also give them tons of data with which they can test what works in the accounting space.

For example, KPMG uses IBM’s machine learning platform, Watson, to help leasing companies comply with the IFRS 16 lease accounting standard. Argus, a tool developed by Deloitte, uses machine learning to review documents for key accounting information. The tool works with many types of documents, including but not limited to sales, leasing and derivatives contracts, employment agreements, invoices, client meeting minutes, legal letters, and financial statements. Deloitte also has introduced a service that monitors risk associated with algorithms and machine learning, which can help early adopters use these innovative technologies with reduced fear of adverse effects.

How to embrace the tech

While the Big Four have the most financial resources to invest in AI-related technologies, Jeanne Boillet, the global assurance innovation leader for EY, believes that smaller practices have a chance to experiment more effectively because they can be more agile to respond to market changes. She recommends that “accountants … start small but think big: Starting out by doing simple proofs of concept that are highly relevant for your business will ensure that your use of AI in the future will be suitable and effective for you specifically.”

As a smaller firm practitioner, you have the opportunity to take advantage of software packages that are pre-built by the developers that are investing in AI and ML innovation. You can research online ecosystem marketplaces that integrate with your software-as-a-service (SaaS) general ledger systems to automate tasks such as bill payment, expense reporting, audit sampling, and more. It is always good to test one application at a time to see if it will help your practice increase efficiency and minimize errors and allow you more time to work on value-added tasks for your clients. It is important when researching applications that you ask about their AI and machine learning road map to ensure they are going to continue making investment in this area.

Ultimately, machine learning and deep learning will allow accounting professionals to spend more of their time helping their clients run their business, rather than focusing on the data entry and computational aspects of their jobs. Our advice has always differentiated us from technology. It’s our time to not only focus on the technology, but also to invest in the learning around soft skills areas such as communicating the analysis of financial ratios and collaborative opportunities with clients and your team. Learning the technology and focusing on the human component are equally essential. You will need to be strong in both areas to thrive in the coming years.  

Amy Vetter, CPA/CITP, CGMA, is CEO of The B3 Method Institute, a keynote speaker and adviser, Technology Innovations Taskforce leader for the AICPA’s Information Management Technology Assurance (IMTA) Executive Committee, and author of the book Integrative Advisory Services: Expanding Your Accounting Services Beyond the Cloud, published by Wiley. Learn more at www.amyvetter.com. To comment on this article or to suggest an idea for another article, contact Jeff Drew, a JofA senior editor, at Jeff.Drew@aicpa-cima.com.