insight magazine

Accounting for AI

IBM’s Watson tackles audit and tax as artificial intelligence takes on the accounting world. By Ryan Watson, CPA | Summer 2017


Ken Jennings was untouchable. Beating back competitor after competitor, he strung together a winning streak for the ages, which, to this day, remains the longest winning streak in Jeopardy history. Then, in February 2011, Jennings squared off against his most formidable opponent in his historic 74th game. Watson.

Weighing in at just under 1,500 pounds and having 200 million pages of information committed to memory, Watson didn’t look like the 144 people Ken had defeated. In fact, Watson wasn’t a person at all. Watson was IBM’s AI (artificial intelligence) supercomputer, and unseating the greatest trivia expert of all time was just the coming out party for what appears to be a long and disruptive livelihood.

Its latest endeavor? Accounting.

Who’s Watson?

A pioneer of cognitive computing, Watson is quickly becoming whoever you need it to be. Unlike basic computing, which only works with structured data in databases, Watson relies on natural language processing to understand and process unstructured data. Meaning, rather than solely looking for keyword matches, it evaluates words and phrases to interpret the context and draw inferences about the meaning and intent—just like you and me.

When Watson begins in a new domain—like tax, for instance—it begins with a basic collection of information as the foundation of its knowledge. Then, with the help of human experts, it refines its knowledge through a process known as machine learning, where it evaluates pairs of questions and answers to understand linguistic patterns in the data. Like humans, Watson uses information to form and analyze options, develop hypotheses, make decisions, and ingest results, getting “smarter” with each feedback loop.

Watson Rising

In Watson’s first commercial application, it outpaced real doctors in successfully diagnosing lung cancer. IBM has since partnered with many industry leaders in education, financial services, healthcare, and more, and is by all accounts looking to be the central AI engine powering all business and consumer applications through its developer ecosystem and open application program interface.

It was just a matter of time before Watson would take on the accounting world, and that time is now.

After all, Watson excels at analyzing terabytes of disparate, unstructured data to apply order and context. Watson processes this data and makes determinations on it in mere milliseconds. In many ways, evaluating data and applying “rules” is the bedrock of traditional accounting work, leaving Watson poised to disrupt a significant amount of the work being done.

Automated Assurance

Not buying it? Several of the largest firms in the industry have already begun deploying Watson in exactly this way. While they’ve not disclosed all of the specifics around its application, KPMG is leveraging Watson in its audit practice to analyze large volumes of structured and unstructured data related to a company’s financial information. Watson is being trained by KPMG’s auditors to fine-tune assessments, giving them faster access to precise measurements used to analyze anomalies.

An increasingly important component of assurance work is performing analytic procedures on individual transaction data. Historically, these procedures have been limited and basic, restricted by the computing power of humans and their laptop computers. But with Watson, all bets are off. And as the technology behind the AI supercomputer improves, it will continue to make more sense for firms, like KPMG, to use Watson to develop a much richer understanding of the enormous amounts of data that we’re increasingly producing.

Tackling Tax

H&R Block is another noteworthy player leveraging Watson to disrupt traditional business practices, but this time it’s for tax preparation.

“By using Watson, H&R Block’s tax professionals highlight additional areas of possible tax implications based on a personalized cognitive interview,” explains George Guastello, H&R Block vice president of client experience. “The client can follow along with the tax professional on a monitor that brings to life key focus areas for deductions, making tax preparation more engaging and transparent.”

What’s interesting about the H&R Block/Watson partnership is that it offers us a glimpse into the future of human and machine interaction. According to Guastello, Watson isn’t replacing tax preparers, but rather helping them do their jobs better.

“Tax professionals are the heart and brain of the client experience, that role will not change,” says Guastello. “Watson augments and expands human intelligence, it does not replace it. This is man and machine working together; the best tax professionals working with the best technology to deliver the best outcome for clients.”

In this, we begin to see a future for how AI and humans will exist symbiotically. In professional service industries where trust is foundational to the relationship, automation alone is unable to service the need. However, when automation and client service combine, the promise is a higher quality product at a lower cost, benefiting both the accountant and the client in the long run.

AI Ahead

It’s unclear what IBM’s true intent is for Watson in the accounting space, but clearly opportunities for AI in the industry exist broadly. Machine learning systems are increasingly being implemented into accounting platforms, automating manually repetitive tasks. For instance, cloud-based accounting software provider Xero recently applied machine learning capabilities to its “Find and Recode” function, a function designed to help accountants and bookkeepers tidy up their clients’ incorrectly coded entries.

The model learns from invoice coding behaviors, noting any mistakes recognized and corrected by the advisor, and then offers suggestions and predictions. Next time the user tries to affix an incorrect code to an invoice, the machine learning system suggests a label that should be used instead.

What’s next? Imagine a big data engine that can monitor credit worthiness against thousands of indicators and approve financing in seconds. Consider the possibility of an audit quality system that can perform real-time benchmarking against a perfectly comparable peer group. Perhaps we’ll one day see a program that can do smart forecasting based on company-specific and macro-economic trends. They’re all likely scenarios that will bring change to accountancy.

There will never be a future without human accountants and bookkeepers, but those who will succeed in the age of automation and machine learning will welcome the disruption. In this future, the accounting professional will be focused on offering intuitive insights and more value to their clients. And the mundane, dirty work? Well, we’ll leave that for the machines.

Ryan Watson, CPA is a founder and principal of tech-savvy accounting firm Upsourced Accounting and Xero’s Midwest ambassador. He can be reached at [email protected].

Leave a comment