Few ideas in the last decade have provoked as much excitement — or as much misinformation — as the introduction of artificial intelligence in tax and accounting. Starting around 2015, the business world became captivated with the technology, drawn in by headlines that ran the gamut of emotion from hype (HR Block is Using AI to Do Your Taxes) to fear mongering (e.g., “The Robots Are Coming For Your Job”).
Today, five years into the journey to improve tax and accounting workflows through data, analytics and AI, I’m here to tell you that the real-world progress of this technology is far more encouraging — and productive — than either of those early storylines would suggest.
AI is not taking jobs and it is certainly not doing anyone’s taxes. What it is doing is helping tax and accounting professionals keep up with the onslaught of new tax legislation, tax code changes and global trade disruption that would otherwise overwhelm them. It is doing this by streamlining the tedious, repetitive, labor-intensive tasks and freeing up brain space for tax pros to think strategically and offer more value to their clients and colleagues.
We’re seeing this pattern play out across all areas of knowledge work — not just accounting. It has become clear through our introduction of a range of new AI-powered research tools for legal, tax and accounting professionals that the value of this technology as it evolves will be focused squarely on productivity. In short, the technology so many people feared would replace humans is actually creating opportunities for more billable hours applied to higher value/cost activities and less on grunt work that no one ever really wanted to do in the first place.
Beyond Boolean
Take, for example, the first phase of AI integration into professional-grade research tools, which has been focused primarily on using natural-language processing technology to streamline the search process. In a traditional keyword search-based environment, law librarians and junior-level accountants would spend many hours creating complicated Boolean search strings to ferret out the information they needed on a particular precedent. These search strings require the researcher to create elaborate logic that’s far from intuitive.
For example, a Boolean search for information on the latest IRS guidance pertaining to the Qualified Business Income Deduction might look something like this: ((“Qualified Business Income Deduction” OR Q.B.I. OR 199A) AND (“Tax Cuts #and Jobs Act”) AND (I.R.S. OR “Internal Revenue Service”) OR (“Section 11011”). By contrast, an NLP search for the same information would be plain English: What is the latest IRS guidance on the QBI deduction?
This is clearly not a robot doing taxes, but it is a major time-saver and workflow improvement. This is the kind of evolutionary improvement AI is bringing to knowledge work.
Tackling information overload
As we look to the second phase of AI integration into knowledge tools, we start to see the technology not only reading and interpreting text, but also surfacing meaningful insights from that text. An example from currently available technology would be an application that scans legal briefs and memoranda and uses AI to suggest other relevant cases to include or weak cases to omit based on cognitive analysis of the underlying case law.
Just as we see with NLP, this technology is not a replacement for the legal or accounting professional; it is a tool that makes their workflow more manageable and serves as a valuable gut-check for their manual case selection processes. This capability becomes increasingly valuable as knowledge workers are deluged with more data. That’s a big deal when you consider that the entire digital universe is projected to reach 44 zettabytes by 2020. To put that in perspective, there will be 40 times more bytes of data than there are stars in the observable sky.
The future of AI
As we look to the next phases of development for AI in tax and accounting and other knowledge-intensive fields, it is clear that the technology will follow a similar evolutionary path. The technology used today to surface meaningful insights from unstructured data sets is now being developed to actively monitor for anomalies and spot emerging risks. That capability will open the door to more advanced, predictive scenario planning and analytics, and, ultimately, automation of many of these capabilities.
AI skeptics will say that means job losses will not be far behind. But that narrow view of the world ignores the human capacity to do more things and evolve our capabilities along with technology. Very few knowledge workers today miss the good old days when they had to toggle through reams of microfiche to find old news articles or wait for a bike messenger to deliver important documents from a client. Throughout history, as technology has evolved, our capacity to do more with the time we’ve gained through increased efficiency has grown with it. That will continue in the AI future, opening up new levels of achievement that would have looked superhuman just a few years ago.
For reprint and licensing requests for this article, click here.