Imagine you’re on a business trip. You pull off the highway, buy a donut and coffee, then head to your next meeting. Sounds like a routine expenditure, but it’s still important to your company’s tax function. The good news: There’s now an opportunity to use artificial intelligence to automate the tax treatment of this expense.
Simply put, AI is a set of approaches that enables software to adapt automatically to input without explicit instructions. It combines machine learning technologies, large datasets, and the use of sensors to mimic human behavior, giving AI the potential to revolutionize the way we live and work.
The case for using AI is compelling. There is greater reliability in decision making due to the consistency and transparency of computer functions. Additionally, AI’s speed and capacity to process large volumes exceed what humans can reasonably accomplish.
AI’s key benefits for tax can be summarized as follows:
Predicting: One primary function of AI and machine learning is prediction — and forecasting is the simplest form. How does it work? The most basic forms include regression, rolling averages, exponential smoothing, and even the infamous “same as last year” (SALY) approach. More advanced methods are capable of detecting trends within tax filing periods or quarterly, monthly, or annual reporting trends, including increases or decreases in cash tax payments and changes in accruals and prepaid accounts.
The tax function can access the power of AI’s predictive functionality to facilitate more accurate forecasting for tax calculations, including the forecasted effective tax rate for interim financial reporting. Scenario tax planning can be accomplished with better data, greater speed, and better accuracy. The ability to forecast more efficiently and effectively can improve the quality and accuracy of tax financial results and provide future insights needed for decision making.
Clustering/Classifying: AI and machine learning functionality can identify clusters of data points faster, with higher accuracy, on vaster quantities of data, and with a higher number of variables. This capability has significant applicability for the identification of tax-sensitive accounts within financial systems to facilitate proper income tax return compliance and reporting treatment.
For instance, clustering and classifying features can more easily categorize expenses qualifying for the research and development (R&D) credit and other tax incentives. In addition, because of the significant volume of transactional data, these AI features can have a material impact on indirect tax (e.g., sales and use taxes) calculations and reporting.
Expert support systems — tools and techniques: In certain AI systems, decision rules are encoded and later queried for accurate responses. A knowledge base is developed with relevant information from human tax experts and automation applied for processing. This capability can be used by the tax function to automate the calculation of complex tax adjustments.
Data sets can be created to capture appropriate general ledger account details with rules applied to calculate tax amounts. A simple example is the treatment of meals and entertainment expenses where several general ledger account details are aggregated based on tax requirements, with the 50% deductibility limitation then applied for tax purposes. Similar treatment can apply to foreign tax credits or fixed asset depreciation deductions where relevant data are spread across various accounts and financial systems and subject to specific tax rules and calculations.
Robots no longer are limited to manufacturing, R&D, and other “high-tech” functions and processes. For finance and tax, AI and machine learning capabilities can produce “bots” that mimic human behavior and, with the appropriate rules applied, can perform routine tasks, freeing up humans for higher-risk, higher-value activities. Robots can now perform automated data entry, integration of data from multiple systems, repetitive tasks, reconciliations, and data validation or control checks related to direct or indirect tax tasks.
AI and machine learning prototype systems already are being used by taxing authorities, academic researchers, businesses, and PwC. Within the next few years, we can expect greater deployment of advanced tax analytics — and related benefits. Technology may facilitate tax research through natural language processing (NLP) where queries are made by human voice, and AI interprets the question, performs document searches, and delivers a response. We believe AI will accomplish three main things — reduce risk, raise value, and help drive a cost-effective approach to core reporting.
In the not too distant future when you stop for a donut and coffee, pull up the bot on your phone, have it order “the usual,” and allow the machine learning application to predict the type of donut and size of coffee you want. AI applications can have the same powerful impact within your tax function. The future is now.
Michael Shehab is tax technology and process leader for PricewaterhouseCoopers.