Harnessing the Strategic Potential of Tax Data with Machine Learning

Other than simplifying jobs, what can ML do to improve functions in data-intensive industries?


Given the rise of machine learning (ML), the disruptive potential on every industry has become commonly accepted knowledge. What’s been referred to as the merging of humans and machines is poised to transform society, politics, industrial processes, medicine, business and even war. In fact, IDC projects that 75 percent of future business software will include some form of artificial intelligence (ML being a form of AI) features within the year.

The benefits of this type of technology are most significant within data-intensive, algorithm-based industries. Advanced and predictive, some even fear that these advancements will eventually eliminate the human factor from these industries altogether. In reality, these are tools not solutions and they won’t remove the need for people, but will simply serve to augment their purposes. H&R Block, for example, emphasizes that its tax professionals are becoming more powerful thanks to robotic helpers, freeing employees from tedious tasks and low-level customer service requests that can be conducted through familiar functions like speech analytics, chatbots and text analytics. Employees can then use their time to build relationships and better serve customers.

Other than simplifying jobs, what can ML do to improve functions in data-intensive industries? To start, take a look at the world of tax. Here are three business implications it can help support:

Identifying opportunities for internal efficiencies

Take as an example a U.S.-based manufacturer of soap, which sells its product worldwide. The end product may be taxed differently in every U.S. state and in each of the hundreds of countries where they operate. Since each jurisdiction and country has a different set of rules for how that soap will be taxed, reporting becomes increasingly complex as the business expands. Yet ML tools can help by identifying similarities, differences and patterns in compliance requirements to develop the most efficient internal processes.

Keep in mind that this is a simple example, as a company that only sells one product. In the majority of cases, businesses are dealing with products that are also assessed and taxed differently adding another layer of complexity that may be difficult and time consuming to decipher for an individual, but is easy for a machine to analyze. As taxation rules and regulations change over time, ML solutions could use data to automatically adjust and refine processes accordingly.

Assessing audit risks

Tax professionals can apply these tools to data such as state revenues, number of audits per jurisdiction, and changes in taxing and reporting legislation to assess risk of reporting errors and potential audits. Global tax authorities are demanding more transparency with each piece of legislation – flawless reporting for international manufacturers is more important than ever. For a multinational organization the complexities of tax make it easy to overlook indicators or anomalies that might flag an organization for an expensive and time-consuming audit, but ML tools could actually use tax data to identify issues. Coupled with reliable tax technology software, these insights could help eliminate the frequency of audits at large organizations and even help simplify the process in the event of an audit.

Handling complexity

Tax depends on data and algorithms that rely on dynamic rules or content, requiring processing speed and power to meet increasingly aggressive tax authority deadlines and exceptions to routine algorithms (negotiated exemptions, for example).  Augmenting tax professionals with solutions that incorporate ML approaches can accelerate and enable the tax professional to focus on the most complex tax problems. Many of these concepts are already embodied in tax engines that can process an immense number of business transactions, while examining tax consequences in a consistent, audit-ready manner (machines do not get tired and make errors, for example.

Simplifying and improving routine processes

Another form of ML, referred to as RPA, or robotic process automation, consists of the ability to automate routine tasks across several tax processing functions. These tasks may include the ability to calculate tax and automatically process it though the compliance (tax return process) and treasury (money movement) functions of a company.  RPA augments routine and time-sensitive processes associated with the tax calculation to compliance process, while freeing up time for the tax professional to focus on planning. For example, while logic embedded in RPA processes automate quality assurance, they can also build analysis statements that compare current tax outcomes to previous compliance activity, highlighting potential anomalies and reconciliation opportunities.

As tax authorities globally request more detailed and timely data, augmenting tax professionals’ abilities with ML, rather than hiring more staff, can mitigate the potential for error stemming from data volume and variations, and can create a repository of information that enables better planning and audit preparedness.

These are just a few examples of the business value tax data can provide when connected with the power of ML, but the applications are endless and just beginning to be explored. Is your business using tax data to its full strategic potential?

John Viglione is the Executive Vice President at Vertex.