The United States Tax Code has either been revised or amended over 4,000 times in the last decade alone, according to an estimate from eFile. Some multinational corporations, wealthy filers, and their advisors take advantage of the movements to shield – in some cases aggressively – taxable income from regulators. They’re banking that overworked staff won’t find abuses in the mountains of disconnected data they’re tasked with reviewing.

It’s a cat-and-mouse game. A data chase where advisors find and exploit weaknesses in the tax code as revenue agents, criminal investigators, and other investigating data analysts race to plug holes or prevent and prosecute fraud.

How could it be like this? How can crucial data hide in an era where anyone with an internet connection can get the computing power of an ultrafast server for a few dollars at a time?

Raw information rarely provides a full picture. Only when human analysts and investigators manually connect data from different sources can it be properly analyzed. And it’s only at that point that it is possible to spot the games that some filers and their advisors play.

In this blog post, we dive a bit deeper into how this evolving cat-and-mouse game is played, how it affects agents, investigators, and analysts, and the steps regulators are taking to bring relief to and ease the burden of staff caught in the chase.

The surprising prevalence of paper filings

In 2021 alone, the United States Treasury handled 257.7 million tax filings and supplemental documents. Of those, 21.8% were filed by paper and scanned into a data warehouse by hand. That is over 56 million filings that includes:

• Over 31 million for individual, estate, and trust taxes
• Over 1.6 million for business taxes
• Over 16.6 million for employment taxes
• Over 6.8 million supplemental documents

Importantly, all this paper – amounting to at least tens of millions, if not hundreds of millions of pages – had been received and scanned into a data warehouse by hand. Exactly zero context or information about the data in those pages would be added at the time of the scan. That analysis (if needed) would come later, at the discretion of an agent, investigator, or analyst.

If this sounds like a painful process, rest assured that it can be.  Each tax year agency leaders select cases to hand out to investigators, agents, and analysts. The discovery process that informs their choices is often complicated by paper filings.

They force a question: is there enough here to warrant assigning a team to sift through mountains of raw material and make connections by hand? If all goes well, the team might eventually uncover a filing mistake or even fraud. Or not. In each case, investigating tax compliance starts the same way: with a Tax ID number.

For example, a criminal investigator assigned to determine whether a multinational company is underreporting profits to avoid taxes would start by entering a Tax ID to recall as many documents and records as possible related to the company. Judging the contents of those documents, and – more importantly – their relation to each other and outside entities can consume hundreds of hours of manual research.

That’s when volume becomes a serious problem. If no patterns of obvious noncompliance surface after weeks of sifting through digitized files, investigators may be tempted or even coerced to move on to the next case. It’s a frustrating cycle of incomplete investigation after incomplete investigation and burnout is common.

Better tools for the chase

Knowing all this, it’s easy to understand why some multinational businesses and ultra-high-net-worth individuals might choose to have their advisors file by paper. It’s just easier to hide. To tip the game in their favor, regulators are investing in knowledge graphs and artificial intelligence to improve the process of ingesting and adding context to filings.

What’s this mean practically? Revisit the workflow of the criminal investigator. Work on the case still begins with a search of filed documents tied to a single Tax ID. But now, instead of having to make all the relevant connections to other Tax IDS, entities, and documents by hand, it’s all automated and entered into a knowledge graph. Crucial relationships are easy to spot, and so is potential noncompliance.

Consider the case of two wealthy individuals who choose to set up an entity together where one party contributes assets and the other contributes cash. Then, in the same tax year, they choose to dissolve the entity with each taking the assets apportioned to them. A transfer of assets that probably should have been classified as a sale goes untaxed.

The underlying knowledge graph might catch that the two individuals actually have a long-standing relationship due to their joint involvement in a charity. The system could flag the potential unfair dealing and point the investigator to other documents that fill out the story and produce a fairer result.

GraphGrid, a powerful Graph + A.I. Platform, is helping Treasury investigators, agents, and analysts do this work today. To learn more, download and read the case study, “How to Hunt a Cheat.”

Get Started for FREE – Schedule a Solutioning Session

Schedule a free solutioning session with a GraphGrid professional. Tell us about your highest impact data-intensive challenge or opportunity so we can align our platform and professionals to accelerate your graph-enhanced solution journey.

In the meantime, feel free to download GraphGrid at our Freemium license level and check out our step-by-step tutorials, video tutorials, or read more about GraphGrid in our use cases, blogs, and documentation.