Soon, perhaps as early as next year's tax season, a major milestone will be officially achieved in the goal of fully digitizing and automating Federal tax filings, resulting in significantly shorter processing times and, ultimately, delivery of tax refunds. But as of today, San Francisco Bay area robotic digitization company Ripcord announced that it has entered the next stage of a new pilot program to digitize archived tax filings under an IRS tech pilot program, improving the IRS's use of data to reduce the manual process of entering and verifying tax forms and the reliance on paper filings.
Under the Pilot IRS Solution Challenge, in collaboration with the IRS's Enterprise Digitalization and Case Management Office (EDCMO), Ripcord said it has advanced into phase two, where it will complete a six-month program to securely digitize tax Form 709, which reports transfers of assets that may be subject to federal gift tax and certain generation-skipping transfer taxes.
"This is a small step towards the giant leap of making the IRS fully digital," said Ripcord president and CEO Sam Fahmy in an interview with ZDNet. "This year, we are making sure that everything is optimized, that everything works seamlessly through a multi-phase process with the IRS. So, my sense is it will be fully operational the next tax season," he said.
Over the years, the IRS has made progress in using digital technologies to streamline the tax filing process, specifically in its e-File program, but the technology to properly digitize, process and archive tax forms hasn't been there. Ripcord is using a combination of machine learning and robotics in its drive to digitize tax filings. "You need advanced machine learning, which is the ability to fully understand the context of the data," Fahmy says. What does that mean? Let's say you extract a piece of data from a tax form. To be able to check, validate and classify the data requires using machine learning technology – the same technology used in self-driving cars. "Consider that in the world of taxes, the equivalent of a road is a tax document, so we built machine learning models that become very good at understanding what's in the document," Fahmy said. So, just as a self-driving car knows everything about a road, from curves and bumps to signs and signals, the same concept applies to a tax document. Every time the machine learning model sees the document again, it understands what an amount is that's entered in a specific field and can check it back against other similar documents and be able to automate that document.
Also: Best free tax software: Free online tax filing
On the flip side, Ripcord uses robots to take paper filings and turn them into digital documents. That includes making sure that stapled attachments are removed from the filings, determining if the document is aligned the correct way, sizing the document, and so on. "It's a combination of these two technologies that allows for an end-to-end automation of the tax forms," Fahmy said. "We're able to handle paper and digital documents equally, so if it's a paper document, the first thing we do is digitize it in a very easy way for the next step, which is machine learning extraction."
So what happens to that data? Ripcord has a system in place to make it easy for the IRS to extract that data. Because new machine learning technologies can complicate and disrupt other technologies set in place in an organization as big as the IRS, Ripcord handles all of the intelligent extraction; the IRS tech staff doesn't need to be trained on Ripcord's technology, and Ripcord doesn't have to install its technology on their tech stack; it's completely non-disruptive, according to Fahmy. "We call that 'modernize-in-place' so they can keep everything in place but at the same time modernize the process without having to change their own systems and staff," he says. What's more, Ripcord's system adheres to the highest government security standards.
Also: Income tax guide for tax season
The lessons learned from digitizing Form 709 will then be applied to other IRS tax forms, ultimately to the point where all IRS forms will be digitized and archived using machine learning technology, at least in theory. It's a lofty goal but one that has value. The Covid pandemic showed that the IRS can't always rely on people to process forms as delays mounted when it came time to issuing refund checks. And, because more Americans had to file their tax forms remotely during the pandemic, there's a greater need to automate the filing process.
"This is a small step; we don't want to claim that we are tasked to completely modernize what is left to me modernized at the IRS," Fahmy says, noting that getting the contract to help the IRS was a highly competitive process. "It's a small step that we hope and plan to build on toward that path to full digitization. There have been more initiatives [by the government] to seek out new technologies and smaller companies like ourselves. We're a small company of 100 to 150 people. It's fantastic to see that our government has open arms to companies like us," he said.