WHITE PAPER
Artificial intelligence, tax preparation, and public accounting’s future
1040 tax automation technology is evolving. Now that advanced engineering has hit a ceiling, the next leap in tax preparation is artificial intelligence. Since 2002, SurePrep has led the innovation curve in the tax automation space. SurePrep’s vice president of AI has assembled a dedicated Innovation Team to integrate AI into our tax preparation technology.
What is AI and why is it relevant to tax preparation?
AI is the theory and development of computer systems that are able to perform tasks normally requiring human intelligence. Several underlying processes can comprise AI, including:
- Machine learning (ML). This is a computer’s ability to algorithmically teach itself tasks by absorbing data and observing processes.
- Computer vision (CV). This process allows a computer to visually scan images and documents and recognize patterns.
- Natural-language processing (NLP). The ability of a computer to recognize patterns in text and understand how different areas of the text relate to each other. For example, AI trained in tax terminology should realize that a form titled “Property Tax Statement” and a form titled “Real Estate Property Tax Bill” may be related.
These processes rely on pattern recognition. The more quality data you feed your AI, the more samples it can draw from to develop pattern detection — the same way humans learn. This is good news for firms preparing 1040s because the tax world is made of repeating patterns. Think of standardized forms, prior-year data, and repetitive tasks. Our industry is uniquely positioned to reap the benefits of AI.
Historically, effective AI was difficult to develop because it required immense computational power. In recent years, advancements in cloud computing have brought us closer to accurately replicating human-like decision-making at scale.
This white paper explores how AI can deliver unprecedented savings and efficiency by eliminating menial tasks throughout the 1040 process. We’ll highlight real AI solutions being applied at SurePrep and reveal the unique advantages that position our Innovation Team as leaders of tax preparation AI.
How AI is impacting optical character recognition
Optical character recognition (OCR) already utilizes CV to read letters, numbers, and symbols and convert them into data. Scan-and-populate solutions use OCR to extract data from standard tax documents and export that data to tax software. Automating data entry with OCR boosts efficiency and reduces costs.
Without AI, this technology has a fixed ceiling. To extract data, a computer must be told where to “look.” For example, a taxpayer’s Social Security number (SSN) is always in the upper left-of-center on a W-2. OCR software needs fixed instructions to understand that numbers in this location are always SSNs. These instructions are called “OCR templates,” which are grids manually created by humans.
Standard documents are easy to build templates for because information is always in the same place. One W-2 template can serve all W-2s. But, some tax documents have thousands of variations, such as property tax statements, which vary based on county. These inconsistent formats are called “non-standard documents.” Before AI, there wasn’t an efficient way to make non-standard documents compatible with OCR.
SurePrep uses NLP to recognize more tax documents
SurePrep’s Innovation Team is hard at work on OCR enhancements that incorporate NLP. This technology will be used to create dynamic OCR templates that help computers locate data anywhere on a document. Instead of relying on a grid, a program trained with NLP can interpret contextual clues to find information. For example, when a human sees the phrase SSN or identification number, followed by a number in XXX-XX-XXXX format, they understand this is an SSN. NLP empowers AI to make similar judgment calls. 1040SCAN already recognizes four to seven times as many documents as the alternatives, but AI is making template limitations obsolete.
If this change sounds transformative, it is. The more tax documents your firm can automate, the fewer hours your staff will spend on data entry. Decreasing the overall hours per return can increase profit margins and free up billable hours for value-added work.
How AI is enhancing OCR verification
It’s best practice for humans to verify data captured by OCR. Verification takes a fraction of the time your staff would have spent on data entry. The more accurate the scan, the faster verification goes. An accurate scan also decreases the chances mistakes will carry forward to the preparation or review phases.
OCR currently relies on CV alone. SurePrep is building a more holistic AI layering in ML. Using prior-year data, OCR programs can be trained to make judgment calls about what they see. For example, the value $120.76 might have scanned with a faded period. Traditional OCR would simply read that value as $12,076 and move on. OCR enhanced with ML has learned from historical data. If the dollar amount in this field is in the hundreds in 99% of similar cases, a sudden jump to the tens of thousands is unlikely. AI-enhanced OCR can infer the missing period. In other words, the program catches mistakes before a human verifier.
In addition to training our OCR with ML, the SurePrep Innovation Team is refining the software’s CV. The goal is OCR results requiring minimal verification time. Currently, 1040SCAN’s patented, AI-powered technology auto-verifies OCR data for 65% of standard documents. That said, AI can never replace human verification 100%. Vendors that suggest otherwise are misrepresenting the technology.
Only SurePrep combines AI, auto-verification technology, and verification outsourcing
You don’t have to wait for advancements in AI to decrease verification time. In 2019, SurePrep patented new technology that auto-verifies data extracted from native PDFs. Native PDFs processed with 1040SCAN are the only exception to the human verification rule for OCR. 1040SCAN reads the metadata layer in the PDF and compares it to the OCR results with better-than-human accuracy.
All digital tax documents imported directly from financial institutions are in native PDF format. SurePrep firms gathering documents with TaxCaddy Smart Links report that over half of their client documents are native PDFs. That’s over half the verification work, already done.
If you want to eliminate verification work completely, consider selective outsourcing. Our trained staff perform your verification in secure facilities so your firm can focus on preparation and review. Because our staff use SurePrep technology, the benefits of our upcoming AI enhancements will be passed on to you.
How AI is reducing preparation and review time
Artificial intelligence can’t learn tax preparation, but it can learn to anticipate a preparer’s needs. This new generation of work-paper management software is learning as it’s used.
Do It Like Last Year
Do It Like Last Year (DILLY) is a SurePrep project training AI to perform actions in the binder based on a client’s prior-year returns. This action is useful for documents that change very little from year to year. For example, a client may submit a property tax statement for the same house every tax season. The only field your preparer needs to reference is the amount due. DILLY recognizes the document from previous years, remembers the preparer’s past actions, and preemptively references the correct field. All your preparer needs to do is confirm.
In the same way, DILLY can help preparers sort non-standard documents. 1040SCAN already sorts standard documents into an index tree that follows the flow of the return. Non-standard documents populate the thumbnail panel, where preparers can click and drag them into the right folders. DILLY uses this folder information for auto-indexing, remembering how your preparers sorted those documents to perform that organization automatically.
This automation was not possible before AI because document layouts often change from year to year. DILLY uses CV and NLP to recognize documents, even if they change. To train DILLY, your preparers should simply process 1040s the way they normally would. After one tax season, DILLY will have enough information to begin performing useful tasks, becoming more responsive and intuitive as time goes on.
Automatic K-1 Reclass
1040SCAN is the only OCR software that recognizes state K-1s. In 2020, we expanded recognition to K-1 supplemental pages and released a K-1 reclass tool. The K-1 reclass interface in SPbinder helps preparers sort K-1 data into the appropriate income and expense categories.
Preparers are accustomed to manually selecting where to reclass each line item from a picklist. Now, an AI engine can pre-select suggested destinations. Unlike DILLY, this AI engine is continuously learning ways to reclass as it observes human actions and makes suggestions even where no prior-year data exists.
Fewer errors for reviewers
When AI is implemented throughout the scanning, verification, and preparation phases of the 1040 tax process, human error will decrease. Fewer errors mean a faster review process. Because reviewer time is expensive, decreasing review time is the most effective way to increase your profit margin per return.
Will AI threaten tax preparation jobs?
Short answer — no.
Medium answer — AI may rescue the tax preparation industry. Employee burnout, staffing shortages, and high turnover plague public accounting. Offloading menial, repetitive tasks to AI and automation can only benefit tax professionals. Job satisfaction will increase when employees can focus on work that honors their expertise. Long hours during the busy season will drop when computers handle the low-skill labor. These positive changes may attract more candidates to the profession and convince veterans to stay. AI will never replace a college-educated CPA.
Why SurePrep is uniquely positioned to bring AI to tax preparation
SurePrep is the industry leader in 1040 tax automation. Our scan-and-populate, work paper management, and client collaboration solutions integrate with leading tax software to streamline the 1040 process.
We achieved our market position by emphasizing technological development and innovation. SurePrep’s OCR technology recognizes four to seven times as many documents as alternatives. We own patents on OCR auto-verification technology. SurePrep is the only vendor with established mobile apps for both taxpayers and tax professionals — with a four-and-a-half star rating. We’re also the only vendor to offer advanced preparation and review tools like K-1 reclass and digital lead sheets.
We are able to dedicate our resources to technological advancement because tax automation is all we do. Other vendors in the space only develop tax automation as ancillary solutions to their tax software.
The secret ingredient in good AI is data, data, data
A top reason AI adoption can bottleneck is a lack of data or data quality issues, according to a 2020 survey by O’Reilly. It makes sense — data is the backbone of ML. Businesses without a strong data emphasis lack the foundation to train AI.
SurePrep is the only vendor in the tax automation space with a data advantage. Our solutions operate as software as a service (SaaS) and our preparers use SurePrep software for outsourced returns. This means we have more than 10 years of anonymized data from our staff. Competitor software is installed locally, which means all useful data is lost. To put it another way, SurePrep’s AI is about to earn its doctorate, while others in the space are enrolling in undergrad. There is simply no substitute for data generated over time, and no way to close a 10-year gap.
Ironclad privacy and security policies are part of the reason why 33,000 tax professionals trust SurePrep. The data that trains our AIs is completely anonymized. For example, AI can learn that K-1s are in the income category because preparers always drag K-1s to the income folder — but it does not remember the contents of any individual form.
Next steps for AI and tax preparation
This is an exciting moment in tax preparation. AI has rounded the corner and is now performing tasks at scale with better-than-human accuracy. The opportunities for automation are limitless. SurePrep customers experienced the first wave of AI enhancements in the 2021 tax year.
SurePrep is the leader in 1040 tax technology, and we feel a responsibility to lead the curve in AI. Dynamic OCR templates, enhanced verification, refined OCR, DILLY, and automatic K-1 reclass are some of the AI-driven technologies available to SurePrep customers.
With AI, your firm can reduce redundant tasks and boost efficiency throughout the 1040 process.
Experience the entire SurePrep end-to-end 1040 process from start to finish. For further details, contact us to schedule a one-on-one demonstration.
Start saving over 90 minutes per 1040 return
If outdated 1040 tax technology isn’t doing your firm any favors, you need SurePrep — the most powerful tax automation software on the market