Why We Automated Tax Season

There are scores of things that humans can still do better than computers. But mechanical computers were specifically built to crunch numbers and they’ve only widened the gap on humans in the almost hundred years since. These machines, at their core, are created to demolish problem-sets of rules and numbers. Which is why it seemed obvious for us to create a fully automated tax service: neo.tax.

In many ways, taxes are the perfect dataset: they’re a collection of deeply specific, often archaic rules and numbers. For example, the IRS's Publication 970, Tax Benefits For Education is 86 pages long — most families end up missing out on their maximum tuition writeoff. Humans can get lost in the minutia, but a well-trained algorithm can breeze through the data and quickly find the maximum credits for a client.

For many startup founders and small business owners, April is the worst month of the year. The tax code is often purposely obscured — it’s why it takes years of schooling and an exam to become a CPA. But what if a machine-learning algorithm could be taught the entire tax code? And what if it was built and trained by the former Head of Tax at Kruze, who also spent seven years at the IRS?

With Neo.Tax, we’ve started by solving a relatively simple problem-set: the R&D tax credit. The credit, expanded in 2016 to better serve startups, is far-too-rarely claimed by young tech companies. The 2016 switch meant that the R&D costs could be deducted from payroll tax rather than income tax, which meant, suddenly, thousands of early, pre-revenue startups could get relief for their biggest burn area.

We built an algorithm that could pore over a startup’s accounting and payroll information and instantly identify the maximum R&D credit filing. From there, it was simple to automatically create the paperwork for the filing and send it to the client. Computers were built to scan through mountains of data and deliver a clean, concise and correct solution. Neo.Tax has brought that solution to the R&D tax credit.

Because the Neo.Tax algorithm is automated, the entire process of filing for the R&D credit takes about 15 minutes for our clients. Human or half-human/half-automated services charge between 15% and 25% for the same service — for a problem-set like this, a human touch makes the process take longer and cost more, without actually making the results more accurate. Because our process is fully automated (though each return is double-checked by our ex-IRS cofounder), we thought it only fair to charge just 5% of what you save. Our clients are often pre-revenue and the credit can extend the runway for burgeoning companies.

Taxes are often the worst part of running a business. They’re also the ideal problem-set for computers to power through. That means they’re the perfect thing to automate.

Neo.Tax is starting with the R&D Tax Credit, but we plan to continue honing our algorithm. Eventually, much of tax season can be automated. A mess of numbers and pages of rules? This is what computers were built for.