We're giving a billion hours back to tax professionals

Hey there,

I’m Thomas, one of the founders of Magnetic.

People often ask why we started the company, and the answer is pretty simple. I had already seen what tax prep could look like when the right technology was built around the actual work, and I couldn’t stop thinking about how much better the process could be for firms everywhere.

Before Magnetic, I led product at Keeper, where we built an AI-powered tax prep system that filed tens of thousands of returns with a relatively small team of EAs and CPAs. What stuck with me from that experience was how different the work looked when the system actually worked. Preparing the return became much less about grinding through manual tasks and much more about review. Tax professionals could spend more of their time where they were actually valuable: talking with clients, applying judgment, and thinking through the parts of the return that really needed expertise.

Later, a friend who had just started their own firm reached out because they wanted help making their practice more automated. They had already been doing taxes for years. They had already tried outsourcing. They had already tried the standard OCR tools. And they were still buried in the same prep work, return after return. That conversation really brought the problem back into focus for me: "If I still have to review everything, it doesn't save me any time"

The issue is that most tax automation tools only solve a small part of the job. They can pull some data off some forms, and that can be helpful, but real tax prep is rarely just about extraction. The hard part is everything around it - figuring out what belongs where, reconciling information across documents, carrying things forward correctly, handling exceptions, and making judgment calls when the file is messy, incomplete, or ambiguous. That is where the work actually is, and it is also where most tools start to fall apart.

Even worse, when the software is only partially reliable, the firm still has to do a detailed review of everything. Someone still has to trace numbers back to source documents, check that carryovers were handled correctly, make sure values were assigned to the right taxpayer, and confirm that nothing subtle was missed. At that point, the software may have changed the shape of the work, but it has not really taken much of it away.

That distinction matters a lot. There is a big difference between a high-level review and a detailed review. A high-level review is what experienced preparers, managers, and partners actually want to be doing. They want to look at the return, understand the story, spot the important issues, and apply judgment based on what they know about the client. A detailed review is something else. It is prep work wearing a different label. And if you still have to do that, you have not really created leverage. You have mostly just rearranged the same workload.

That is the problem we started Magnetic to solve.

We built Magnetic for the real, messy reality of tax work. Not just the clean forms. Not just the easy cases. The actual work firms deal with every day: multiple documents, related entities, ambiguous situations, edge cases, and returns that do not fit neatly into a template. We wanted to build something that could handle more of that complexity in a way that was genuinely useful.

At the same time, we have a pretty strong belief about this market: AI by itself is not enough for tax prep. The standard for accuracy is just too high. It is not enough for a system to be mostly right if the mistakes are scattered and hard to find, because then the firm still has to check the whole thing in detail. Once that happens, a lot of the promised time savings disappear.

So from the beginning, we built Magnetic as a human-plus-AI system. Our AI does the preparation work, and then every return is reviewed by a U.S.-based EA or CPA before it goes back to the firm. The goal is not to hand firms raw output and ask them to validate everything from scratch. The goal is to return work that is already much further along, so the firm can focus on the final review, the client-specific questions, and the judgment calls that only they can make.

That is the difference we care about. We are not trying to give firms another tool that works on the easy parts and gives up on the hard ones. We are not trying to automate a thin slice of the return and hand back the uncertainty. We want to help firms get past the detailed review and move faster into the high-level review that actually makes use of their expertise.

That is where the real time savings come from. That is where firms get leverage. And that is what Magnetic is for.

— Thomas Shelley, co-founder & CEO