Magnetic Tax Is Automating the Work CPAs Hate Most

The accounting profession is in crisis, and most people outside the industry don't know it.

There are not enough CPAs. The pipeline is drying up. Fewer students are entering accounting programs. Experienced practitioners are retiring faster than new ones arrive. Firms are overworked, understaffed, and turning away clients they should be serving.

During tax season, 80-hour weeks are not the exception. They are the norm. Partners burn out. Staff turn over. And the work that consumes the most hours isn't the strategic advisory work that CPAs are trained for. It's data entry.

Reading documents. Extracting numbers. Typing them into tax software. Checking for errors. Repeating thousands of times across hundreds of clients.

This is the problem Magnetic Tax is solving.

The Talent Crisis Is Real

The numbers tell a stark story. The AICPA has been sounding alarms for years about the declining CPA pipeline. Accounting graduates are down. CPA exam candidates are down. Meanwhile, tax complexity only increases. New regulations, new forms, new filing requirements.

The result is a supply-demand imbalance that gets worse every year. Firms can't hire enough people. The people they do hire leave for better-paying roles in tech or finance. The ones who stay are buried in work that doesn't require their expertise.

A senior CPA spending hours entering K-1 data into tax software is like a surgeon spending the morning filing paperwork. It's a misallocation of skilled labor on a massive scale.

Firms have tried to address this with offshore outsourcing. Send the data entry work overseas, where labor costs are lower. But outsourcing brings its own problems: quality control, turnaround time, communication overhead, and security concerns around sensitive financial data.

What firms actually need is not cheaper labor. They need to eliminate the labor entirely.

What Magnetic Tax Does

Magnetic Tax built an AI agent that scans client tax documents and inputs the extracted data directly into legacy tax software. The system achieves over 90% accuracy across a wide range of document types.

This isn't OCR with a nice interface. It's a full automation pipeline.

The agent handles everything from standard W-2s and 1099s to complex K-1 partnership schedules, handwritten notes, and unconventional document formats. It reads the document, understands the context, extracts the relevant fields, maps them to the correct lines in the tax software, and enters the data.

The "directly into legacy tax software" part is critical. CPA firms use established tax preparation platforms like UltraTax, Lacerte, ProSeries, and Drake. These are not modern cloud applications with open APIs. They are desktop-installed software with complex interfaces built over decades.

Magnetic Tax doesn't ask firms to switch platforms. It works with what they already use. The AI agent interacts with the existing software the same way a human would, but faster and without fatigue.

This is the key insight: meet firms where they are. Don't ask them to change their entire workflow. Automate the painful part of their existing workflow.

Why 90% Accuracy Changes Everything

A 90% accuracy rate might sound imperfect. In reality, it transforms the economics of tax preparation.

Consider the current workflow: a human reads every document, enters every field, and (hopefully) catches their own errors. The accuracy rate of manual data entry by tired humans working overtime isn't 100% either. Studies consistently show that manual data entry error rates range from 1% to 5%, and that's under normal conditions. During tax season crunch time, the rate climbs higher.

With Magnetic Tax, the agent handles the bulk entry, and the CPA reviews the output. Instead of entering 100% of the data manually, the CPA verifies 100% of the data with corrections needed on roughly 10% of fields.

The time savings are enormous. Verification is dramatically faster than entry. A CPA can review a pre-filled return in a fraction of the time it takes to create one from scratch.

This shifts the CPA's role from data entry clerk to quality reviewer. They spend their time on what they're actually trained for: identifying issues, applying judgment, advising clients.

The Document Challenge

Tax documents are a nightmare for automation.

They come in every conceivable format. Crisp PDFs generated by financial institutions. Photographed receipts. Handwritten notes from elderly clients. K-1s from complex partnership structures with dozens of line items. Foreign income statements in non-standard formats. Brokerage summaries with hundreds of transactions.

Traditional OCR struggles with this variety. It works well on standardized forms in clean formats. It fails on the messy reality of what clients actually send to their accountants.

AI changes this. Modern vision and language models can interpret documents contextually. They don't just read characters. They understand what the document is, what information is relevant, and where it belongs in a tax return.

A handwritten note saying "rental income $2,400/mo, 123 Main St" isn't just text. It's a data point that belongs on Schedule E, Line 3, with associated property information. The AI can make that inference. OCR cannot.

Thomas Shelley, Magnetic Tax's founder, previously served as Head of Product at Keeper, where he built AI tax filing systems and OCR scanning technology. He understands both the technical challenges and the practical realities of tax document processing. Patrick Fay, co-founder and first engineer at FunCraft, brings the engineering velocity needed to ship and iterate quickly.

The Economic Model

The economics for CPA firms are straightforward.

A staff accountant costs $60,000 to $90,000 per year. During tax season, firms also hire seasonal workers, pay overtime, and engage outsourcing services. A mid-size firm might spend hundreds of thousands annually on the labor that goes into data entry.

If Magnetic Tax can replace even a portion of that labor, the ROI is immediate and measurable. Firms don't need a business case meeting. They can calculate the value in minutes.

Better yet, the real value isn't cost savings. It's capacity.

A firm that currently serves 500 clients during tax season isn't limited by expertise. It's limited by data entry throughput. If Magnetic Tax doubles the number of returns each preparer can handle, the firm can serve 1,000 clients without hiring. That's revenue growth without proportional cost growth.

In an industry with a structural talent shortage, capacity is the binding constraint. Anything that relaxes that constraint directly translates to growth.

Why Legacy Software Integration Is the Moat

The decision to integrate with existing tax software rather than replace it is strategically brilliant.

CPA firms are famously change-averse. Their workflows are built around specific platforms. Their staff are trained on specific tools. Their clients' data lives in specific systems. Asking them to switch platforms is a non-starter for most firms.

By working with legacy software, Magnetic Tax eliminates the adoption barrier. There's no migration. No retraining. No disruption to existing workflows. The AI agent simply does the data entry that a human would have done, using the same software the human would have used.

This also creates a defensibility layer. Building reliable AI agents that can navigate complex desktop applications is technically challenging. The interaction patterns, error handling, and edge cases of each platform require deep integration work. A competitor can't just replicate this by wrapping an API.

The firms that adopt Magnetic Tax don't need to change anything about how they work. They just do less of the worst part.

The Timing

Several factors make this the right moment:

The CPA talent shortage is at its worst point in decades. Firms are desperate for solutions. They're willing to try new technology in ways they wouldn't have been five years ago.

AI vision and document understanding models have reached the quality threshold needed for reliable document processing. The 90%+ accuracy rate wasn't achievable with earlier technology.

The IRS is increasing complexity, not reducing it. New regulations around digital assets, international income, and pass-through entity taxation create more data entry work every year. The problem is getting worse.

And CPA firms have seen AI work in adjacent domains. They're no longer skeptical about whether AI can read documents. They want to know when they can use it.

The Broader Pattern

Magnetic Tax fits a thesis about where AI creates the most value: automating skilled-adjacent labor in talent-constrained industries.

CPAs are skilled professionals doing unskilled work because no one else is available to do it. The AI doesn't replace the CPA. It replaces the data entry that the CPA shouldn't be doing in the first place.

This pattern repeats across professions:

• Lawyers spending hours on document review instead of strategy.
• Doctors spending time on charting instead of patient care.
• Engineers spending time on boilerplate instead of architecture.

In each case, the professional's time is being consumed by work below their skill level. AI automation doesn't threaten the profession. It frees it.

For CPA firms, Magnetic Tax represents something they haven't had before: a way to grow without hiring. In an industry where hiring has become the binding constraint, that's not just a convenience.

It's a structural advantage.

The firms that adopt this technology early won't just be more efficient. They'll be able to serve more clients, generate more revenue, and attract better talent by offering work that's actually interesting.

And in a profession that's struggling to attract the next generation, making the work better might matter as much as making it faster.