Business Idea Audit
AI Firm That Files Your Taxes
There is something here but it needs serious rework. Do not quit your job yet.
This is a proven, gigantic market with a clearly defined wedge, but the incumbents are not weak or asleep. Both giants and funded startups already ship the exact 'upload docs, AI does it, return gets filed' experience, so any edge has to be sharper execution for a narrow niche rather than a new category.
DEMAND — Does anyone actually want this?
13/20Nobody questions the demand: the U.S. individual tax-prep market is roughly $23 billion (cited around FlyFin's coverage on PitchBook and PR Newswire), and people obviously pay to make tax pain go away every single April, which makes urgency and willingness-to-pay strong. But search traffic for the angle runs through branded terms like 'TurboTax Full Service' and 'FlyFin', not an unmet 'just file it for me' gap, so the pull is for existing products, not a missing one. FlyFin's reported 270,000+ users and 95% satisfaction prove the specific done-for-you angle already has a happy customer base. I could not surface raw Reddit threads venting the exact pain through search, so the unfiltered community-complaint signal is thinner than the market-size signal.
COMPETITION — Who's already doing it?
7/20The market is validated to the maximum and that is exactly the problem under an inverted-U: it is saturated by both giants and funded startups. Intuit's TurboTax now markets agentic, 'done-for-you' filing in as little as two hours, H&R Block shipped AI Tax Assist built on Azure, and Thomson Reuters' Ready to Review automates source documents through Q1 2026. On the startup side FlyFin (about $8M raised, CPAs file start to finish), TaxGPT (full filing for $50), Instead (IRS-approved across 48+ jurisdictions), Filed (2025 AICPA/CPA.com accelerator) and Prime Meridian all occupy the identical wedge, so the exploitable gap is tiny and the differentiation has essentially no moat against incumbents with the data, brand and capital.
REVENUE — Where's the money?
13/20People already pay real money for this exact service: TurboTax Full Service starts around $129, independent CPAs charge roughly $300-800 and chain preparers $200-600 per the 2025 price guides, and FlyFin reports about 93% gross margins, so the unit economics can be healthy. The trouble is pricing power: the entire pitch is 'cheaper than a CPA', and TaxGPT has already anchored full filing at $50 while TurboTax bundles a near-free path, so this wedge competes mostly on being the cheapest, which erodes the margin advantage. The revenue model is clear and seasonal, and a thin operation could reach revenue without massive scale, but the price war caps how good those margins stay.
FEASIBILITY — Can you actually build this?
9/20The software is buildable with today's tools, and FlyFin, TaxGPT and Instead prove an LLM-plus-e-file pipeline works, with relatively low capital needs outside compute and compliance. The hard wall is that you cannot actually run a pure-AI 'firm delivering filed returns': under Treasury Reg 301.7701-15(f) AI only gets the mechanical-assistance carve-out, so a human PTIN holder must sign as the paid preparer and carries Section 6694 penalty liability for errors, which the GAO flags as common among unenrolled preparers. That forces you to staff licensed CPAs or EAs exactly like FlyFin does, plus secure IRS e-file authorization, making licensed-human supply and accuracy liability the real bottleneck, not the model.
TIMING — Is now the right time?
16/20The 'why now' is genuinely strong: agentic AI tax filing went mainstream in 2025-2026, with Intuit launching an all-in-one agentic consumer platform, H&R Block deploying AI Tax Assist, and Thomson Reuters expanding Ready to Review's automated source documents through Q1 2026. The enabling tech (document-reading LLMs plus established e-file rails) is clearly ready. The catch is that this same window is why every incumbent already shipped the feature, so the trend is accelerating away from a new entrant rather than toward one, and the IRS regulatory posture on preparer liability is static, not opening up for automation.
The Honest Take
“Here's the coffee-shop truth: the demand is real and the timing is hot, but you are about two years late to a party that Intuit, H&R Block, FlyFin and a $50 TaxGPT are already throwing. The thing you are not seeing is that 'AI files your taxes cheaper than a CPA' is no longer a wedge, it is the table everyone is already sitting at, and the one part you cannot automate away, a licensed human who signs the return and eats the penalty when the AI is wrong, is exactly the cost that kills your price advantage. You would be entering a saturated market with no moat, fighting funded incumbents on price, while still having to hire the CPAs you were hoping to replace. If you do this, do not build a generic filer; pick one ugly, high-pain niche the giants ignore (messy multi-state freelancers, crypto traders, expats with FBAR/foreign income) where a human-in-the-loop CPA and a sharp AI actually beats TurboTax, and win that slice before anyone notices.”
What To Do Next
Today, sign up and run your own real return through TurboTax Full Service, FlyFin and TaxGPT back to back, and write down exactly where each one fails or feels painful, that gap list is your only possible wedge.
Pick one narrow, underserved niche (multi-state 1099 freelancers, crypto/DeFi traders, or US expats with foreign income) and go read 20 recent posts in that group's Reddit or forum to confirm the incumbents genuinely fail them.
Call or email two CPAs or enrolled agents this week to price what it costs to put a PTIN holder behind every filed return, because that human-signing cost, not the model, decides whether your 'cheaper than a CPA' math actually works.
Want to test your idea?
Get this same honest, scored audit for your own business idea in 60 seconds — free, no signup.
Test my idea →