The IRS is moving deeper into AI-assisted audit selection, and the shift is already changing how tax software, CPA firms, fintech platforms, and high-complexity filers think about compliance.
The story is not that every taxpayer should suddenly expect an audit notice after filing season. The more useful point is that the IRS is building a faster, more data-driven way to decide which returns deserve a closer look. That matters because modern tax filings now leave a much larger paper trail across W-2s, 1099s, K-1s, brokerage reports, payment platforms, and digital asset activity. When those records do not line up cleanly, the mismatch is easier to find than it was in the old paper-heavy system.
According to a March report from the Government Accountability Office, the IRS had 126 active artificial intelligence use cases in its internal inventory as of June 2025, with most aimed at operational efficiency or tax compliance and fraud detection. GAO also warned that staffing cuts had weakened the agency's ability to manage those tools, including the loss of 63 employees in the Research, Applied Analytics and Statistics group who had worked on AI. That is an important caveat. The IRS is using more AI, but it is also doing that with real workforce and governance constraints.
Even with those limits, the direction is clear. Audit selection is becoming less about isolated line items and more about patterns across related data. A return can look ordinary on its own and still raise questions when third-party forms, prior-year filings, pass-through income, digital asset reports, or business deductions point in a different direction. For taxpayers, the practical risk is messy records, unexplained swings, missing forms, or positions that cannot be documented when an examiner asks for support.
The Inflation Reduction Act helped fund the modernization push, although later spending changes reduced the pool of money available to the agency. The IRS has still prioritized enforcement against high-income taxpayers, large partnerships, complex pass-through structures, and corporations where the potential revenue impact is larger. Large partnership audits are a useful example. These entities can involve hundreds of partners, layered ownership, special allocations, and cross-border issues, which makes them difficult to review manually. Machine learning and scoring models help narrow the field before human specialists spend months inside a return.
Businesses Brace
The agency has repeatedly said it does not intend to raise audit rates for taxpayers earning under $400,000 solely because of new enforcement funding. That promise should not be confused with immunity from automated matching. If a small business leaves out a 1099-K, if a contractor misses platform income, or if a crypto trader fails to reconcile proceeds, the system does not need a special campaign to notice. Basic mismatches remain basic mismatches.
Digital assets are a major part of the new compliance surface. IRS rules for Form 1099-DA require brokers to report gross proceeds for digital asset transactions effected on or after January 1, 2025, with basis reporting for certain covered transactions phased in for sales on or after January 1, 2026. That first stage creates a familiar problem for crypto users: the IRS may see proceeds before it has a complete view of cost basis, transfers, wallet history, or the full economic result. Anyone who traded across multiple exchanges or moved assets between wallets will need records that explain the story clearly.
Federal Blueprint
This is why tax technology is becoming a more serious category. The obvious products are audit-risk dashboards, document matching tools, crypto basis engines, and pre-filing review systems that flag inconsistencies before a return is submitted. CPA firms already do versions of this work manually. Software can make it faster, especially for clients with many entities, many forms, or multiple sources of income.
There is also a second layer for fintech companies. Payroll platforms, business banking apps, expense tools, creator-payment networks, and crypto exchanges sit close to the data that later becomes tax evidence. If they can help users reconcile income, classify transactions, preserve documentation, and export clean reports, they are no longer just moving money. They are lowering compliance risk.
The pressure will not stop at the federal level. State tax agencies already compare returns with IRS data, employer filings, sales records, and other government sources. As these agencies modernize their own systems, the same logic applies: better matching leads to faster questions. The gap between filing a return and being asked to explain it is likely to shrink.
The smart response is not panic. It is better preparation. Businesses should keep cleaner books, reconcile tax forms before filing, document unusual deductions, and make sure digital asset activity is tied to a defensible basis calculation. Builders should assume that compliance will become a workflow, not a once-a-year scramble. The IRS may still be working through staffing shortages and oversight problems, but the market has already received the message. Tax risk is becoming more data-driven, and the companies that help taxpayers see problems before the government does will have a clear opening.
Also read: Anthropic publishes Claude system prompts, setting new AI transparency bar • OpenAI's model timeline trends as GPT-5.5 marks efficiency shift • OpenAI's Project Mimic gaslights millions in sycophancy stress test
