Observing the federal government reduce its human workforce while investing heavily in artificial intelligence is unsettling. However, that is precisely what is taking place at the IRS, where the Trump administration’s fiscal 2027 budget proposal calls for a more algorithm-driven, leaner agency with fewer auditors.
The figures paint a striking picture. The administration wants to cut the IRS’s current budget by $1.4 billion to $9.8 billion. $1.2 billion is set aside in that smaller budget for technological advancements, such as AI-powered tax enforcement instruments. The White House presents this as emphasizing customer service and streamlining operations. Opponents question whether it is possible to genuinely enhance customer service while reducing personnel and placing all of your bets on technology that is still largely unproven at scale.
| Category | Details |
|---|---|
| Budget Proposal | Fiscal Year 2027 Federal Budget |
| Administering Authority | Trump Administration / Office of Management and Budget |
| IRS Proposed Budget | $9.8 billion (down $1.4 billion from FY 2026) |
| AI Investment | $1.2 billion for technology modernization and AI-driven enforcement |
| Workforce Reduction | 27% reduction completed in FY 2026; additional 17% enforcement cut proposed |
| Total Nondefense Discretionary Cut | $73 billion (10% reduction) |
| Reference | White House Office of Management and Budget |
It’s important to keep in mind the past. Last year, the IRS lost 27% of its workforce, ending what the administration referred to as the “staffing surge” of 20,000 new hires during the Biden administration. If you can find an IRS office that is still open today, you will see that the ranks are dwindling. 22% of the organization’s taxpayer services staff departed last year. The new plan calls for an additional 17% cut to the enforcement division, which actually generates revenue by apprehending tax evaders.
Whether artificial intelligence can truly fill the void left by thousands of departing human workers is the question that hangs over all of this. The White House maintains that it can, citing “technology improvements” that will enable the agency to deliver better services while guaranteeing that tax laws are applied equitably. “Fairly administered” is a powerful phrase, particularly from an administration that has accused the IRS of being “weaponized against the American people.”
However, there is a tension that is difficult to overlook. Recently, the Government Accountability Office issued a warning that significant staffing cuts at the IRS “could greatly affect its ability to use AI.” To put it another way, the very technology that is meant to replace them requires skilled individuals to design, implement, and manage. It’s similar to dismissing your mechanics before the self-driving car is complete.
The administration is adamant that AI is the way of the future for tax enforcement. Theoretically, machine learning algorithms are far faster than human auditors at scanning millions of returns for patterns of fraud or evasion. They don’t complain to their union or take lunch breaks. They don’t require retirement benefits or health insurance. The reasoning is almost alluring from the perspective of pure efficiency.
However, anyone who has worked with automated customer support systems is aware of how easily that effectiveness can give way to annoyance. After failing to meet its hiring targets, the IRS already had to place 1,500 IT and HR staff on 120-day leave this year to cover front-line taxpayer service tasks. Those were people who were proficient in payroll processing or coding, and all of a sudden they were taking calls regarding standard deductions and W-2 forms. When the agency continues to lose employees, it is difficult to see that situation getting better.
The administration’s actual priorities are made clear by the larger budget proposal. A few areas receive protection or even increases, while the majority of civilian agencies experience drastic cuts—the Small Business Administration, EPA, Labor Department, and NASA all saw reductions of more than 20%. For example, Veterans Affairs receives nearly $145 billion in discretionary spending, which includes $130 million for AI and automation in claims processing and $4.2 billion for a problematic electronic health records system.
These proposals seem to follow a pattern: bureaucracy is bad, automation is good. Due in part to mandatory overtime and in part to technological investments, the Veterans Benefits Administration has already cut its backlog of claims by over 60%. According to the White House, these tools provide quicker, more accurate results while doing away with “the costly practice of relying on surge staffing”.
Maybe they’re correct. It’s possible that AI will be able to process tax returns and benefit claims more quickly than humans, who have been doing so for decades. However, it’s also possible that what we’re seeing is a massive experiment, carried out in real time using actual taxpayers and veterans as test subjects.
The administration’s celebration of closing Direct File, the free online tax filing platform introduced in 2024, makes the IRS situation especially ironic. The program only processed 300,000 returns, or roughly $140 per return, despite costing $41 million, according to the White House. Yes, that is pricey. However, it’s also a small-scale pilot program, the kind of thing you would anticipate becoming more affordable with experience and scale.
Rather, the administration eliminated it while simultaneously placing a $1.2 billion wager that alternative technology—enforcement tools driven by artificial intelligence—would somehow perform better. That decision contains optimism, or maybe it’s just a different set of priorities. Free filing facilitates system navigation for taxpayers. The government uses AI enforcement to apprehend those who attempt to manipulate it.
As you go through the specifics of this budget proposal, you begin to see how much depends on technology to produce outcomes that government IT initiatives have not in the past. After receiving an additional $800 million, the VA’s electronic health records system was put on hold for three years because of “persistent outages and usability issues.”
According to the IRS’s own chief information officer, the agency lost 80% of its tech leadership and 40% of its IT staff last year. These aren’t particularly encouraging beginnings for a significant AI revolution.
However, the administration appears confident that this is the correct course. The budget calls for “streamlining operations” and “eliminating bureaucratic morass.” There is a sincere belief that technology can save us from decades of accumulated dysfunction and that government has grown bloated and ineffective.
Perhaps it does. Or perhaps we’re going to learn some costly lessons about what happens when you cut first and automate second—that is, when you reduce human expertise before you have demonstrated that the machines can perform the task. The first day of the fiscal year is October 1, 2026. We’ll know by this time next year whether this risk was worthwhile or if American taxpayers are waiting for an algorithm to figure out their issue.