EXPOSED
The bulk of this work — matching reported income against third-party data, flagging arithmetic and eligibility errors, sending deficiency notices, computing interest and penalties, and processing correspondence audits — is structured data comparison that automated matching systems already do at scale and that AI extends further. What survives is the enforcement tier: field audits of complex business books, interviewing taxpayers and representatives, exercising statutory discretion on offers in compromise, liens, levies, and penalty abatement, and defending findings in appeals. Score reflects the modal examiner doing desk review; senior revenue agents auditing partnerships and corporations sit materially higher.
Headcount grew steadily across the period.
Median pay $54,890 → $62,370 -9.1% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
-1.8% 57,600 → 56,500 on the projections basis
Exposed, and shrinking
Both signals point the same way: the tasks are largely automatable and the BLS projects -1.8% by 2034. This is the case where the score and the forecast agree, and it is the one worth taking seriously.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~4,300 openings a year on average, including replacing people who leave.
VerifierCollectorTax AgentTax AnalystReturn AgentTax AdjusterTax ExaminerTax PreparerTax ReviewerBlock AdvisorCustoms AgentCustoms GuardRevenue AgentSpecial AgentState AuditorTax AssociateTax CollectorTax ProcessorCity CollectorPort CollectorTax SpecialistTax TechnicianTreasury AgentRevenue Officer
Holding it up: judgment & accountability . Weakest point: trust premium .
Core tasks are already automatable The modal examiner's day is comparing a 1040 against W-2s and 1099s, recomputing EITC eligibility, issuing CP2000 and CP-series notices, and closing correspondence audits from a queue — work the IRS's own Automated Underreporter and math-error programs already do without a human, which is why this sits at 6 rather than the 10-12 a field revenue agent reconstructing a partnership's books would earn.
Some physical or field component Above pure desk work because field collection and field audit mean going to the taxpayer's place of business to inspect records, inventory, and assets, serving levies and seizing property, and occasional site visits with a POA present — but that's a minority of positions and the rest is a terminal, IDRS, and a phone, holding it to 5.
Certification preferred, not legally required There is no licence to examine returns for the government — Revenue Agent positions require 30 semester hours of accounting, not a CPA, and the assessment is the Service's, not the individual's — but delegated authority under IRC 6320/6330 and 7122 means a specific examiner's signature attaches to liens, levies, and compromise acceptances and is reviewable in Appeals and Tax Court, which is what earns 8 instead of 3.
Meaningful discretion Real discretion exists in reasonable-cause penalty abatement, evaluating collectibility on an offer in compromise, and deciding whether to expand an audit's scope, but it is exercised inside the IRM with mandatory managerial approval thresholds, DATL review, and Appeals as a backstop, which keeps this at 10 rather than the 15+ of someone whose call is final and unreviewed.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (6/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (8/20) is whether the law requires a licensed human to sign. Trust premium (5/20) is whether buyers specifically pay for a person. Judgment and accountability (10/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 23 of this occupation's 34 points (68%).
Embodiment (5/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
Personal Financial Advisors EXPOSED
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 50/100, still EXPOSED.
Automated matching absorbs the desk-review tier (AUR, correspondence audits, math-error notices) so the remaining headcount is concentrated in field examination of pass-through entities, transfer pricing, and crypto/basis reconstruction — the IRS's stated LB&I and partnership audit build-out under the IRA funding already shifts hiring toward revenue agents rather than tax examiners. This is task-mix shift, not new capability resistance.
If statutory discretion over offers in compromise, trust fund recovery penalty assessments under IRC 6672, and penalty abatement for reasonable cause is formally reserved to a named human examiner — e.g. IRM revisions or a Taxpayer First Act-style requirement that adverse determinations be signed by an identified employee with reviewable rationale — the surviving role owns consequential ambiguous calls.
A codified prohibition on fully automated adverse determinations, similar to Section 1001 of the Taxpayer First Act's independent appeals right or state-level bills barring automated final assessment, requiring a named agent's signature on statutory notices of deficiency, jeopardy assessments, and levy approvals, with supervisory personal accountability under IRC 6751(b)-style written-approval rules already litigated in Tax Court.
Post-Loper Bright and Boechler litigation pressure making agency determinations more contestable raises the evidentiary burden on the examiner who signed the case file, since AI-generated workpapers without a human author's testimony are weak in Tax Court and Appeals.
The limit. Trust premium has no route: taxpayers are involuntary counterparties who would generally prefer an automated process to a human examiner, and the buyer (Congress/Treasury) is buying revenue yield per dollar, not human contact. Embodiment is capped near current levels — field audits involve travel and document inspection but nothing physically unpredictable.
| Sacramento-Roseville-Folsom, CA | 3,780 | $67,610 +8% |
| Ogden, UT | 3,240 | $49,440 -21% |
| New York-Newark-Jersey City, NY-NJ | 2,910 | $86,340 +38% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 2,120 | $60,550 -3% |
| Kansas City, MO-KS | 1,980 | $50,250 -19% |
| Atlanta-Sandy Springs-Roswell, GA | 1,700 | $56,350 -10% |
| Los Angeles-Long Beach-Anaheim, CA | 1,520 | $97,680 +57% |
| Fresno, CA | 1,490 | $56,890 -9% |
| San Francisco-Oakland-Fremont, CA | 420 | $110,780 +78% |
| Vallejo, CA | 30 | $110,780 +78% |
| Worcester, MA | 50 | $109,420 +75% |
The U.S. Government Accountability Office reported on the IRS's use of artificial intelligence in its operations, including audit selection and taxpayer service functions.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.