The IRS Is Running AI on Your Clients' Returns — Here's the 7-Step Workflow That Cuts Notice Response in Half
Published: April 18, 2026 | By: The Crossing Report
Summary
Tax season ended April 15. Now the IRS's AI is running. The agency is using artificial intelligence to cross-reference returns against bank records, 1099-K data, and third-party income reports — flagging discrepancies faster than any human examiner could. At the same time, the IRS has cut approximately 27% of its workforce. The result: more AI-generated notices, fewer humans to resolve them, longer delays for clients. For accounting firms, this is both a compliance alert and a revenue signal. A documented 7-step AI workflow cuts notice resolution time from 35 days to 14 — and saves a typical firm 22 hours per month.
The IRS AI System That Just Finished Its First Full Filing Season
The IRS has been adding AI to its return-processing infrastructure for years. What's different in 2026 is that the system is now processing the full filing volume — and doing it with a dramatically smaller human workforce.
Get the full picture. Go premium.
Weekly intelligence briefings, deeper analysis, and direct access to the full archive.
Here's what the IRS's AI does during return processing:
Income cross-referencing. The system compares what a taxpayer reported on their return against every W-2, 1099-INT, 1099-DIV, 1099-NEC, 1099-K, and third-party report the IRS received for that taxpayer. Discrepancies — even small ones — are flagged for review. What took IRS examiners days now happens in seconds.
Payment app income matching. The $600 1099-K threshold change (in effect for the 2025 tax year, reported on 2026 returns) means millions of taxpayers who received money through Venmo, PayPal, Cash App, and Zelle received 1099-Ks for the first time this filing season. Many of them — and some of their accountants — may have handled these inconsistently. The IRS's matching AI will find those inconsistencies.
Audit-target selection. The agency's AI flags statistically unusual returns — entries that deviate from historical patterns for a given income level, industry, or deduction type — and routes them for examiner review. The formal governance policy (IRS Internal Revenue Manual 10.24.1, February 2026) documents this use. A separate GAO audit report (GAO-26-107522, 2026) confirmed AI's role in audit selection.
Here's the context that makes this urgent for accounting firms: the IRS has simultaneously cut roughly 27% of its workforce — from approximately 102,000 employees in early 2025 to 74,000–75,000 today. The AI is flagging more returns. There are fewer humans to process the fallout. Processing delays are already appearing, and the National Taxpayer Advocate's 2026 Annual Report explicitly flagged "significant service challenges ahead."
More flags. Longer queues. More confused clients calling their accountant.
What's Coming to Your Clients' Mailboxes
The specific notices that AI income-matching generates:
CP2000 — The most common AI-generated notice. Proposes changes to a return when the IRS's records of income differ from what the taxpayer reported. Almost all 1099-K mismatches, unreported freelance income, and investment account discrepancies generate CP2000s.
CP14 — Balance due. Sent when the IRS calculates that additional tax is owed after matching. Often follows a CP2000 if the taxpayer doesn't respond or the resolution results in additional liability.
CP504 — Intent to levy. Sent when a balance due remains unresolved. Creates urgency and requires immediate action.
Letter 12C — Additional information requested. The IRS needs documentation to complete processing — most commonly for deductions, credits, or income items that were flagged but need supporting evidence before a determination is made.
CP5071C — Identity verification. Triggered when the AI detects patterns that suggest possible identity fraud or an anomalous filing pattern.
The volume of these notices has historically tracked IRS enforcement capacity. With AI running the matching and flagging — and fewer humans to slow the output — the volume this year will be higher, not lower.
For accounting firms: the clients who've been managing their own IRS correspondence — the ones who call you only when things escalate — are the ones most likely to be overwhelmed by this year's notice volume. They are the inbound demand signal.
The 7-Step AI Workflow That Cuts Resolution Time from 35 to 14 Days
CPA Pilot published a detailed workflow analysis showing how AI handles IRS notice response across seven steps. The documented results: firms handling 30 notices monthly save approximately 22.5 hours per month (45 minutes per notice), and average resolution time drops from 35 days to 14.
Here's how the workflow operates:
Step 1: Triage and Auto-Classification
When a notice arrives, AI classifies it automatically: notice type (CP2000, CP14, CP504, Letter 12C, CP5071C), tax year, issue type, and response deadline. The triage step determines which response template and exhibit protocol to use, and assigns a priority level based on the notice type and deadline.
What this replaces: A CPA or admin manually reading the notice, determining what it requires, figuring out the response deadline, and routing it to the right person. At 5–10 minutes per notice for 30 notices, that's 2.5–5 hours of routing work per month.
Step 2: Research
For most notice types, the IRS is citing a specific code section, regulation, or discrepancy type. AI research tools pull the relevant tax code provisions, Treasury regulations, IRS guidance, and any applicable precedent for the issue type flagged. The research output gives the responding CPA the relevant framework before they've read the first source document.
What this replaces: 20–40 minutes of manual research per notice — longer for unfamiliar notice types or novel issues.
Step 3: Draft Response Letter
Using the notice classification, the tax year, the client's filing data, and the research output, AI generates a first-draft response letter in the format and tone the IRS expects. The draft includes the relevant code citations, the specific factual basis for the response, and the requested action (accept, modify, no change).
What this replaces: The most time-intensive writing step in the notice response workflow. For a standard CP2000 response, writing from scratch typically takes 60–90 minutes. AI gets you to a 90%-complete draft in minutes.
Step 4: Exhibit Building
AI organizes the supporting documentation — W-2s, 1099s, bank statements, receipts, prior-year returns — into the exhibit format the IRS expects for the specific notice type. Documents are labeled, sequenced, and cross-referenced to the response letter.
What this replaces: Manual document gathering and organization, which is often the step that creates the most delay and error risk in the notice response workflow.
Step 5: Human Fact Confirmation — The Non-Negotiable Gate
This is the only step that doesn't accelerate. The CPA reviews the AI draft, verifies every factual claim against the actual client records, corrects anything the AI got wrong or missed, and applies professional judgment to anything that requires interpretation.
This step cannot be abbreviated. The AI drafts; the CPA verifies. The liability is yours, not the software's.
The efficiency gain here is upstream: because the AI did the research, drafted the letter, and organized the exhibits, the CPA is reviewing a near-complete package rather than building from scratch. The review takes 15–30 minutes instead of 90–120.
Step 6: Client Communication in Plain English
After the response is finalized, AI drafts the client update: what the IRS said, what your firm responded, what happens next, and what — if anything — the client needs to do. Written in plain English, not tax language.
What this replaces: Time spent translating the IRS notice and your firm's response into something a non-accountant can understand. A task that's often skipped when firms are busy — to the detriment of client relationships.
Step 7: Case Tracking Until Resolution
AI tracks the open case through the IRS's response cycle: when the response was sent, what the IRS acknowledged, when a determination is expected, and when to follow up. Cases that go past their expected response window are flagged automatically.
What this replaces: Manual case tracking on a spreadsheet or in the firm's project management system — a step that frequently falls through the cracks during busy periods.
What This Looks Like for a Typical Accounting Firm
At 30 notices per month (a reasonable baseline for a 10-person firm with active tax practice), the math is straightforward:
- Time saved per notice: ~45 minutes
- Monthly time recovery: ~22.5 hours
- Annual time recovery: ~270 hours
- Value at $200/hour: ~$54,000 in recovered staff capacity annually
The 14-day average resolution time (vs. 35-day baseline) also matters for client relationships. An IRS notice that lingers for five weeks breeds anxiety and erodes trust. A notice resolved in two weeks demonstrates that your firm has the workflow to handle these efficiently.
The Representation Opportunity Your Firm Should Consider
Most small accounting firms handle IRS notices reactively — when a client calls. The structural shift underway creates an opportunity to get ahead of it.
With the IRS's AI generating more flags and the agency's human workforce reduced, clients who've been managing their own IRS correspondence — the small business owners, freelancers, and self-employed professionals who receive a notice and spend weeks trying to figure out what it means — are the most likely to need professional help this year.
Three services that become more valuable in a depleted-IRS environment:
Penalty abatement. The IRS's AI can generate notices; the humans who approve penalty relief are fewer. Knowing the process — first-time abatement, reasonable cause, administrative waiver — is a differentiator when clients are facing penalties.
Audit reconsideration. For clients who received an assessment and didn't respond effectively the first time, audit reconsideration is the formal mechanism to reopen the case. Understanding the workflow puts you in front of clients who tried to handle this themselves and failed.
Power of Attorney and transcript access. Before you can respond to any notice effectively, you need IRS transcript data. Clients who come to you mid-notice process often don't have the documentation the IRS expects. Being the firm that can pull transcripts, identify what the IRS has, and respond systematically is a competency worth building before the notice volume peaks — which is typically May through July for the prior-year filing season.
Your Next Client Conversation
The proactive version of this doesn't require a new service, a new system, or a new team member. It's two sentences:
"The IRS lost a quarter of its workforce this year, and it's using AI to flag returns it doesn't have enough examiners to resolve quickly. If you get a notice this year — any notice — call me before you respond."
That's it. The clients who hear that from you are less likely to spend three weeks responding to a CP2000 on their own, making the situation worse. The clients who don't hear it from anyone are the ones who call you in month five of an escalated situation.
What to Do This Week
For accounting firms handling any volume of IRS notices:
Look up CPA Pilot's published workflow breakdown (cpapilot.com) and review the seven-step structure against your current process. Identify which steps you're doing manually that AI could accelerate.
Build a standard exhibit protocol for your three most common notice types (CP2000, CP14, Letter 12C). Document the format the IRS expects and create a checklist for what client documents need to be gathered at each step. This one-time investment pays off every time a notice of that type arrives.
Draft the two-sentence proactive client message. Send it to your 10 most likely candidates — small business owners, freelancers, anyone who received their first 1099-K this year. You're not selling them anything. You're telling them to call you before something escalates.
The notice volume that follows a filing season with AI-generated flagging at scale will materialize over the next 90 days. The firms that have a workflow ready will handle it efficiently. The ones that don't will handle it slowly, expensively, and reactively.
Related Reading
- How Accounting Firms Are Growing Revenue Without Growing Headcount
- The 3-Step AI Framework for Accounting Firm Client Advisory Services
- AI Capacity Multiplier: How Small Accounting Firms Are Growing Without Hiring
- Accounting Firms Replacing New Hires With AI: The Math Behind the Decision
The Crossing Report covers the AI transition for professional services firm owners — including accounting, law, consulting, staffing, and marketing agencies. For weekly intelligence, subscribe here.
Frequently Asked Questions
Is the IRS really using AI to audit tax returns in 2026?
Yes. As of the 2026 filing season, the IRS uses AI to cross-reference returns against third-party data — bank interest, freelance income, payment app transactions (1099-K) — in seconds. The agency's internal governance policy (IRM 10.24.1, February 2026) formally documents AI's role in tax return processing and audit-target selection, and a GAO report (GAO-26-107522) confirmed its use. At the same time, the IRS has reduced its workforce by approximately 27% — from roughly 102,000 to 74,000–75,000 employees — meaning more AI-flagged returns are being generated by fewer humans available to resolve them.
What IRS notices are most likely to increase because of AI flagging?
The notices most commonly triggered by AI cross-referencing are CP2000 (proposed changes when reported income doesn't match IRS records), CP14 (balance due), CP504 (intent to levy), Letter 12C (additional information needed), and CP5071C (identity verification). CP2000 notices are the primary output of the income-matching AI — they're generated when the IRS's system identifies a discrepancy between what a taxpayer reported and what a third-party payer (employer, bank, payment platform) reported.
How does an AI-powered IRS notice response workflow work?
The CPA Pilot workflow breaks notice response into seven steps: (1) triage and auto-classification of the notice type; (2) AI research pulling relevant tax code, regulations, and prior rulings; (3) AI-drafted response letter using client facts; (4) exhibit building — organizing supporting documents into the format IRS expects; (5) human fact confirmation — the CPA reviews and verifies the AI output; (6) client communication drafted in plain English; (7) case tracking until resolution. AI handles steps 1–4 and 6–7 at speed; the CPA concentrates on step 5. This structure is what compresses resolution time from an average of 35 days to 14.
How much time does AI save on IRS notice response?
Based on CPA Pilot's workflow analysis, a firm handling 30 notices monthly saves approximately 22.5 hours per month — roughly 45 minutes per notice. At 30 notices per month, that's 270 hours annually. For a 10-person accounting firm billing at $200 per hour, that's approximately $54,000 in recovered staff capacity annually. The time savings compound as the firm's notice templates and exhibit libraries improve through reuse.
Should accounting firms expand their IRS representation services after tax season?
The data supports it. With the IRS deploying AI to flag discrepancies while simultaneously reducing its workforce, the combination means more notices will be generated, and fewer IRS staff will be available to resolve them quickly. Clients who've handled IRS correspondence on their own for years — especially 1099-K mismatches, side-gig income, and investment income reporting — are the most likely to need professional help this year. Enrolled agents and CPAs who build or expand an IRS problem-resolution capability before June 2026 are positioning ahead of demand.
Get the weekly briefing
AI adoption intelligence for accounting, law, and consulting firms. Free to start.
Free weekly digest. No spam. Unsubscribe anytime.
Related Reading
This is the kind of intelligence premium subscribers get every week.
Deep analysis, cross-sector patterns, and the frameworks that help professional services firms make the crossing.