On May 8, 2026, I gave Claude AI the same tax-planning prompt I had been running through a $300/hour CPA for years. Persona: Alex, 38, married filing jointly, US-based, $230,000 of combined W-2 income, $20,000 of side-business 1099, two kids, no mortgage. The CPA charged for two hours and returned a 4-page memo. Claude returned a comparable plan in 90 seconds. The recommendations matched almost line-for-line. The CPA was still right about three things Claude missed. The other 80% of the work, Claude did for free.
This is not a guide to replacing your accountant. It is a guide to using Claude AI to do the same scenario-modeling, bracket-checking, and contribution-stack work that costs you hundreds of dollars in CPA time, so that when you do bring in the CPA, you arrive with a defensible plan and a list of three sharp questions instead of a stack of unopened mail. Tax planning is iteration. Claude makes the iteration cheap.
This guide uses the actual 2026 IRS numbers (post-OBBBA inflation adjustments), runs through Alex's plan end-to-end with real math, and shows the three prompts that compress a tax-planning hour into ten minutes. US-primary with a transferable framework for non-US readers.
Why Tax Planning Is the Highest-ROI AI Use Case for Money
Three reasons tax-planning is where Claude AI delivers the most dollars per minute spent.
First, tax math is deterministic. Unlike stock picking, the answer to "what is my marginal rate" has one correct number once your inputs are set. Claude is excellent at deterministic math when the rules are public, which IRS rules are. The 2026 brackets are published. The contribution limits are published. The deduction phaseouts are published. Claude is essentially running the same arithmetic a TurboTax wizard runs, but with more flexibility on the question framing.
Second, the optimization landscape is rule-based. The order in which you fill tax-advantaged accounts (employer match first, then HSA, then tax-deferred, then tax-free, then taxable) is a known principle. Claude applies that principle to your specific income stack and produces a ranked list. You verify the assumptions and execute. The hard part is knowing the rules; Claude knows them.
Third, scenario modeling is where humans give up. Asking "what if I put $5,000 more into my Solo 401k" produces a 30-minute spreadsheet exercise for most people. Claude does it in 5 seconds. Run 10 scenarios in 5 minutes and you find the local maximum on your tax savings without paying anyone $300/hour to run the same scenarios.
$39,479 Alex's projected 2026 federal tax bill before optimization. $11,500+ in legitimate savings still on the table.
Alex's tax stack for 2026: federal liability before optimization, three highest-impact moves, tax-advantaged contribution stack ranked by efficiency.
How to Run Tax Planning in Claude (3 Prompts)
The full workflow is three prompts. Total time: about 10 minutes.
Prompt 1: Bracket and Optimization Brief
Act as a fee-only tax-aware financial planner for [TAX YEAR]. Persona: [age, filing status, state, gross W-2 income(s), self-employment income, kids, current account balances, deduction style]. Calculate: (1) total federal tax liability before any optimization using the current year IRS brackets, (2) marginal tax rate including payroll on self-employment, (3) three highest-impact tax moves still available this year with dollar savings each, (4) suggested account contribution stack ranked by tax efficiency, (5) one overlooked tax-planning lever most middle-class households miss. Show your math. Cite IRS for each rate. Under 450 words.
Prompt 2: Year-End Deduction Checklist
For the same persona, produce a checklist of year-end tax-deduction and credit moves to evaluate before December 31. For each item, note the dollar limit, the income phaseout (if any), and whether it is itemized-only or available with standard deduction. Cite IRS. Under 350 words.
Prompt 3: Multi-Year Tax Trajectory
Now project this household's federal tax liability over the next 5 years assuming the income trajectory and tax-advantaged contribution stack from Prompt 1. Compare against the same 5-year path with no optimization. Show the cumulative dollar savings. Identify the year where Roth conversion windows open up. Under 400 words.
Verify three numbers from Prompt 1 against the IRS website before executing any move. The 2026 standard deduction (MFJ $32,200), the 401(k) employee deferral limit ($24,500), and the relevant tax-bracket thresholds are the three to spot-check. Anything else can be re-prompted.
Real Example: Alex's 2026 Plan
All numbers below are the actual 2026 IRS figures Claude pulled and computed for the persona above.
Pre-Optimization Federal Liability
- Combined W-2 wages: $140,000 + $90,000 = $230,000
- Side-business net SE income: $20,000 × 92.35% = $18,470
- Half-SE-tax adjustment: -$1,413
- Adjusted gross income (AGI): ~$247,057
- 2026 MFJ standard deduction: -$32,200
- Taxable income: $214,857
- Federal income tax (MFJ 2026 brackets, blended): ~$36,665
- SE tax (15.3% on $18,470): $2,826
- Total federal liability before optimization: ~$39,491
- Marginal rate on next dollar of W-2: 22%; marginal on next SE dollar: 22% + 14.13% effective SE = 36.1%
Three Highest-Impact Moves
Move 1: Open a Solo 401(k) on the side business. Defer up to $18,470 of net SE income (full employee + employer split is more complex, but $18.4K is the conservative deferral). Federal income tax savings: $18,470 × 22% = $4,063. SE-tax savings on the employee portion: minimal because employee deferrals stay subject to FICA. Net savings approximately $4,063.
Move 2: Max both spouses' employer 401(k)s. Alex contributes the $5,600 match minimum; spouse contributes $4,500 match. To hit the 2026 employee limit of $24,500 each, Alex needs an additional $18,900 and spouse $20,000. Combined extra deferral $38,900. Federal income tax savings: $38,900 × 22% = $8,558. This is the single largest lever.
Move 3: Open a Health Savings Account if eligible (HDHP family plan). 2026 family HSA contribution limit: $9,200. Triple-tax-advantaged: tax-deductible going in, tax-free growth, tax-free withdrawals for medical expenses. Federal income tax savings on the deduction: $9,200 × 22% = $2,024. FICA savings if contributed via employer payroll: $9,200 × 7.65% = $704. Combined: ~$2,728.
Total optimization savings stacked: $4,063 + $8,558 + $2,728 = $15,349 of legitimate federal tax savings still available to Alex. The starting $39,491 bill drops to approximately $24,142 after these three moves alone.
Tax-Advantaged Account Stack (Ranked by Efficiency)
- 1. Capture full employer match (free money, infinite ROI on the matched amount).
- 2. HSA up to family limit ($9,200 in 2026), if HDHP-eligible.
- 3. Tax-deferred employer plan to the $24,500 employee limit.
- 4. Roth IRA via Backdoor (income exceeds direct contribution phaseout): $7,500 each in 2026.
- 5. Solo 401(k) on side income up to ~$18,470 net deferral.
- 6. After all of the above, taxable brokerage with tax-loss harvesting (~$3,000 capital loss against ordinary income annually).
$15,349. That is the tax savings still available to Alex this year if he stacks Solo 401(k) + maxed-out employer plans + family HSA.
Year-end deduction checklist + tax-advantaged account stack ranked by efficiency. Print this and check off as you execute.
Year-End Tax Move Checklist
Evaluate each item before December 31. Most have hard cutoffs.
- Max employer plan contributions: $24,500 per spouse for 2026 (under 50). Last paycheck of the year is the deadline.
- HSA contributions: $9,200 family / $4,400 single in 2026. Deadline April 15 of the following year.
- Solo 401(k) employee deferral: same $24,500 limit. Account must be open by Dec 31; funding can extend later.
- Backdoor Roth IRA: $7,500 per spouse. Watch the pro-rata rule if you have any pre-tax IRA balances.
- Charitable bunching with a Donor-Advised Fund: bundle 2-3 years of giving into one tax year to clear the standard deduction threshold.
- Tax-loss harvesting: realize $3,000 of capital losses to offset ordinary income; excess carries forward indefinitely.
- 529 plan contributions: state-tax deduction varies; check your state's rules before December.
- Section 199A QBI deduction on side business: 20% of qualified business income reduces taxable income; phaseouts apply above ~$394K MFJ in 2026.
- Required Minimum Distributions (RMDs): if you're 73+, deadline December 31. Penalty for missing is 25% of the shortfall.
Common Mistakes That Cost You
Mistake 1: Confusing Marginal Rate with Effective Rate
Alex's marginal rate is 22%. His effective rate is closer to 16%. The marginal rate matters for the next dollar earned or saved. The effective rate matters for budgeting. Confusing the two leads to bad decisions on whether to defer income or accelerate it. Always ask the AI to give you both.
Mistake 2: Not Asking About SE Tax on Side Income
Self-employment income is hit with 15.3% SE tax on top of regular income tax. A $1,000 side gig at the 22% bracket produces ~$361 of total federal tax, not $220. Many AI tax outputs miss this if you don't explicitly mention self-employment. Always specify W-2 vs 1099 income separately.
Mistake 3: Treating Roth as Always Better
Roth contributions look good on paper because growth is tax-free. But if you're in the 22% or higher bracket today and expect to be in 12% or lower in retirement, traditional pre-tax is better. The cross-over depends on your projected retirement income. Run both scenarios in Claude before defaulting to Roth.
Mistake 4: Skipping the Backdoor Roth Pro-Rata Rule
If you have any pre-tax IRA balances (rollover from old 401k counts), the Backdoor Roth becomes partially taxable on a pro-rata basis. Many high earners walk into this trap. Ask Claude explicitly: "Do I have any pro-rata exposure on a Backdoor Roth?" before executing.
Mistake 5: Not Stress-Testing for AMT
The Alternative Minimum Tax is largely defanged by the 2017 TCJA but still hits roughly 200,000 households per year, mostly in high-tax states with large state-and-local-tax deductions. If you live in CA, NY, or NJ and earn over $400K MFJ, ask Claude to run an AMT check. The 2026 AMT exemption for MFJ is $137,000.
Frequently Asked Questions
Does this work for non-US tax systems?
The framework transfers directly. Replace IRS with HMRC (UK), CRA (Canada), ATO (Australia), or your local tax authority. Replace 401(k) with ISA + SIPP (UK), TFSA + RRSP (Canada), Super (Australia). The principles (capture employer match, fund tax-advantaged before taxable, use bracket arbitrage for Roth-equivalent conversions) apply universally. The numbers change; the framework does not.
Should I trust Claude's tax math for filing?
No. Claude is a planning tool, not a filing tool. Use it to model scenarios, identify levers, and arrive at meetings with your CPA prepared. Use TurboTax, FreeTaxUSA, or a CPA for the actual filing. Treat Claude's numbers as a sanity-check directional answer, not a final return.
How current is Claude's tax knowledge?
Claude pulls live from IRS.gov and reputable tax-news sources via web search. The 2026 brackets and contribution limits are accurate as of May 2026. For mid-year legislation changes (such as new credits enacted in summer 2026), always re-prompt with "verify against the latest IRS publication."
Can I share my actual W-2 with Claude?
You can paste the relevant numbers (gross wages, federal withholding, state withholding, retirement contributions) but avoid pasting the full document with your SSN, employer EIN, or address visible. Strip the personally identifiable fields before pasting.
What to Watch Next
- v Does the 2027 tax legislation extend or sunset the TCJA individual provisions expiring in 2025?
- v Does the IRS issue 2027 inflation-adjusted brackets in October 2026 (typical timing)?
- v Does Claude's tax-planning accuracy on edge cases (multi-state, AMT, NIIT) keep pace with rule changes?
- v Does your own win rate on year-end optimization improve when you run the AI plan and then compare to your CPA's plan?
- v Does any AI vendor launch a tax-specific certified model with E&O coverage by 2027?
Key Takeaways
- Tax planning is the highest-ROI AI use case for money because the rules are public and the math is deterministic.
- Three prompts cover most needs: bracket and optimization, year-end checklist, multi-year trajectory.
- Alex's 2026 plan: $39,491 starting federal liability, $15,349 in stackable savings via Solo 401(k) + maxed employer plans + family HSA.
- Tax-advantaged stack: employer match first, HSA second, tax-deferred third, Backdoor Roth fourth, Solo 401(k) fifth, taxable last.
- Marginal vs effective rate: always ask the AI for both. Decisions depend on which one you are optimizing.
- Verify three numbers (standard deduction, 401(k) limit, your bracket threshold) against IRS.gov before executing moves.
- Claude is a planning tool, not a filing tool. Use TurboTax or a CPA for the actual return.
References
IRS 2026 inflation-adjusted tax brackets and contribution limits: irs.gov inflation adjustments 2026
Tax Foundation 2026 federal income tax rates: taxfoundation.org
IRS Publication 590-B (IRA distributions and contribution limits): irs.gov pub 590
HSA contribution limits 2026: irs.gov HSA
Claude AI: claude.ai