What is a self-improving AI tax agent? A self-improving AI tax agent is an automated system that drafts tax returns, learns from every correction a human accountant makes, and rewrites its own logic to get more accurate over time. On May 27, 2026, OpenAI and Thrive Holdings revealed exactly that: a Codex-powered Tax AI that climbed from roughly 25% accuracy at launch to as high as 97% within weeks. For commercial real estate investors who live and die by after-tax returns, AI tax agents are about to reshape how cost segregation, depreciation, and partnership reporting actually get done. For the bigger picture, see our pillar guide on AI tools for commercial real estate.
Key Takeaways
- OpenAI and Thrive Holdings launched a self-improving AI tax agent on May 27, 2026 that improved drafted return accuracy from about 25% to up to 97% in roughly six weeks.
- The pilot ran through Crete Professional Alliance, a network of more than 30 accounting firms, processing 7,000 returns focused on complex 1040 and 1041 filings with messy documents and K-1 data.
- The system saved practitioners about one third of their time and raised throughput roughly 50%, with one accountant cutting tax prep from 180 hours to 15 hours.
- For CRE investors, the impact lands on cost segregation, depreciation strategy, and K-1 preparation across multi entity partnership structures, all of which drive after-tax cash flow and IRR.
- AI tax agents change after-tax returns, not NOI or cap rate, because depreciation and tax treatment sit below the line that defines a property's operating value.
Why AI Tax Agents Matter for Commercial Real Estate
Commercial real estate is one of the most tax-driven asset classes in the economy. Headline returns often come less from rent growth than from depreciation, cost segregation, 1031 exchanges, and how income flows through partnership structures to limited partners. That is exactly the work AI tax agents are now learning to automate.
OpenAI took an equity stake in Thrive Holdings in December 2025, and the two teams spent six months building Tax AI on Codex. According to OpenAI, the pilot deliberately avoided simple W-2 returns that practically file themselves and targeted medium to high complexity 1040 and 1041 filings, the trust and estate returns and individual returns stuffed with K-1 schedules that consume the most billable hours and produce the most errors, the same returns CRE syndicators, fund managers, and family offices wrestle with every spring.
The result was a system that records the full trace of each return: source file, extracted field, citation, tax-engine mapping, the accountant's correction, and the final filed value. Repeated corrections become evaluation targets, and Codex writes and tests code to fix them. Every human fix makes the next return better.
The Numbers Behind the 97% Accuracy Claim
The accuracy story is the part CRE professionals should study closely, because it shows how fast a narrow AI tax agent can improve when it is fed real corrections:
- Cold start: At launch, only about 25% of returns hit 75% correct field completion.
- Six weeks later: That figure climbed to 86%, with gains stacking up at the 90% and 100% accuracy thresholds.
- Time saved: Practitioners reclaimed about one third of their tax prep time and lifted throughput by roughly 50%.
- One real example: A senior accountant who spent 180 hours on tax prep last year spent only 15 hours this year, then reinvested that time into clients and new business.
This mirrors the trajectory CRE accounting tools are already on. AI bookkeeping platforms cut CRE transaction categorization time by 70% to 85% and auto-match 90% to 95% of transactions. To see where dedicated platforms stand, compare options in our guide to AI tools for CRE accountants.
Where AI Tax Agents Hit a CRE Portfolio
The four classic tax levers in commercial real estate are cost segregation, depreciation strategy, 1031 exchanges, and property tax appeals. AI tax agents touch all four:
- Cost segregation: Traditional studies require engineering analysis and run $15,000 to $30,000 per property. AI pre-screening tools from providers like CSSI and Capstan Tax Strategies can estimate potential savings before you commission a full study, helping you decide which assets in a portfolio justify the cost.
- K-1 and partnership reporting: Syndications and funds generate stacks of K-1s every year. A self-improving agent that learns your entity structure can draft these faster and flag inconsistencies a tired associate might miss.
- Depreciation modeling: AI can model the after-tax impact of a sale, a cost segregation study, or a 1031 exchange before you commit capital.
- Property tax appeals: AI can scan assessments against comparables to surface over-assessed properties worth appealing.
For a deeper look at the planning side, see our analysis of AI for CRE tax planning and 1031 exchanges. CRE investors looking for hands-on AI implementation support can reach out to Avi Hacker, J.D. at The AI Consulting Network.
An Important Distinction: After-Tax Returns vs. NOI
It is worth being precise about what AI tax agents change and what they do not. Net operating income (NOI) is gross revenue minus operating expenses, and it excludes debt service, depreciation, and income taxes. Cap rate is NOI divided by purchase price. Because depreciation and tax treatment sit below NOI, smarter tax work does not change a property's NOI or its cap rate.
What it does change is the investor's after-tax cash flow and after-tax internal rate of return (IRR), the discount rate that sets the net present value of all cash flows to zero across the hold period. Accelerating depreciation through a well-timed cost segregation study can defer taxes, boost early-year after-tax cash flow, and lift the after-tax IRR limited partners actually receive. AI tax agents make that optimization cheaper and faster to run, which is where the real CRE value sits.
The Bigger Pattern: Agentic Accounting Is Arriving
The OpenAI and Thrive announcement is not an isolated event. Agentic accounting has been one of the hottest categories in proptech and fintech. Basis AI raised a $100 million Series B in February 2026 at a valuation above $1 billion, and roughly 30% of the top 25 U.S. accounting firms already use its platform. Read our breakdown of agentic accounting platforms like Basis AI for the competitive picture.
Adoption data backs the trend. According to the Thomson Reuters Institute 2026 report, 34% of tax firms already use generative AI and another 47% are planning or considering it. EY notes that AI is reshaping real estate tax functions specifically, from data extraction to compliance. Meanwhile, the IRS is deploying AI of its own for audit selection, with a stated focus on complex partnerships and high-wealth filers, exactly the structures CRE investors use. That cuts both ways: AI can sharpen your filings, but it can also sharpen the scrutiny they receive.
Roughly 92% of corporate occupiers have initiated AI programs, yet only about 5% report achieving most of their AI goals. The lesson from the Thrive pilot is that narrow agents that learn from real corrections are the ones that cross from pilot to production. If you are ready to transform your underwriting and back-office processes with AI, The AI Consulting Network specializes in exactly this.
What CRE Investors Should Do Now
You do not need to wait for a self-improving agent to reach your accounting software. Start by mapping which parts of your tax workflow are repetitive and rules-based: K-1 drafting, depreciation schedules, cost segregation screening, and document extraction. Pair tools like ChatGPT, Claude, or Gemini with a CPA in the loop, and treat each correction as training data for next year. The firms winning in 2026 are not replacing accountants. They are giving each one the leverage of a team. For personalized guidance on implementing these strategies, connect with The AI Consulting Network.
Frequently Asked Questions
Q: What is a self-improving AI tax agent?
A: It is an AI system that drafts tax returns and then updates its own logic based on the corrections human accountants make. OpenAI and Thrive Holdings demonstrated one on May 27, 2026 that improved from about 25% to up to 97% accuracy in roughly six weeks by learning from real production data.
Q: How does AI tax automation help commercial real estate investors specifically?
A: CRE returns depend heavily on cost segregation, depreciation, 1031 exchanges, and partnership reporting through K-1s. AI tax agents speed up and reduce the cost of all of these, which lifts after-tax cash flow and after-tax IRR even though it does not change a property's NOI or cap rate.
Q: Will AI replace CRE accountants and tax advisors?
A: Not in the near term. In the Thrive pilot, accountants used reclaimed time to serve more clients and add services rather than disappear. Complex CRE structures still require professional judgment, and a human should review every filed return, especially given that the IRS is using AI to target complex partnerships.
Q: Is it safe to use AI tools for sensitive tax data?
A: Data security is the most cited barrier to AI adoption among finance executives. Use enterprise-grade tools with clear data handling terms and never file an AI-drafted return without human verification of the numbers and citations.