Learn how real-time AI documentation could have saved $2.5M in at-risk R&D credits.
George v. Commissioner: Where It All Went Wrong
In today’s hyper-scrutinized tax environment, simply doing legitimate research and development is not enough to be able to claim R&D tax credits. A company must be able to prove it did legitimate research and development. Often, companies build the case for R&D tax credits retroactively, using a combination of existing data and employee interviews as proof for the work. But the case of George v. Commissioner shows what can go very wrong with this approach.
George’s of Missouri, Inc. (GOMI) — one of the largest fully integrated poultry processors in the U.S. — claimed $4.47 million in R&D tax credits across 2012–2014 for seven research trials on feed additives,vaccines, probiotics, and genetic lines. alliantgroup conducted the retroactive R&D study.
The Tax Court’s verdict was a blow. They disallowed three of seven trials, and the credit rate was slashed from 14% to 6%. Not because the research wasn’t real — but because the documentation couldn’t prove it.
GOMI’s research was legitimate — trials on Ross 708 genetic lines, probiotic additives like Sporulin and Calsporin, and vaccine protocols were real scientific work conducted by qualified personnel. But the R&D study was built retroactively by alliant group, relying on employee interviews conducted months or years after the research occurred. When the IRS compared those interviews narratives against GOMI’s own operational records — feed logs, flock data, procurement records — the contradictions were fatal.
The court found that feed records showed constant, unchanged dosages during periods where consultants claimed active experimentation was occurring. Entire flocks that allegedly received test additives had no documentation whatsoever. And credits were claimed for research years after the underlying uncertainty had already been resolved.
On top of the individual trial failures, GOMI couldn’t substantiate its base-year qualified research expenses, triggering a start-up limitation under IRC §41(c)(5)(B) that cut the credit rate from 14% to just6% — compounding the losses across every surviving claim.
THE DOCUMENTATION FAILURES THAT COST MILLIONS
Of seven research trials, three were completely disallowed and two were only partially allowed. Every failure traced to the same root cause: retroactive documentation that contradicted operational records.
● DISALLOWED / PARTIAL
Salinomycin & Phytase — Interview claims of experimentation contradicted by feed records showing constant, unchanged dosages (~$1.2M in QREs at risk)
Floramax — No records existed showing which flocks received the additive; the court refused to estimate (~$680K in QREs unsubstantiated)
HatchPak & Tylan — Credit claimed for 2013 after uncertainty was already resolved in 2012 ($2.87M credit year partially disallowed)
LT Vaccines — Priming trials allowed, but method-of-administration lacked timing records (partial credit reduction)
Combined with the base-year failure cutting the rate from 14% to 6%, the total credit impact was devastating across all $4.47M claimed
● ALLOWED — THE STANDARD THAT WON
Ross 708 Genetic Line — Called “the cleanest example of the scientific method” — control groups, data analysis, clear conclusions
Sporulin & Calsporin — Clear hypothesis, test, analyze, iterate pattern with documented next steps upon failure
These trials survived because their documentation matched what Exactera captures automatically for every project
The Base-Year Penalty
Beyond individual trial failures, GOMI couldn’t substantiate base-year QREs (2009–2011). This triggered the start-up limitation under §41(c)(5)(B), slashing the credit rate from 14% to 6% — losing more than half the credit value on every qualifying trial. Exactera’s platform maintains rolling base-year data automatically, making this trap structurally impossible
WHY THIS KEEPS HAPPENING: THE RETROACTIVE STUDY MODEL
The failures in George aren’t unique — they’re structural. alliantgroup and traditional R&D firms all use the same retroactive model:
1. Research happens. No credit-specific documentation is captured.
2. 6–18 months later, consultants arrive and interview employees from memory.
3. Narratives are manually reconstructed — often contradicted by operational records the consultants never reviewed.
4. A static report is delivered. When the IRS finds discrepancies, there’s no way to fix them.
This is the model that produced three disallowed trials and a credit rate cut in half. The new Form 6765 makes this approach even riskier by requiring project-level detail and contemporaneous evidence that retroactive studies were never designed to produce.
TRADITIONAL METHODS v. EXACTERA
DIY / IN-HOUSE
No formal documentation process
Engineers focus on research, not tax compliance — critical evidence is never captured
No QRE methodology
No understanding of what qualifies; expenses lumped together or guessed at filing time
Static point-in-time report
Can’t be updated or verified against live data
No four-part test awareness
Activities claimed without mapping to §174, business component, or experimentation tests
No base-year tracking
Historical QRE data doesn’t exist; start-up limitation almost guaranteed
Maximum audit exposure
No defensible position — IRS can disallow the entire credit with minimal disruption
RETROACTIVE STUDY
Interview-based narratives
Written from fading memories months after research ends
Manual QRE allocation
Spreadsheet estimates with no audit trail
No four-part test awareness
Qualification assessed retrospectively
Qualification assessed retrospectively
Four-part test applied by consultants, not validated during research
Base-year records not maintained
Left to the client — often lost or incomplete
Partial, fragile defense
Retroactive narratives collapse when compared to operational records
EXACTERA’S REAL-TIME AI
Live documentation capture
Project data ingested as research happens — no recall required
Automated QRE engine
Wages, supplies, and contracts tied to projects from payroll and GL data
Living, audit-ready output
Continuously updated documentation that maps to Form 6765 sections
AI-powered qualification
Four-part test evaluated continuously — issues flagged before filing
Rolling base-year maintenance
Platform tracks base-period QREs automatically from year one
Minimized audit risk
Contemporaneous, system-generated evidence designed to withstand IRS scrutiny
NEW FORM 6765 COMPLIANCE: THREE APPROACHES COMPARED
REQUIREMENT
Contemporaneous documentation
- DIY / In-House: Non-existent
- Retroactive Study: Structurally impossible
- Exactera AI: Core architecture
Project-level narratives
- DIY / In-House: Not attempted
- Retroactive Study: Interview recall; contradiction risk
- Exactera AI: Auto-generated from live data
QRE allocation by project
- DIY / In-House: Guesswork or omitted entirely
- Retroactive Study: Manual spreadsheets
- Exactera AI: Automated with audit trail
Process of experimentation
- DIY / In-House: No awareness of requirement
- Retroactive Study: Failed in 3 of 7 George trials
- Exactera AI: Structured hypothesis to result capture
Officer attestation support
- DIY / In-House: Officer signs blindly
- Retroactive Study: Sign based on consultant summary
- Exactera AI: Full dashboard visibility
Base-year substantiation
- DIY / In-House: No historical data kept
- Retroactive Study: Often missing; cost GOMI 8% of credit rate
- Exactera AI: Maintained automatically
THE BOTTOM LINE
The George case isn’t ancient history — it was filed February 3, 2026. It proves that retroactive R&D studies produce documentation that crumbles under IRS scrutiny. Three trials disallowed. Credit rate cut in half. Millions in credits lost.
Every one of those failures traces back to the same root cause: documentation was reconstructed from memory instead of captured in real time.
Exactera’s AI platform doesn’t reconstruct the past. It documents the present. And that’s the difference between bcredits that survive an audit and credits that don’t.
Based on George v. Commissioner of Internal Revenue — T.C. Memo. 2026-10
This material is for informational purposes only and does not constitute legal or tax advice
