Answer in brief: All four base-case payment calculations were within 0.2% of the local formula. The model’s initial ranking changed when fees and the holding period were made explicit, so no single offer was labeled best.
Can AI organize loan terms without converting an incomplete comparison into a recommendation? A useful answer has to be narrower than a product claim. This article tests a bounded workflow, publishes the scoring surface, and keeps consequential approval with a person. It does not turn a controlled result into personalized financial advice.
What we tested or analyzed
We gave OpenAI GPT-5.6 in a Codex editorial session four fictional offers, recomputed amortized payments locally, and varied the holding period and fee treatment. We prohibited borrower-specific advice.
The original asset is a four synthetic offers and independently recomputed payment/cost table. The complete machine-readable table is available as CSV. The evidence visual below summarizes the primary criterion; its values are also written in text and shown in the table, so the chart is not the only way to obtain the result.
The measured result
All four base-case payment calculations were within 0.2% of the local formula. The model’s initial ranking changed when fees and the holding period were made explicit, so no single offer was labeled best.
The row-level outcome distribution was calculation pass: 4. Those labels are deliberately more descriptive than one blended score. A partial, review, stale, exception, or unsupported row can carry a different operational risk from a plainly wrong row, so the CSV preserves the reason beside the disposition.
| Item | Outcome | Evidence or note |
|---|---|---|
| Offer A | calculation pass | fixed rate, higher fee |
| Offer B | calculation pass | fixed rate, lower fee |
| Offer C | calculation pass | variable rate |
| Offer D | calculation pass | balloon feature |
Reading the evidence row by row
- Offer A was recorded as calculation pass. The evidence note is “fixed rate, higher fee”; the disposition remains visible so it cannot be averaged away.
- Offer B was recorded as calculation pass. The evidence note is “fixed rate, lower fee”; the disposition remains visible so it cannot be averaged away.
- Offer C was recorded as calculation pass. The evidence note is “variable rate”; the disposition remains visible so it cannot be averaged away.
- Offer D was recorded as calculation pass. The evidence note is “balloon feature”; the disposition remains visible so it cannot be averaged away.
The expected label or control was fixed before review. The visible note explains why the row received its disposition. The chart uses the published primary criterion, but the table is authoritative because it preserves exceptions that a single percentage would hide.
How to reproduce the check
- Download the CSV and read its labels, units, and synthetic/public-data notice before using it.
- Write the expected answers or decision rule before looking at a model response.
- Use the same bounded prompt and record the model or tool, access surface, and date.
- Preserve the raw response. Break prose into atomic claims rather than grading the tone of the whole answer.
- Recompute arithmetic with deterministic formulas and verify definitions against the linked primary sources.
- Record correct, partial, wrong, uncertain, and refused outcomes separately. Do not silently repair the model output before scoring it.
- Repeat material checks after a model, source, or workflow changes.
What the result means
The value of this result is diagnostic. It shows where a structured assistant can reduce search, formatting, or first-pass review work. It does not transfer responsibility for the underlying decision. A “pass” means the row met the published rule in this test, on this date, with these inputs.
The errors and open items matter more than a polished average. In money and business workflows, one missed assumption, stale fact, false match, or overconfident definition can dominate many correct low-risk rows. That is why the artifact keeps row-level outcomes and why a human reviews exceptions rather than receiving only a percentage.
Reproducibility also has limits. A reader can repeat the steps and inspect the same answer key, but a probabilistic model may not return identical wording. A useful rerun should therefore compare atomic claims, calculations, citations, and escalation decisions—not superficial phrasing.
Why this topic needs its own boundary
Banking and lending decisions can affect access to money, housing, and essential services. A model score or explanation must therefore sit inside data-quality, validation, notice, review, appeal, security, and monitoring controls rather than becoming the decision by itself.
The examples organize public terms and simplified scenarios. They do not evaluate eligibility, affordability, identity, creditworthiness, or a real product, and they do not replace the official disclosure for the reader’s jurisdiction.
A safer operating workflow
- Compare like-for-like terms.
- Include fees and timing.
- Show adjustable or balloon risk separately.
- Recompute formulas outside the model.
- Use official disclosures and a qualified adviser for a real decision.
How each control changes the decision
Control 1: Compare like-for-like terms. For this test, that control answers the bounded question “Can AI organize loan terms without converting an incomplete comparison into a recommendation?” without extending the result into an untested decision.
Control 2: Include fees and timing. For this test, that control answers the bounded question “Can AI organize loan terms without converting an incomplete comparison into a recommendation?” without extending the result into an untested decision.
Control 3: Show adjustable or balloon risk separately. For this test, that control answers the bounded question “Can AI organize loan terms without converting an incomplete comparison into a recommendation?” without extending the result into an untested decision.
Control 4: Recompute formulas outside the model. For this test, that control answers the bounded question “Can AI organize loan terms without converting an incomplete comparison into a recommendation?” without extending the result into an untested decision.
Control 5: Use official disclosures and a qualified adviser for a real decision. For this test, that control answers the bounded question “Can AI organize loan terms without converting an incomplete comparison into a recommendation?” without extending the result into an untested decision.
Keep data collection, model preparation, deterministic validation, and approval as separate stages. Use the least sensitive input that can answer the question. If removing personal or confidential data makes the result ambiguous, route the case to an approved human process instead of restoring secrets to an unapproved tool.
Calculations need an independent formula; current facts need a current primary source; classifications need an “uncertain” route; and irreversible actions need explicit authorization outside the model. Logs should capture the version, prompt, source date, output, reviewer, correction, and final disposition without retaining unnecessary personal data.
Limitations and professional boundary
The offers are synthetic and omit credit eligibility, tax effects, insurance, currency risk, jurisdiction, and personal affordability.
This publication provides general educational information. It does not know a reader’s finances, duties, jurisdiction, contracts, tax treatment, credit position, or risk tolerance. A qualified financial, accounting, tax, legal, lending, security, or other professional should review decisions with material consequences.
Primary sources
Verified 2026-07-26. Primary-source links can change; use the publication date and linked source to check for a newer version.
Bottom line
All four base-case payment calculations were within 0.2% of the local formula. The model’s initial ranking changed when fees and the holding period were made explicit, so no single offer was labeled best. The practical lesson is to make AI produce inspectable work inside a controlled process—not to make fluency the final control.