Answer in brief: The workflow produced 36 exact matches, eight useful review candidates, four correctly retained exceptions, and two false suggested matches. No item was auto-cleared.
Where can AI reduce reconciliation work without being allowed to clear an exception? 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 asked OpenAI GPT-5.6 in a Codex editorial session to propose matches across 50 synthetic pairs using amount, date, and reference. A deterministic pass confirmed exact pairs; a human-reviewed queue handled the rest.
The original asset is a synthetic ledger/bank pairs and an exception disposition 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
The workflow produced 36 exact matches, eight useful review candidates, four correctly retained exceptions, and two false suggested matches. No item was auto-cleared.
The row-level outcome distribution was exact: 36, exception: 4, review: 8, wrong: 2. 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 |
|---|---|---|
| R01 | exact | date and amount |
| R02 | exact | date and amount |
| R03 | exact | date and amount |
| R04 | exact | date and amount |
| R05 | exact | date and amount |
| R06 | exact | date and amount |
| R07 | exact | date and amount |
| R08 | exact | date and amount |
| R09 | exact | date and amount |
| R10 | exact | date and amount |
| R11 | exact | date and amount |
| R12 | exact | date and amount |
| R13 | exact | date and amount |
| R14 | exact | date and amount |
| R15 | exact | date and amount |
| R16 | exact | date and amount |
| R17 | exact | date and amount |
| R18 | exact | date and amount |
| R19 | exact | date and amount |
| R20 | exact | date and amount |
| R21 | exact | date and amount |
| R22 | exact | date and amount |
| R23 | exact | date and amount |
| R24 | exact | date and amount |
| R25 | exact | date and amount |
| R26 | exact | date and amount |
| R27 | exact | date and amount |
| R28 | exact | date and amount |
| R29 | exact | date and amount |
| R30 | exact | date and amount |
| R31 | exact | date and amount |
| R32 | exact | date and amount |
| R33 | exact | date and amount |
| R34 | exact | date and amount |
| R35 | exact | date and amount |
| R36 | exact | date and amount |
| R37 | review | date shift |
| R38 | review | date shift |
| R39 | review | date shift |
| R40 | review | date shift |
| R41 | review | date shift |
| R42 | review | date shift |
| R43 | review | date shift |
| R44 | review | date shift |
| R45 | exception | missing counterpart |
| R46 | exception | missing counterpart |
| R47 | exception | missing counterpart |
| R48 | exception | missing counterpart |
| R49 | wrong | false suggested match |
| R50 | wrong | false suggested match |
Reading the evidence row by row
- R01 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R02 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R03 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R04 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R05 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R06 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R07 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R08 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R09 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R10 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R11 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R12 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R13 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R14 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R15 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R16 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R17 was recorded as exact. The evidence note is “date and amount”; the disposition remains visible so it cannot be averaged away.
- R18 was recorded as exact. The evidence note is “date and amount”; 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
Business-finance workflows combine documents, accounting policy, timing, and deterministic arithmetic. An assistant can help prepare a queue or explanation, but the chart of accounts, reconciliation rule, formula, materiality threshold, and posting authority need accountable ownership.
Synthetic records make the test repeatable without exposing suppliers, employees, customers, bank accounts, or tax information. They also make the boundary visible: passing a clean example does not validate a production feed with duplicates, foreign exchange, split transactions, and incomplete documents.
A safer operating workflow
- Auto-match only deterministic exact rules.
- Keep suggestions separate from posted entries.
- Require evidence before clearing an exception.
- Use no real account numbers in unapproved tools.
- Retain an audit trail of overrides.
How each control changes the decision
Control 1: Auto-match only deterministic exact rules. For this test, that control answers the bounded question “Where can AI reduce reconciliation work without being allowed to clear an exception?” without extending the result into an untested decision.
Control 2: Keep suggestions separate from posted entries. For this test, that control answers the bounded question “Where can AI reduce reconciliation work without being allowed to clear an exception?” without extending the result into an untested decision.
Control 3: Require evidence before clearing an exception. For this test, that control answers the bounded question “Where can AI reduce reconciliation work without being allowed to clear an exception?” without extending the result into an untested decision.
Control 4: Use no real account numbers in unapproved tools. For this test, that control answers the bounded question “Where can AI reduce reconciliation work without being allowed to clear an exception?” without extending the result into an untested decision.
Control 5: Retain an audit trail of overrides. For this test, that control answers the bounded question “Where can AI reduce reconciliation work without being allowed to clear an exception?” 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
Real feeds contain duplicates, fees, split payments, foreign exchange, reversals, and timing differences not represented here.
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
- U.S. SBA: manage your finances
- Federal Reserve revised model risk guidance SR 26-2
- NIST AI Risk Management Framework
Verified 2026-07-26. Primary-source links can change; use the publication date and linked source to check for a newer version.
Bottom line
The workflow produced 36 exact matches, eight useful review candidates, four correctly retained exceptions, and two false suggested matches. No item was auto-cleared. The practical lesson is to make AI produce inspectable work inside a controlled process—not to make fluency the final control.