Industry field guide
Commercial bank returned-item records for AI review
A deposit may appear in a ledger before an item in it is returned.
- Sector
- Finance and insurance
- 2022 NAICS
- 522110
- Practice pack
- Example and answer key (JSON)
- Updated
What this guide tests
Which original item is affected by R1, and does D2 establish a replacement posting for the returned item?
See the specimen and fields · Work through the fictional records
Task and source context
A deposit may appear in a ledger before an item in it is returned. An assistant investigating a discrepancy needs to link the return to that item and preserve any unaffected entries. FFIEC check-imaging materials describe the processing context for the fictional benchmark. FFIEC remote deposit capture guidance
Create a dummy deposit with two items and a later return that applies to one. Add a third entry with the same amount but a different check reference. Give the reviewer a mock statement and reconciliation note. Ask which item was affected, whether the records establish a new posting, and which source backs the conclusion. Test the distinction between the original item, a return, and a separate deposit.
Use synthetic documents and nonvalid account and routing placeholders. Keep actual check images, signatures, customer information, account balances, and internal fraud rules outside the sample. A bank reviewer should confirm the fictional labels and use case.
Industry research
The U.S. Census Bureau’s industry definitions establish which activities these codes cover. The workflow sources in this guide describe the task or record type. Fiund’s pilot below uses fictional material to test a specific decision within that scope.
522110 · Commercial Banking
This industry comprises establishments primarily engaged in accepting demand and other deposits and making commercial, industrial, and consumer loans. Commercial banks and branches of foreign banks are included in this industry.
Build a pilot sample
Set up this pilot
Create two synthetic deposit items with equal amounts and one return reference. Add an unlinked reposting and grade matching the reversal separately from proving a replacement.
Prepare the test packets
- Start with the worked case. Use the fictional record types below as a guide to the source files you need. Give each packet a stable ID and save its source versions with it.
- Change one fact. Make one packet with enough evidence to answer the question and one where the evidence points to a different answer. Keep the other fields fixed so a reviewer can tell what changed.
- Remove the decisive evidence. Give the reviewer a third packet with a plausible near-match in its place. Have them write what can still be established and what evidence would resolve the rest.
Fields the reviewer needs
- Deposit and item references
- Preserves component-level identity.
- Return target and amount
- Limits the adjustment's scope.
- Replacement link and reconciliation note
- Distinguishes an established reposting from coincidence.
Use How to prepare a financial workflow data sample for AI to turn the specimen into a brief with a defined task, source and review owner.
Fictional exercise
One returned item does not reverse every equal-value deposit
An inert ledger fixture uses invented item aliases and practice units. It includes no valid account or routing information and cannot initiate a banking action.
| Record ID | Source | What it records |
|---|---|---|
| BI1 | Deposit D1 | Items C-A=70 units and C-B=30 units; total 100. Posted in mock ledger day 1. |
| BI2 | Return R1 | Day 3 return of 30 units explicitly references C-B within D1. Reconciliation note maps R1 to that item. |
| BI3 | Deposit D2 | Day 3 separate 30-unit item C-C. No reference to C-B, D1 or R1. |
Question for the model. Which original item is affected by R1, and does D2 establish a replacement posting for the returned item?
Show the reference answer
BI2 explicitly associates the 30-unit return with C-B in D1. BI1's C-A remains a separate 70-unit item unaffected by the supplied return record. BI3's D2 is also 30 units, but its item C-C has no replacement link to C-B. Preserve D2 as a distinct deposit and leave replacement status unresolved. Matching amounts and dates do not establish a shared transaction identity.
How to check the result
Ask for the answer, the record IDs that support it and any fact the evidence leaves unresolved. Have a reviewer compare the response against the reference before deciding whether the sample is ready to expand.
Errors to flag in this case
- Reversing the entire 100-unit deposit for a return linked to one component.
- Treating an equal-value later deposit as a replacement without a source relationship.
Change the inputs and run it again
Add an explicit correction linking a new item C-D to C-B. Require the replacement relationship without merging unrelated C-C.
Score the decision and the evidence separately
Save the reviewer’s expected answer, supporting record IDs and unresolved fields in a separate answer key. Score the model on the decision, its citations and its handling of the deliberate gap. A correct answer citing a different case counts as an evidence error.
Before expanding collection, have a second reviewer run the same checks on the held-out packet. Keep related versions, views and exports of the same case together so they stay on one side of the training and evaluation split.
Check permission and scope
Use fully synthetic cards and nonvalid identifiers. Exclude checks, signatures, customer histories, actual balances and internal fraud procedures.
Exclude real personal and financial data, credentials, underwriting decisions and individual eligibility conclusions.
Describe an authorized archive through Fiund’s existing-data assessment, or describe a new capture capability through the collection brief. The contributor agreement covers the terms for participation.
Sources and further reading
Resources for this workflow
Includes worksheet
How to prepare a financial workflow data sample for AI
Build a fictional or authorized finance example with the reporting unit and time cutoff intact.
Includes worksheet
Plan a sample of business data for AI
Describe the evidence without transmitting private account or customer information.
Company and program reviews
Use these reviews to compare sourcing, capture or evaluation approaches. Confirm the applicable program, accepted sources and terms with the provider.
Compare the approach
Scale AI review: data partnerships, Outlier and buyer services
Read the public financial-services partnership criteria and scope controls.
Compare the approach
Invisible Technologies review for expert-led AI training
Compare expert calibration and reviewer agreement for narrow financial document exercises.
Related industry examples
Industry guide · NAICS 522130
Credit union share reconciliation data for AI testing
An amount may appear in a control account before it is attributed to the correct share account.
Industry guide · NAICS 522180
Savings account rate-history records for AI evaluation
A savings product’s advertised rate today may differ from the rate in force for an earlier statement period.
Industry guide · NAICS 522220
Dealer-arranged finance contract data for AI testing
A dealership worksheet may describe a proposed deal while a finance company’s acceptance refers to a later contract.