Screen deal · PCT-069 · Sourced card · live demo
Screen an ABF data tape
Screens each new loan tape before an advance: duplicate collateral, assets that should have left the base, eligibility breaks and delinquency hidden by extensions.
Worked example: 266 of the 271 loans modified in August now report as current. Counting extended loans that now report current moves the 60+ day delinquent balance from 4.1% to 5.7% of the pool. See the walk-through ↓
Today vs automated
- Open the tapeAssociate · manual
- Sort for duplicatesAssociate · manual
- Check rules by handVP · manual
- Email / meeting hand-offAssociate → VP
- A new loan tape arrives from the originatorTrigger: starts on its own
- Test unique IDsAutomated · Document store
- Reconcile balancesAutomated · LLM
- Flag impossible fieldsAI draft, human review · Pipeline CRM
- Whether a break is a data error or something worse, and whether to fundApprove / override · VP
AutomatedAI draft, human reviewHuman onlyTools: Deal inbox, Document store, LLM, Pipeline CRM
Estimated valueestimate
100–900 analyst hours a year
$15k–$225k of time, vs an assumed $20k–$100k tool cost
Net: −$85k to $205k a year net of tool cost; at the low end of the assumptions it does not cover the tool cost.
Assumptions
- 8–16 analyst hours per tape today; 1–3 hours with an agent, including review
- 20–60 tapes a year for one team
- $150–250 per hour loaded cost
- Tool or build cost $20,000–100,000 a year (both)
Estimate, based on the assumptions shown; your numbers will differ. How we estimate
What else it can do for the team
- Reconcile tape totals to the servicer report
- Profile the pool against eligibility criteria
- Build roll-rate and vintage curves for underwriting
- Draft the question list for the originator
- Compare this month's tape with last month's for silent changes
Worked walk-throughKey catch: 266 of the 271 loans modified in August now report as current. Counting extended loans that now report current moves the 60+ day delinquent balance from 4.1% to 5.7% of the pool.
1. The input
A public loan-level tape: the Reg AB II asset-level data (Form ABS-EE, file cart20231.xml) for CarMax Auto Owner Trust 2023-1, filed September 15, 2026 for the August 2026 reporting period. Filing index on EDGAR. Public auto ABS tapes use the same fields a private ABF buyer gets in a data room.
2. What the AI produces
| Screen | Count |
|---|---|
| Asset records in the file | 75,490 |
| Loans with a balance at period end (Aug 2026) | 24,008 |
| Active loans 60+ days delinquent (currentDelinquencyStatus) | 589 |
| Active loans modified in the period | 271 |
| Active loans flagged repossessed | 33 |
| Active loans with payment-to-income above 25% | 437 |
| Active loans with no numeric credit score | 3,308 |
| Active loans with original amount above 150% of vehicle value | 3 |
| Active loans modified in August that now report 0–29 days past due | 266 of 271 |
| Active loans with at least one payment extension (paymentExtendedNumber) | 262 ($3.7m balance), 259 of them reporting current |
| 60+ day delinquent balance, as reported | $9.2m (4.1% of $224.3m active balance) |
| Same, adding extended loans now reporting current | $12.9m (5.7%) |
Counts computed from the filed XML with a short script (2026-10-18; extension rows 2026-10-19). Delinquency uses currentDelinquencyStatus (days); extensions use paymentExtendedNumber. Method for the extension figures: from the filed cart20231.xml (Form ABS-EE, CarMax Auto Owner Trust 2023-1, August 2026 period), active loans are those with an end balance above zero (24,008; $224.3m). 271 have reportingPeriodModificationIndicator = true, and 266 of those report 0–29 days past due. 60+ days past due: $9.2m of balance (4.1%). Adding the 259 loans with paymentExtendedNumber above zero that report under 30 days past due ($3.7m) gives $12.9m (5.7%). Re-verified from the file on 2026-10-19. The thresholds (25% payment-to-income, 150% of vehicle value) are illustrative screen settings, not standards. A flag is a row to look at, not a judgment on the pool.
3. What a person still checks
- Ask the servicer for the extension policy and roll rates of extended loans: if they re-default at a high rate, the reported 60+ day figure understates stress.
- Ask why 3,308 active loans have no numeric score: thin-file borrowers, a coding convention, or missing data.
- Reconcile the tape totals to the servicer report for the same period before trusting any ratio.
- Look up the field definitions in the SEC technical specification; delinquency is reported in days, not buckets.
- Review the flagged rows by hand and decide which are errors and which are real risk.
4. Where it breaks
- Codes differ by issuer and change over time, so a screen built on one shelf misreads another.
- Paid-off loans stay in the file; forgetting to filter them inflates every denominator.
- An anomaly screen finds outliers, not slow drift; roll rates by vintage need several periods.
How you get thereFour steps from prompting to an agentic workflow; most teams start at step 2.
Needs: an approved AI tool, a confidentiality policy.
Risk: unchecked answers, data pasted outside policy.
Needs: field definitions, a template, document parsing.
Risk: breaks silently when formats change.
Typical starting point
Needs: inbox/doc-store access, system write-back, logging.
Risk: acting on a misread; needs review gates.
The automated lane above
Needs: shared data model, permissions, audit trail, exception queue.
Risk: errors compound across steps; needs monitoring.
Solution optionsBuy (5 vendors listed), build or hybrid, with the trade-offs of each.
Lumonic, dv01, Intain, Oxane Partners, Setpoint
Fastest start; you adapt to the product's data model and pricing.
An LLM API, a document parser, a vector or document store, your data warehouse, and a review queue.
Fits your definitions exactly; you own maintenance and controls.
Vendor extraction feeding your own checks, templates and systems.
Common middle path; two things to integrate and govern.
Listing is not endorsement.
What changes for the team8–16 h → 1–3 h per tape (estimate); coverage goes from a sample of deals read in full to every inbound deal screened the day it arrives.
ABF tape hunter demo (SAMPLE) →
How it works
Full title: ABF data-tape anomaly screen: duplicate collateral, stale borrowing-base assets and impossible loan fields
“Tricolor's lenders found the double pledge after the money was out. A uniqueness join finds it before.”
Scenario
Public: an asset-based lender wants every new tape screened before an advance for duplicate collateral IDs, assets that should have left the borrowing base, and fields that break the facility's own eligibility rules.
What the human still decides
Whether a break is a data error or something worse, and whether to fund. The agent stops at the evidence.
Why it matters
Tricolor filed for Chapter 7 on 10 Sep 2025 with about 100,000 loan accounts as collateral after lenders found alleged double pledging and tape irregularities. At First Brands, receivables and inventory were allegedly used for on-balance-sheet debt, SPV financing and factoring at once.
What the record shows
Checked 2026-10-16.
Tricolor's Chapter 7 (10 Sep 2025) left about 10,000 vehicles and 100,000 loan accounts as collateral, after warehouse lenders found alleged double pledging and loan-tape irregularities. Its trustee alleges falsified performance data and fabricated receivables. At First Brands, receivables and inventory allegedly backed on-balance-sheet debt, off-balance-sheet SPV financing and factoring at the same time (Morgan Lewis, Feb 2026). Wolters Kluwer cites a case where weekly tape reconciliations caught a $100m double pledge. On this site's SAMPLE tape, deterministic rules catch 33 of 33 planted breaks with 0 false positives.
Try it: prompt and data
Screen {tape} against {facility reports}: list any asset ID, VIN, invoice or obligor ID appearing in more than one active facility; any asset sold or paid off but still in the borrowing base; and any row breaking the eligibility criteria in {agreement section}. Cite the row and rule for each break.Data
Public inputs only (SEC/EDGAR filings, published reports, public news). Every figure is cited.
Who uses it, vendors and build
- Segments
- ABF / asset-based, Fund finance
- Who uses it
- ABF / structured analyst, Portfolio manager
- Build
- Vendor + in-house
- In-house
- Deterministic joins plus rules. Estimate: 1 week for one asset class.
- Last verified
- 2026-10-15
Vendors that claim this capability
- dv01: abf-tape-review, borrowing-base
- Intain: abf-tape-review, borrowing-base
- Setpoint: abf-tape-review, borrowing-base
- Lumonic: borrowing-base
- Oxane Partners: borrowing-base
Tags reflect each vendor's own public claims at the linked page (checked 2026-10-15). Not tested, ranked or endorsed.
Sources and related
Next step
Related use cases
Get new walk-throughs
Not investment advice. Names marked SAMPLE are fictional. Found an error? See the corrections log.