What counts as a recovery loop?
A recovery loop has five parts: an attempt, a check that fails, a diagnosis, a specific fix, and a retry, repeated until the check passes. The record is valuable only when all five survive and link to each other. A pile of error messages is not a loop, and neither is a clean final result with no trace of how it was reached.
AI researchers study the same shape in software agents. The Reflexion paper, published in the NeurIPS 2023 proceedings, describes agents that improve across attempts by writing reflections on feedback into a memory that guides the next try, rather than by retraining. Data vendors sell work built around failures and fixes too: Appen markets failure-mode taxonomies, golden trajectories and verifier design among its agentic AI data services. That shows the category is real. It does not show that any particular company's records are wanted.
In companies, recovery loops pile up wherever data moves between systems and someone has to make the numbers agree: ERP migrations, integration rollouts, bank file changes, month-end loads. Import tools record a lot of it. Microsoft's documentation for Dynamics 365 data import jobs says each run keeps execution details, including records that succeeded and failed per entity, and an execution log; failing records can be fixed and reimported. The same page notes that by default, job history and related staging data older than 90 days are deleted automatically. That single fact is why the questions below ask where reports went, not what the tool shows today. A full worked case is on the child page listed at the end.
What should an introducer ask about recovery loops?
Use this card with a controller, IT lead or migration project manager. None of the questions requires opening a file, and each has a typical answer that points to kept records and a typical answer that points to an opportunity that exists only in theory.
| Question | Answer that points to kept records | Answer that stays theoretical |
|---|---|---|
| 1. When a data load or integration fails, where does the error report go? | A shared mailbox, ticket queue or project folder that was kept | "The tool shows it," with no copy elsewhere |
| 2. Do analysts record what they changed between attempts? | Change tickets, versioned mapping files or a fix log | "We just fix it and rerun" |
| 3. Is there a written pass condition for each load, such as a control total? | Reconciliation workbooks with sign-off | "We check it looks right" |
| 4. How many migrations, integrations or system changes in five to ten years, and were project files archived? | Several projects, archived with the project record | Projects happened; nobody knows where the files went |
| 5. Who could authorize a conversation about licensing these records? | The CEO by name, and whether a parent or lender must agree | "Not sure who would decide" |
Question 1 matters most because of tool retention: if failed runs exist only inside an import tool with a short history window, the loop may already be gone. Question 2 separates a loop from repetition; without the change between attempts, a reviewer sees the same failure several times and cannot tell what fixed it. Question 3 decides whether anyone can say when the loop ended. Question 4 measures volume, since several imperfect projects usually count for more than one perfect one. Question 5 is not about records at all, but no introduction goes anywhere without it.
Reading the answers: kept records or a theoretical opportunity?
Count the answers in the middle column. Three or more suggest the company may hold real recovery loops and is worth introducing. One or two suggest the owner should first look for archived project files; the guide on historical data decisions in a QuickBooks to NetSuite migration helps an owner work out what older detail still exists. None means the opportunity is theoretical for now, however interesting the systems sound.
Illustrative. A controller at a 300-person distributor that changed ERPs three years ago gives these answers:
| Question | Answer given (Illustrative) | Column |
|---|---|---|
| 1. Error reports | Import errors were emailed to a project mailbox that IT archived | Kept records |
| 2. Fixes between attempts | Mapping spreadsheet saved as a new version before each rerun | Kept records |
| 3. Pass condition | Control totals per balance, signed off in the cutover workbook | Kept records |
| 4. Volume and archive | Two later integrations; nobody is sure their files were kept | Still theoretical |
| 5. Authority | CEO decides; no lender consent needed | Kept records |
Four of five point to records. That is enough to introduce the CEO, with the CFO as sponsor and a note that the later integrations need checking. It is not evidence that the records pass SourceX's acceptance check, that a buyer wants them, or what any payout would be.
Keep the conversation at metadata level throughout. Partners never ask for error files, workbooks or screenshots; what was migrated, roughly how much, over how many years, in which system and who owns it is all an introducer needs. Finance teams already value numbers that trace back to their source records, as the CFO dashboard resource shows, so the questions rarely feel unusual to a controller. Reconciliation breaks of other kinds are covered under reconciliation exceptions.
Who holds these records, including in a wind-down?
Migration and integration records usually sit with finance, IT, or an outside implementation partner. The implementation partner's copies belong to whoever the engagement contract says; the operating company should confirm it holds or can obtain them.
When a company is sold, restructured or wound down, its systems may pass to a buyer, an assignee or an investor. An assignee for the benefit of creditors may treat records as intangible assets of the estate, as the guide on ABC assignee intangible asset recovery explains, and special-situations investors sometimes find data assets in a portfolio company's systems. In those cases, establish who now holds legal authority over the records before anything else, and ask question 5 of that person.
The company must have 50+ full-time employees at peak (contractors excluded), a documented operating history, rights to license, and an authorized contact. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English. See who qualifies.
Unknown until SourceX reviews it: SourceX's current data needs for these records, the rights review result, any redaction of customer details, the company's offer and therefore any reward. Other record families are listed under data types, and agent failure patterns in general are covered in why AI agents fail.
Go deeper on recovery loops
- Recovery loops in an ordinary business workflow: a worked import case, attempt by attempt, from first failure to sign-off.
- Record gaps that make recovery loops hard to assess: the missing runs and undocumented fixes that most often break a loop, and how to spot them early.
- Recovery loops review packet: what an evidence packet for SourceX's review should contain, item by item.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
If three or more of a client's answers point to kept failed runs, fixes and sign-offs, register as a partner, check who qualifies, and read the program terms before introducing the CEO, with the CFO or CIO as sponsor.