What records connect a failed automation to the human action that repaired it?
Three groups of records, joined by one key. Without the key, they are three unrelated piles.
| Record | Typical system | What it shows | Join key | Survival risk |
|---|---|---|---|---|
| Failed run | Automation platform run history | When it failed, which step, the error | The item it was processing | Short retention; lost if the automation is deleted |
| Item record | ERP, CRM or helpdesk | The order or case the run should have handled | Order or case number | Usually kept for years |
| Manual tracker | Shared spreadsheet or list | Which items staff handled by hand, by whom | Order number in a column | Often overwritten or archived informally |
| Fix ticket | IT or operations ticketing | Diagnosis, decision to repair or retire | Automation name, error text | Kept, but rarely linked to items |
| Audit entries | System people typed into | That a person, not the integration, created the record | Record ID and user | Depends on whether auditing was on |
Run history is the fragile part. Microsoft's documentation for Power Automate points users to a 28-day run history to find a failed run, where the failed step is marked and its error shown. Zapier documents run statuses such as Errored, On hold and Safely halted, and requires errored steps to be replayed within 60 days of the original trigger event. It also notes that failed runs cannot be replayed once the Zap has been significantly changed. In both cases, the detail of a failure is easiest to capture while it is recent.
An abandoned automation, week by week (Illustrative)
A fictional 170-person specialty food distributor built an automation that copied web-shop orders into its ERP. Here is what happened, and which record each event left behind.
- Week 0. The automation runs cleanly. Run history shows successful runs, one per order.
- Week 1. The web shop changes its order format. Runs start failing at the step that maps product codes. Each failed run records the order number it was processing.
- Weeks 1 to 6. Customer service rekeys orders into the ERP by hand. A shared spreadsheet lists each order number, who entered it and any corrections. The ERP's audit trail shows those orders created by named staff, not by the integration user.
- Week 3. An IT ticket records the diagnosis: the product code field moved. Fixing it needs a developer the company does not have.
- Week 6. Operations decides to switch the automation off. The manual process, with its spreadsheet, becomes standard.
- Month 5. A new integration replaces it. The old automation is deleted, and with it, its run history.
Events 2 and 3 are the valuable pair: a machine failing on a specific order, and a person completing the same order correctly, with corrections noted. Event 6 is why timing matters. If nobody exported the run history before deletion, only the human side remains.
Ownership: the question that stops most conversations
Abandoned automations are often owned by nobody in particular. The person who built one may have left, and the account it runs under may be theirs.
That has practical consequences. Microsoft notes that in Power Automate's automation center, seeing a flow's run history requires being its owner, co-ownership not being enough, or having environment-wide access. So the first question is not "what is in the runs?" but "who can still see them, and who owns the process today?"
For an introduction, establish at metadata level:
- Which platforms ran automations for the process, and roughly when they were built, failed and retired.
- Who owns the process now, usually an operations, finance or customer service manager.
- Whether run history still exists, or only the human side (tickets, trackers, audit entries).
- What key links them: an order, invoice or case number present in both the failed run and the manual record.
- Whose data it is. Orders from the company's own customers are usually the company's records. Automations run on behalf of clients may belong to the client.
- Who can authorize a licensing discussion: the CEO, usually the owner in an owner-run company.
The company also needs 50 or more 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.
Why the human side carries the value
An automation's success records show the easy path. Its failure records show where the easy path broke. The human records show what a capable person did instead, including the corrections they made along the way, such as a discontinued product swapped for its replacement. That last part is the judgment automation could not handle, and it is what our explainer on workflow data means by records of how work actually gets done.
Quality checks done by hand during the workaround, if recorded, are related evidence; our page on quality assurance workflows covers that record type. If the company once wrote a design document for the automation, as described in our guide to the RPA process design document, it states the intended steps, which makes the gap between plan and reality visible.
Limits and unanswered questions
- Run details can include personal data. Step inputs and outputs often carry names and addresses. Partners never ask for exports; rights and de-identification are handled in SourceX's review.
- Volume is unknown until reviewed. Six weeks of rekeyed orders may be a few hundred records or many thousands. No one should estimate value from a conversation.
- Fragile joins. If the manual tracker used customer names instead of order numbers, linking may not be possible.
- Two separate payments. Any company compensation is agreed between the company and SourceX. Your reward is funded separately and never reduces it.
For examples of the wider record family, see the software workflow example pack, and for how agent connectors differ from classic automation tools, MCP vs Zapier. More guides are in the guides hub.
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 an operations leader mentions an automation that failed and a team that covered by hand, with order numbers on both sides, register, check the company on who qualifies, and read the terms before introducing the owner.