Check twenty inspection records against six house rules. Then decide what they flag.
In real work, many checks are plain rules like these. Where rules fall short, such as reading free-text notes, a language model can do that part, tested first on your own past records, and your team reviews what it flags. Here a fixed set of rules does every step, so you can see how it flows. The records are invented.
- Run the checks.
- Open an exception. Accept it, or override it with a reason.
- Change a rule and run again.
20 records ready · not checked yet
See the resultsThe records
Invented rail and power assets. Dates count back from today, so the results are the same whenever you visit.
| Asset | Category | Location | Last inspected | Due | Signed off | Result | Notes |
|---|
The results
This sample is deliberately seeded so that every rule has something to find.
Run the checks to see what reaches your team.
Run the checks to see which records clear.
Run the checks to create receipts.
Your rules, your amendments.
v1.0, as writtenNew receipts cite the new versions. Old receipts keep theirs.
Limits
A word list misses 'Slight seepage at door gasket' (AF-0112). Where wording varies like this, a language model can read the notes, tested first against your own records, with anything it's unsure of going to a person.
Try the amendmentPlenty of checking needs no model at all. Part of a Proof is finding out which part does, and proving it on your records.
Field names follow AssetFlow, TinyPhi's asset and maintenance platform (early stage). AssetFlow does not run this demo.
Want this on your own records?
Start with a Proof: one workflow, run beside what you have, switched over only when your team decides.