Agree the expected outcome before you look at the result
An acceptance case records the inputs, the outcome your team expects and why, decided before the rule is run. Running the new rule and copying its answer into the test proves nothing, so Automation never treats a rule’s own output as its expected result.
The cases every rule needs
A rule that adds a support above 1,200 mm needs a case at 1,200 mm and one just above, not only an ordinary order. Automation asks for the cases that catch most mistakes.
- A match and a non-match.
- The boundary value itself.
- Missing or unknown data, kept separate from false.
- An invalid value.
- Overlapping rules and no matching rule.
- A blocked item, and a record that has changed since the draft began.
Compare a change on the same inputs
Before review, Automation runs the current and proposed rules over the same frozen inputs and summarises which cases are unchanged, newly allowed, newly blocked or routed differently, with each case one click away. Where past inputs were not kept, it says so instead of inventing them.
Tests that never print, book or reserve
A test evaluates the rule and nothing else. It never prints a label, books a courier, reserves stock or sends an order, so you can test against realistic data without anything happening on the floor.
When a required case fails or is skipped
Every required case must pass before the set can support an approval. A skipped, cancelled or unfinished case leaves the set incomplete, and removing a failing case creates a new reviewed case set rather than a green result. Suggested cases from AI stay suggestions until someone sets the expected outcome.
What SmartFACT Automation does not do
- Claim complete coverage. Tests and checks cover the declared cases and values, not every possible input.
- Perform production actions from a test.
- Accept a change while a required case is failing, skipped or unfinished.
Common questions
Can AI write the tests?
It can suggest unusual and boundary cases. A person decides the expected result before a suggested case counts towards approval.
What is a regression test for rules?
A case whose expected outcome comes from the approved release. If a change alters it, you see the difference and decide whether it was intended.
Does a test print labels or book couriers?
No. Tests evaluate the rule only and never act on the floor.
Related pages
Trials are by invitation and start with one agreed workflow. Scope and arrangements are agreed with you before access.
