How it works
Describe the requirement in your own words. Automation proposes a rule built only from the facts and outcomes the product allows, restates it in plain English from the rule itself and suggests examples. Your team answers any questions, checks the examples, and a reviewer approves the release. The product then applies that exact rule to every job.
- Describe: the requirement, as written by the person who owns it.
- Clarify: specific questions, such as which field identifies a customer group.
- Propose and restate: a rule, with a plain restatement generated from the rule.
- Test and approve: agreed examples, then an independent review.
- Apply: the product runs the approved rule, the same way every time.
A rules engine or an AI agent?
An AI agent decides each case afresh, so the same input can produce a different answer and the reasoning is hard to inspect afterwards. An approved rule gives the same answer for the same facts every time, and its explanation comes from the rule and the recorded facts. Automation uses AI where it helps, in writing the rule, and approved rules where predictability matters, in production.
What the AI is never allowed to do
The AI cannot invent a fact, an engineering limit, a permission or a physical result such as ‘inspection passed’. It cannot approve or activate a rule, and it is never called when a product makes a decision. Every part of your requirement becomes a proposed rule, an open question or a stated gap; nothing is quietly dropped.
Plain English or the structured editor
The structured editor builds the same rule by choosing the decision, the approved outcome and typed conditions. Both routes go through the same validation, tests and review, so a team can use AI for some rules and not others, or switch it off entirely.
Explaining a decision afterwards
Open any held job or decision to see the rule version and the facts it used. An optional assistant can answer follow-up questions from those records and the Help guides, linking to each record it relies on. When records are missing, it says so rather than guessing.
What SmartFACT Automation does not do
- Make decisions in production with AI. Approved rules run without a model call.
- Choose outcomes a product does not allow, or run scripts, SQL or free-text conditions.
- Change a live rule without a passed test set and an approval.
Common questions
Does AI make decisions on the shop floor?
No. AI helps write the rule. Each product applies the approved rule exactly, with no AI call when the decision is made.
Can the AI change a live rule?
No. A proposal, a test result, an approval and an activation are separate recorded steps, and only an approved release reaches a product.
What if the AI misunderstands the requirement?
It asks a question when the requirement is unclear. If it still proposes the wrong rule, the plain restatement and the agreed examples show the difference before anyone can approve it.
What happens if the AI service is unavailable?
Approved rules keep running in every product, and the structured editor works without it.
Related pages
Trials are by invitation and start with one agreed workflow. Scope and arrangements are agreed with you before access.
