AI Business Rules
AI Business Rules let you define data validation and quality checks in plain language, then have Adeptia automatically generate the logic needed to evaluate your JSON data against those rules.
What AI Business Rules do
Instead of writing validation logic by hand, you describe what a rule should check in natural language. Adeptia's AI layer interprets that description and produces executable logic that processes your JSON data and returns a clear pass or fail result for each record it evaluates. This makes it practical to enforce complex data quality standards across integrations without requiring deep technical expertise for every rule.
How it fits into your integrations
AI Business Rules operate on JSON data that flows through your integration pipelines. When a rule runs, it receives a JSON payload, applies the logic derived from your natural-language description, and returns a structured result indicating whether each record passed or failed the check. You can also specify which fields from the data should be included alongside the pass/fail result, giving downstream processes the context they need to act on failures.
What you provide when configuring a rule
When setting up an AI Business Rule, you supply several pieces of information that shape how the rule is generated and applied.
| What you provide | Purpose |
|---|---|
| Rule name and description | Identifies the rule and provides the natural-language definition of what the rule should check |
| Result fields | The data fields whose values should be returned alongside the pass/fail outcome, so you can trace exactly which record or value triggered a result |
| Data path | The location within the JSON structure where the rule should iterate, used when your data contains repeating elements such as line items or order rows |
| Additional instructions | Optional guidance that further refines how the rule logic is generated |
| Sample JSON data | A representative payload used to validate that the fields referenced in your rule definition actually exist in the data |
Single-record and multi-record evaluation
AI Business Rules support two evaluation modes. In single-record mode, the rule evaluates the entire JSON payload as one unit and returns a single pass/fail result, optionally accompanied by up to three result fields. In multi-record mode, the rule iterates over a repeating section of the JSON, evaluating each element individually and returning a pass/fail result for each one. Multi-record mode supports up to ten result fields per record, giving you detailed traceability across large datasets.
Data validation before rule generation
Before generating rule logic, Adeptia checks that every field path referenced in your rule description is actually present in the sample JSON data you provide. If a referenced field is missing, you receive a clear message identifying which field could not be found, so you can correct the rule definition or update the sample data before proceeding.
Results and traceability
Every record evaluated by an AI Business Rule receives a result of either pass or fail. If a record cannot be evaluated due to a data issue, it is automatically marked as failed rather than causing the overall process to stop. For multi-record rules, each result also includes location information that identifies where in the JSON structure the evaluated element came from, making it straightforward to trace failures back to their source.