DecisionManager

AI-assisted rule authoring

Multi-provider AI authoring with ordered fallback & content-hash cache

Generate BAL 1.0 rules, decision table drafts, and test scenarios from plain-English prompts — across Anthropic, OpenAI, Gemini, Bedrock, or your own self-hosted endpoint. One provider's outage doesn't stop a migration run. Every draft is human-reviewed before it ever touches a decision service.

Supported AI providers

Configure an ordered chain — the engine tries each in priority order. On rate-limit or 5xx, it falls through transparently.

Anthropic Claude

claude-sonnet-4-5

OpenAI GPT-4o

gpt-4o

Google Gemini

gemini-2.5-pro

AWS Bedrock

claude-3-5-sonnet

Self-hosted LLM

OpenAI-compatible API

+ any OpenAI-compatible endpoint (e.g., Ollama, vLLM, Azure OpenAI, LM Studio)

How it works

Ordered provider fallback

On rate-limit (429) or server error (5xx), the engine moves to the next provider without interrupting the authoring workflow. The draft BAL rule arrives regardless of which provider generated it.

Content-hash cache

Every AI artefact is stored against a hash of (prompt, model, temperature). Re-runs return the cached draft — same input, same draft, zero additional provider cost. Essential for large migration runs.

Human-in-the-loop enforcement

AI drafts are never auto-persisted. Every generated rule passes through draft → in_review → approved before it can reach a decision service. The audit trail records whether the draft came from cache or a live provider.

What teams use AI authoring for

  • ODM migration acceleration

    Feed an IBM ODM BAL export to the AI; it generates equivalent BAL 1.0 drafts for human review. Content-hash cache means a re-run on a 500-rule set doesn't re-spend tokens.

  • Policy-to-rule translation

    Business analysts describe a regulatory constraint in plain English; the AI generates a decision table draft with test scenarios. A reviewer checks the logic before publish.

  • Test scenario generation

    Given a decision table, the AI generates edge-case test payloads covering boundary conditions — saves hours on manual test authoring for large tables.

  • Rule gap analysis

    Describe an expected coverage area; the AI suggests rules that may be missing from an existing decision table, flagging potential gaps for human review.

Try AI authoring on your rules

The AI authoring console is included in every plan, including the free trial. No credit card required.