Docs / Pricing Advisor / Workflow
Recommendation workflow
The full lifecycle from generation to version history. Human-approved at every step.
Overview
Recommendations flow through six stages. You stay in control: nothing changes until you apply.
Recommendation generated
Chargly evaluates provider cost, current credits, feature, model, and margin target. A recommendation candidate is produced with confidence and estimated lift. The system persists it for review.
Surfaced in dashboard
The recommendation appears in the Chargly dashboard as a card. You see feature, current vs suggested credits, reason, confidence, and estimated monthly lift. Apply and Reject controls are available (Growth plans).
Inspected in detail
Open the detail drawer to see the full recommendation object: reason, metadata, and any supporting data. Use this to decide whether the suggestion aligns with your strategy before acting.
Applied or rejected
You choose. Apply accepts the suggested credits and creates a new pricing rule version. Reject records the decision and preserves audit history. No silent overwrites — every action is explicit.
Immutable version created
Applying creates a new immutable pricing rule version (e.g. v1 → v2). The old version remains in history. You can always see what changed and when.
History preserved
Both applied and rejected recommendations stay in the recommendation history table. Version badges, timestamps, and outcomes are retained for trust and audit.
Human in the loop
No autonomous pricing changes. You apply or reject. MCP tools can surface and explain recommendations, but applying still requires your approval in the dashboard or via an explicit agent action you authorize.