What Is a Decision Card?
A decision card is the structured handoff an AI support agent presents to a human when it is not cleared to auto-send. Instead of sending a risky answer blind or dropping the ticket into a queue with no context, the AI packages the customer request, its recommended reply or action, its reasoning and sources, its confidence, and which guardrail tripped, alongside clear labeled choices. It is the mechanism that makes human-in-the-loop support fast instead of slow.
What a decision card contains
- The request and context: the customer message, history, tier, and sentiment.
- The recommended reply or action: what the AI would do if it acted alone.
- Reasoning and sources: why, grounded in retrieved knowledge.
- Confidence and the tripped guard: how sure it is and what held it back.
- Clear choices: approve, edit, send, escalate, or reject.
Why it beats a plain escalation
A plain escalation hands a human a cold ticket and makes them redo the investigation from scratch. A decision card hands them a near-finished decision, so most cards clear in seconds. It also creates a training loop, because every choice the human makes is signal the model learns from. The card turns oversight from a bottleneck into a flywheel.
How Resolve uses it
In Resolve, the decision card is the core human-in-the-loop surface. Any time Reva falls below its confidence threshold or hits a guardrail, it produces a card rather than acting blind. Edits become preference pairs, approvals confirm patterns, and escalations become new cases, so clearing cards steadily raises how much the AI can safely handle on its own.