A rule receipt explains the agent's decision path.
For high-attention channels like SMS, the user should see more than a message. They should be able to inspect why that message existed at all.
Personal AI agents need rule receipts
As personal AI agents move from chat into action, users need to inspect why the agent texted, browsed, escalated, summarized, or changed a deliverable. Rule receipts make that behavior explainable and correctable.
Receipts are the missing bridge between policy and behavior.
A policy says what should happen. A receipt proves which policy fired, which source was used, and which correction shaped the decision.
Source proof
Attach the message, calendar item, browser event, or task state that triggered the rule.
Policy version
Record which instruction or escalation policy was active when the agent acted.
Correction trace
Show whether a previous user correction changed the current behavior.
Review outcome
Mark whether the decision was useful, late, early, duplicate, or unnecessary.
A personal agent becomes trustworthy when each proactive action can answer: what rule fired, what source proved it, what correction shaped it, and what should change next.
What a useful receipt contains.
Keep receipts compact enough for review, but structured enough to improve the next agent decision.
Trigger and source
The event that caused the action, plus the source evidence. This matters for a text message AI assistant because every proactive text spends attention.
Rule and channel
The active rule, policy version, and chosen channel: interrupt, batch, escalate, or archive. This creates a clear path for review.
Correction and retest
The previous correction that influenced the rule, plus whether the next similar event should be watched before autonomy expands.
Rule receipt checklist
- Record the trigger event and source evidence.
- Record which rule or policy version fired.
- Record the selected channel: interrupt, batch, escalate, or archive.
- Attach any prior correction that shaped the rule.
- Ask the user or operator to score the outcome.
- Turn review failures into prompt or policy changes.
Where Supers fits
Supers workflows can use rule receipts across channels: proactive SMS, computer-use cache workflows, and deliverable updates when an AI agent builds websites. The same receipt shape lets the agent explain its behavior across different surfaces.
Sources
- Microsoft Research on interrupted work for attention-cost framing.
- Nielsen Norman Group on notification design for alert clarity and user-control principles.
- Supers for personal AI agent workflows across SMS, browser work, and generated outputs.
FAQ
Is a rule receipt a debug log?
No. Debug logs are for engineers. Rule receipts are compact explanations for users and operators.
Should every action have one?
Every proactive or autonomous action should. Passive chat replies may need lighter receipts.
What makes receipts useful?
They must include enough source and policy context to change future behavior.
How do receipts affect autonomy?
Permissions should expand where receipts show consistent good judgment and shrink where failures repeat.
What operators notice.
Receipts make review less emotional because the team can inspect source, policy, and correction history instead of arguing about intent.
"The receipt told us the agent used an old rule, so the fix was obvious."
"We stopped debating whether the alert felt noisy and started correcting the exact source gate."
"Autonomy reviews became clearer once every action had a rule trail."
Build agents that can explain why they acted.
Supers can help teams test personal agents that text, browse, and build while keeping action receipts visible and correctable.