The Binary Automation Toggle Trap
TRIGGER
Teams want agents to handle routine work autonomously, but jumping straight to full automation creates trust issues—users don't know which tasks agents handle well, and agents don't have the feedback loops to improve on failures.
APPROACH
Cursor and Linear designed a progression path: start with manual assignment (user explicitly assigns issue to Cursor agent), evolve to label-based triggers ('with this label, let agent handle it'), then to rule-based auto-assignment ('in this project, with this label, and t-shirt size small or less, let the agent handle it'). Each stage gives users control appropriate to their trust level while building toward richer automation.
PATTERN
“Binary "agent on/off" toggles fail both power users wanting more control and cautious users wanting less. Build the automation ladder upfront—manual trigger to label-based to rule-based—so users graduate at their own pace while the agent proves competence at each level.”
✓ WORKS WHEN
- Agent reliability varies by task type, making blanket automation risky
- Users have different risk tolerances and need to opt-in at their comfort level
- Task metadata exists to enable rule-based filtering (labels, size estimates, project tags)
- Failures at higher automation levels can gracefully fall back to lower levels
✗ FAILS WHEN
- All tasks have similar complexity and risk profiles, making graduated automation unnecessary
- Task metadata is unreliable or inconsistently applied across the organization
- Users expect immediate full automation and won't tolerate a trust-building period
- Agent performance doesn't meaningfully correlate with task attributes available for filtering