Create Action Funnel (CREATE)
CanonicalConfidence
Cognitive Load
High
Evidence
production validated
Impact
product
Ethical Guardrail
Agent must cite the specific data source or policy reference before executing any action on behalf of the user.
Design Intent
Every meaningful action in software is the result of a cognitive journey. The CREATE funnel asserts that action is not a moment -- it is a sequence of psychological gates. Users do not fail to click a button. They fail to notice, emotionally engage, believe the action is worthwhile, feel capable, or act at the right moment. This pattern exists to help designers and AI agents identify exactly where behavior breaks down and intervene with precision rather than intuition.
Psychology Principle
Five mental events must align for any user action: Cue, Reaction, Evaluation, Ability, Timing, then Execute.
Description
The CREATE funnel is the foundational diagnostic model for every flow in a product. Each stage represents a cognitive gate the user must pass through before acting. If any gate fails, the action does not happen. For AI agents, the funnel provides a diagnostic framework: when a user stalls, identify which gate is blocked and intervene there. When used correctly, CREATE transforms product design from guesswork into behavioral engineering.
When to use
Every single product flow. This is the default diagnostic lens. If users are not completing an action, the responsible CREATE stage must be identified before UI changes are proposed.
Example
Stripe payment processing flow: email notification triggers Cue, payment summary shapes Reaction, line-item breakdown supports Evaluation, pre-filled payment methods enable Ability, due-date awareness drives Timing, one-click pay now enables Execute.
Autonomy Compatibility
Behavioral Objective
Users complete critical workflow actions before deadlines without abandonment.
- Users complete multi-step workflows without abandonment
- Users escalate ambiguous items earlier in the process
- Users develop confidence in AI-supported decision making
Target Actor
role
Product Manager / Product Analyst / Engineering Team Lead
environment
Time-constrained, outcome-sensitive product workflows
emotional baseline
Mild anxiety, high accountability pressure
ai familiarity
medium
risk tolerance
low
Execution Model
cue
User must notice the opportunity to act.
User receives invoice but does not open payment flow.
reaction
User forms emotional response.
User dismisses alert or postpones action.
evaluation
User decides whether action is worth effort.
User opens flow but abandons before starting.
ability
User must feel capable of completing action.
User stalls in form or navigates away.
timing
User must perceive urgency or appropriateness.
User delays until after the deadline.
execute
Final irreversible or meaningful action occurs.
User hesitates indefinitely at confirmation.
Failure Modes
Excessive friction causes shadow workflows (email/Slack bypass)
Use Confidence Signaling to calibrate urgency
Over-automation reduces user ownership of outcomes
Bind automation to Autonomy Dial position
Poor cue placement leads to missed critical actions
Position cues within existing workflow context
Cognitive overload at Ability stage causes abandonment
Apply Progressive Disclosure for complex workflows
Timing signals too aggressive trigger alert fatigue
Adapt signal frequency based on user response history
Agent Decision Protocol
Triggers
- User fails to act within expected workflow window
- High-risk decision detected
- User repeatedly abandons task
Escalation Strategy
L1: Diagnose failing CREATE stage via behavioral_signals
L2: Nudge -- resurface cue, adjust timing, add confidence signal
L3: Restructure -- simplify form, add progressive disclosure, restructure flow
L4: Constrain -- lock Autonomy Dial to confirm_execution, add Strategic Friction
L5: Yield -- flag for human designer or domain expert review
Example
User abandons invoice payment flow -> ability_failure signal fires (form_focus_time > 90s) -> L1 diagnoses Ability failure -> L2 nudge adds inline help -> still failing -> L3 retrieves Action Structuring -> proposes simplified 3-field payment form with pre-filled amounts.
Behavioral KPIs
Primary
- % of critical actions completed before deadline
- Median time from notification to action initiation
- Workflow abandonment rate
Risk
- Number of post-execution error discoveries
- Escalation frequency
Trust
- User-reported confidence in AI recommendations
- Autonomy Dial usage distribution
Behavioral Signals
cue_failure
alert_viewed=false AND time_since_notification > 600s
scroll_depth < 20% on action page
reaction_failure
alert_dismissed=true
action_postponed=true
evaluation_failure
preview_opened=true AND action_started=false AND dwell_time > 30s
ability_failure
form_focus_time > 90s
field_error_count > 3
navigation_away_from_form=true
timing_failure
action_started=true AND time_to_deadline < 24h
action_completed_after_deadline=true
execute_failure
confirmation_dialog_open_time > 30s
confirmation_abandoned=true
Decay Monitoring
Revalidate when
- Autonomy capabilities expand
- Workflow complexity increases
- New user segments onboard
Decay signals
- Increased abandonment at previously stable stages
- Reduced alert engagement
- Rise in manual workarounds