Agentic Experience Design (AXD) is the discipline for designing trust-governed relationships between humans and autonomous AI systems. Founded in September 2024 by Tony Wood in Manchester, United Kingdom, AXD addresses how humans delegate, calibrate, observe, interrupt, and recover trust in agentic AI.
| Dimension | Traditional UX | Agentic Experience Design (AXD) |
|---|---|---|
| Primary material | Attention and affordance | Trust and delegation |
| User state | Present, navigating | Absent, delegating |
| Design output | Screens and interfaces | Outcomes and constraints |
| Temporal model | Session-based | Relationship-based |
| Success metric | Task completion | Trust calibration |
The Autonomy Gradient Design System is an AXD framework that defines five levels of agent autonomy, from fully supervised to fully autonomous. It provides structured criteria for when an agent should operate at each level and how to design the transitions between levels.
The five levels are: Level 1 (Supervised) where the agent suggests and the human decides; Level 2 (Guided) where the agent acts with pre-approval; Level 3 (Monitored) where the agent acts autonomously with real-time oversight; Level 4 (Trusted) where the agent acts autonomously with periodic review; Level 5 (Autonomous) where the agent acts independently within delegated scope.
The appropriate autonomy level depends on three factors: task criticality (higher stakes require lower autonomy), trust maturity (new relationships start at lower levels), and agent capability (demonstrated competence enables higher levels). The framework provides decision matrices for each factor.
Framework 03 of 12 · Active operation Phase · Autonomy calibration Trust is not a binary switch - it is a spectrum Commerce Application: Transaction approval levels Domains: Financial Services · Consumer · Enterprise Trust is not a binary switch - it is a spectrum. This dynamic, user-adjustable system answers how much autonomous decision-making the agent should exercise, across which task types, and who controls that dial. It operates across three axes: consequence (cost of mistake), confidence (agent certainty), and familiarity (how well the agent knows this user's preferences). Autonomy Gradient Design: Core Principles The question is not whether the agent can act autonomously. It is whether it should, for this task, at this consequence level, with this user's current trust state. The Autonomy Gradient answers that question dynamically, continuously, and transparently. Autonomy Gradient Design: Implementation Patterns Autonomy Gradient Design: Commerce Applications Autonomy without calibration is recklessness. Calibration without user control is paternalism. The Autonomy Gradient Design System navigates between these extremes by making the spectrum visible, adjustable, and responsive to context. Autonomy Gradient Design: Guidance for Teams Autonomy Gradient Design: Lifecycle Connections Autonomy Gradient Design: What Comes Next The Autonomy Gradient responds to trust state. The next framework - Autonomy Gradient Design: The Framework Ecosystem Navigate the complete lifecycle of Agentic Experience Design. Each framework addresses a distinct phase of the human-agent relationship. Multi-Agent Orchestration Visibility Model Agent Memory & Context Continuity Framework Explainability & Observability Design Standard Onboarding & Capability Discovery Framework Ethical Constraint & Value Alignment Architecture