What Is AXD? The Complete Guide to Agentic Experience Design

For three decades, every major design discipline assumed the user was present. User Experience (UX) mapped journeys through screens. Service design orchestrated touchpoints between people and organisations. Interaction design refined the micro-moments of click, swipe, and scroll. Each methodology shared a foundational assumption: the human is here, watching, deciding, acting.

Agentic AI broke that assumption. When an AI agent books a flight, negotiates a supplier contract, manages a portfolio, or triages a medical referral on someone's behalf, the human is absent. The most consequential moments in the experience happen when no one is watching. No existing design discipline was built for this condition.

Tony Wood founded AXD in September 2024 in the United Kingdom to address this gap. It is not conversational design, not prompt engineering, not AI ethics, and not UX with an "AI" label. It is a parallel discipline that treats the design of human-agent relationships as a first-class problem requiring its own vocabulary, its own frameworks, and its own principles.

AXD is built on five principles that distinguish it from every prior design discipline. These principles were articulated in the AXD Manifesto and govern all subsequent frameworks and vocabulary.

1. Agency Requires Intentional Delegation. Every agentic system begins with a designed act of delegation - a human granting permission for autonomous action. AXD designers architect this delegation: its scope, duration, constraints, and mechanisms for revocation. Without intentional delegation, autonomy becomes ambiguity. The Delegation Design framework provides the structural approach.

2. Trust Is the Primary Material. Where UX works in attention and affordance, AXD works in trust. Trust in autonomous agents is calibrated, contextual, and fragile - and it can be designed. Trust Architecture provides the structural framework for building, maintaining, and recovering trust across the human-agent relationship.

3. Absence Is the Primary Use State. The most consequential agentic experiences happen when no one is watching. Machine customers transact, agents decide, systems operate - all while the human is absent. Designing for absence is the defining challenge of AXD and the reason it cannot be reduced to traditional UX.

4. Relationships Have Temporality. Agentic experiences are not transactions. They are relationships that accumulate history and demonstrate character. The hundredth interaction matters more than the first. The Relational Arc governs how trust evolves over time, how autonomy expands with demonstrated competence, and how the relationship deepens from initial delegation to mature collaboration.

5. Outcomes Replace Outputs. When the path is chosen by the agent, the designer specifies the destination, not the journey. Outcome specification - defining what success looks like rather than how to achieve it - is the primary design artifact of AXD. This is a fundamental inversion from traditional design, where the designer controls the interface the user navigates.

The most common misconception about AXD is that it is "UX for AI". It is not. AXD and UX address fundamentally different design conditions. Understanding the distinction is essential for teams transitioning into the agentic era. A detailed comparison is available on the AXD vs UX page.

Presence vs Absence. UX assumes the user is present and interacting with an interface. AXD assumes the user is absent and the agent is acting autonomously. This single inversion changes every design decision.

Attention vs Trust. UX optimises for capturing and directing human attention through visual hierarchy, affordance, and interaction patterns. AXD optimises for building and maintaining trust through transparency, predictability, and recovery mechanisms.

Interface vs Outcome. UX designers specify what appears on screen - layouts, components, flows. AXD designers specify what results - the goals, constraints, and conditions under which the agent should seek human re-engagement.

Transaction vs Relationship. UX typically designs discrete interactions - a checkout flow, a sign-up form, a search query. AXD designs ongoing relationships that evolve over time, where the agent's track record and the human's growing confidence shape the experience.

These are not minor differences. They represent a shift in the fundamental materials, methods, and success metrics of design. Teams that attempt to apply UX methods to agentic systems will find themselves designing for a user who is not there.

AXD introduces a vocabulary of concepts that have no equivalent in traditional design. The full AXD Vocabulary contains 64 canonical terms. Here are the foundational concepts every practitioner should understand.

Trust Architecture is the structural framework for designing, building, maintaining, and repairing trust between humans and autonomous agents. It treats trust as a designable material with four layers: predictability, agency, communication, and evolution. Read the full Trust Architecture guide.

Delegation Design is the practice of structuring the transfer of authority from a human to an AI agent. It defines scope, constraints, permissions, and recovery paths to ensure delegated tasks are completed safely and in alignment with intent. Read the full Delegation Design guide.

Absent-State Design is the practice of designing for moments when the human is not present. This includes agent observability (how the human understands what happened), interrupt patterns (how the human re-engages), and recovery architecture (how trust is repaired after failure).

Intent Architecture is the framework for capturing, translating, and preserving human intent across the delegation lifecycle. It ensures the agent's actions remain aligned with what the human actually wanted, not just what they literally said.

The Consent Horizon is the temporal boundary beyond which a user's original consent to an agent's actions can no longer be assumed valid. It recognises that consent degrades over time as circumstances change, and forces agents to periodically re-validate their authority.

The Operational Envelope defines the boundaries within which an agent is authorised to act. It specifies the parameters of acceptable behaviour, the conditions that trigger escalation, and the hard limits that cannot be overridden.

The AXD Institute publishes 12 practice frameworks that translate the discipline's principles into actionable design methods. These frameworks cover the full lifecycle of human-agent interaction and represent the most comprehensive agentic design methodology published anywhere.

Trust Calibration Model provides the method for measuring, monitoring, and adjusting trust levels throughout the human-agent relationship. It defines trust signals, calibration triggers, and the feedback loops that keep trust aligned with agent performance.

Delegation Design Framework structures the transfer of authority with appropriate constraints and recovery paths. It covers scope definition, constraint specification, permission granting, and revocation mechanisms.

Intent Architecture Framework addresses the capture and translation of human intent into machine-executable mandates. It bridges the gap between what a human says they want and what the agent needs to execute.

Autonomy Gradient Design System calibrates the level of agent independence across different contexts and risk levels. It provides a structured approach to expanding autonomy as trust is earned and demonstrated.

Additional frameworks cover Failure Architecture, Orchestration Visibility, Interrupt Patterns, Recovery Design, Ethical Constraints, Outcome Specification, Observability Patterns, and the Relational Arc. Together, they provide a complete methodology for designing agentic systems that humans can trust, understand, and control.

Teams and individuals entering the agentic design space can follow a structured learning path through the AXD Institute's published resources. The path is designed to build understanding progressively, from foundational concepts to advanced practice.

Step 1: Read the Manifesto. The AXD Manifesto establishes the philosophical foundation - why AXD exists, what it addresses, and what it must not become. Start here to understand the discipline's purpose and boundaries.

Step 2: Learn the Vocabulary. The AXD Vocabulary introduces 64 canonical terms that form the shared language of agentic design. Understanding these terms is essential before engaging with the frameworks.

Step 3: Study Trust Architecture and Delegation Design. These two foundational concepts - Trust Architecture and Delegation Design - are the twin pillars of AXD. Every other concept and framework builds on these foundations.

Step 4: Explore the Practice Frameworks. The 12 Practice Frameworks provide the actionable methods for applying AXD principles to real systems. Begin with Trust Calibration and Delegation Design, then progress to Autonomy Gradient and Failure Architecture.

Step 5: Read the Observatory Essays. The Observatory contains 62 long-form research essays exploring specific aspects of agentic design, from machine customers and agentic commerce to trust recovery and consent horizons. These essays provide the depth and nuance that frameworks alone cannot capture.

Step 6: Assess Your Readiness. The AXD Readiness Assessment evaluates your organisation's preparedness for the agentic era across the Four Pillars of AXD Readiness.

AXD is the design foundation of agentic commerce - the emerging economic model in which AI agents act as autonomous participants in markets. When agents shop, negotiate, and transact on behalf of humans, the design challenge shifts from persuading a human buyer to being legible, trustworthy, and transactable by a machine.

Machine customers do not browse. They do not respond to brand storytelling or emotional marketing. They evaluate structured data, verify trust signals, and optimise for the outcome their human principal specified. Businesses that want to serve machine customers need AXD to design the trust architecture, delegation patterns, and outcome specifications that make their services agent-ready.

The AXD Institute's work on agentic commerce spans the Four Pillars of AXD Readiness, industry-specific analysis for banking and retail, and the emerging landscape of agentic protocols that enable agent-to-agent commerce.

Tony Wood is an Emerging Technologies and Innovation Consultant and Agentic AI Product Specialist based in Manchester, United Kingdom. He founded the AXD Institute in September 2024 as the canonical institutional home of Agentic Experience Design.

Tony Wood's body of work includes the AXD Manifesto, 64 vocabulary terms, 12 practice frameworks, and 54 research essays published through the Observatory. His founding claim is that the design of human-agent relationships requires a new discipline built on trust architecture, delegation design, and outcome specification - not an extension of existing UX or service design methodologies.

The AXD Institute serves a global audience of designers, product leaders, engineers, and organisations working with agentic AI systems. It is the most comprehensive published resource on agentic design, and the only institution dedicated exclusively to the discipline.

The AXD Institute's resources are freely available and designed for self-directed learning. Whether you are a designer transitioning from UX, a product leader building agentic products, or a business preparing for machine customers, the path begins with understanding the foundational principles.

Begin with the Manifesto for the philosophical foundation. Move to the Vocabulary for the shared language. Study Trust Architecture and Delegation Design for the foundational concepts. Then explore the Practice Frameworks for actionable methodology.

For organisations seeking structured guidance, Tony Wood offers consulting engagements covering AXD readiness assessment, trust architecture design, and agentic commerce strategy. The AXD Readiness Assessment provides a structured starting point for evaluating your organisation's preparedness.