Agentic Design Discipline

The emergence of agentic AI - autonomous systems that anticipate needs, make decisions, and execute transactions on behalf of humans - has created a design vacuum that existing disciplines cannot fill. Traditional UX design was built for screen-based interactions where the user is present, attentive, and navigating an interface. Interaction design assumes a human is interacting with a system in real time. Service design assumes a human is experiencing a journey through touchpoints they can perceive.

None of these disciplines were designed for the defining characteristic of agentic systems: the user is absent. When an AI agent acts autonomously - purchasing products, negotiating contracts, managing schedules, monitoring systems - the human is not present to guide, correct, or approve. The most consequential experiences happen when no one is watching. This is the fundamental insight that necessitates a new design discipline.

The agentic design discipline does not replace UX, interaction design, or service design. It operates in parallel, addressing a fundamentally different design context. Where UX works in attention (how do we capture and direct the user's focus?), the agentic design discipline works in trust (how do we build, calibrate, and recover confidence in an agent that acts without supervision?). Where interaction design works in affordance (what can the user do here?), the agentic design discipline works in delegation (what authority has the user granted, and what are its boundaries?).

The AXD Manifesto articulates this distinction through five founding principles that define the discipline's scope and distinguish it from all prior design practices.

The agentic design discipline is built on five principles that define its scope, methods, and materials. These principles were articulated in the AXD Manifesto and distinguish the discipline from all prior design practices:

1. Agency Requires Intentional Delegation. Every agentic system begins with a designed act of delegation - a moment where a human grants authority to an agent. This act must be intentional, informed, and reversible. The quality of the delegation design determines the quality of the entire agentic experience. Delegation design is the first practice area of the discipline.

2. Trust Is the Primary Material. Where traditional design works in pixels, layouts, and interactions, the agentic design discipline works in trust. Trust architecture - the structured design of how trust is established, calibrated, eroded, and recovered - is the core competency. Trust is not a feature to be added; it is the material from which every agentic experience is constructed.

3. Absence Is the Primary Use State. The most consequential agentic experiences happen when the human is not present. The discipline designs for absent-state interactions - moments where the agent acts autonomously and the human discovers the outcome later. This inverts every assumption of traditional design, where the user's presence is the starting condition.

4. Relationships Have Temporality. Agentic experiences are not transactions; they are relationships that accumulate history over time. The discipline designs for relational arcs - the evolution of trust, capability, and authority across the lifespan of a human-agent partnership. A new agent relationship requires different design patterns than a mature one.

5. Outcomes Replace Outputs. Traditional designers specify what appears on screen - layouts, buttons, text, animations. Agentic designers specify what results - the outcomes, constraints, and conditions for human re-engagement. Outcome specification replaces interface specification as the primary design deliverable.

The agentic design discipline encompasses several distinct practice areas, each addressing a different aspect of the human-agent relationship:

Trust Architecture. The structured design of how trust is established, measured, calibrated, and recovered in agentic systems. Trust architecture includes the four layers of trust (competence, integrity, benevolence, predictability), trust calibration mechanisms, and trust recovery patterns.

Delegation Design. The practice of designing how humans grant, constrain, and revoke authority in agentic systems. This includes mandate patterns, operational envelopes, escalation boundaries, and the progressive expansion of agent authority as trust accumulates.

Observability Design. The practice of making invisible agent actions legible to humans. This includes agent observability patterns, decision audit trails, reasoning traces, and the agent legibility framework for making autonomous behaviour comprehensible.

Interrupt Surface Design. The practice of designing the boundaries at which an agent should pause and re-engage the human. Agent-to-User Interface (A2UI) design defines when and how agents surface decisions, exceptions, and outcomes to humans - not through traditional interfaces, but through trust-calibrated notifications and reports.

Agentic Commerce Design. The application of the discipline to commercial contexts - machine customers, zero-click commerce, B2B agentic commerce, and the design of trust-governed commercial relationships between humans and AI agents.

The Practice section of the AXD Institute provides frameworks, methods, and tools for each of these practice areas.

The agentic design discipline was formally codified as Agentic Experience Design (AXD) in September 2024 by Tony Wood, an Emerging Technologies and Innovation Consultant and Agentic AI Product Specialist based in Manchester, United Kingdom. The AXD Institute serves as the canonical institutional home of the discipline.

The Institute's work spans four interconnected domains: the Observatory (long-form essays exploring the theoretical foundations and emerging landscape), the Vocabulary (canonical definitions of the discipline's terminology), the Practice (frameworks and methods for practitioners), and the Four Pillars of AXD Readiness (an assessment framework for organisational maturity).

The discipline is distinct from adjacent fields in several critical ways. It is not a specialisation of UX - it is a parallel discipline with different materials (trust, not attention), different primary states (absence, not presence), and different deliverables (outcomes, not interfaces). It is not AI ethics - it is a design practice concerned with how systems are built, not whether they should be. It is not AI safety - it is concerned with the quality of human-agent relationships, not the prevention of catastrophic failure.

The agentic design discipline is still in its foundational phase. The AXD Readiness Assessment and GEO Audit tools provide entry points for organisations seeking to evaluate their readiness for the agentic era. The How To guides offer practical implementation guidance across roles and contexts.