The AXD Discipline - Four Practitioner Roles
Who Designs the Human-Agent Relationship?
Every agentic system requires four distinct design competencies. These are not job titles borrowed from UX. They are new roles for a new discipline - each responsible for a different structural layer of the human-agent relationship.
Why Four Roles
Traditional UX design assumes a single designer can own the entire user experience. That assumption breaks down in agentic systems. When an AI agent acts autonomously on behalf of a human - making decisions, executing transactions, operating while the human is absent - the design surface expands beyond what any single role can govern.
The trust architecture that governs whether a human can delegate with confidence is a different design challenge from the delegation design that determines what the agent is permitted to do. The observability system that makes agent behaviour legible is a different design challenge from the re-engagement patterns that bring humans back into the loop at the right moment.
These four domains require four distinct competencies. The AXD Institute identifies them as the four practitioner roles of Agentic Experience Design. Each role owns a different structural layer of the human-agent relationship. Together, they cover the complete design surface of autonomous systems.
Trust Architect
Owns: Trust Architecture
The structural foundation of every agentic system
The Trust Architect owns the design of how trust is established, maintained, calibrated, and recovered across the entire human-agent relationship. This is the foundational role in AXD - without trust architecture, every other design decision is built on sand. The Trust Architect does not design interfaces. They design the structural system that determines whether a human can delegate authority to an agent with justified confidence.
Core Design Question
"Can the human delegate with justified confidence?"
Key Responsibilities
- 01Design the four-layer trust model: competence trust, integrity trust, benevolence trust, and predictability trust
- 02Define trust formation mechanisms - how initial trust is established between human and agent
- 03Design trust calibration systems - how trust levels adjust based on evidence and performance
- 04Architect trust recovery procedures - how trust is rebuilt after violations or failures
- 05Establish trust metrics and measurement frameworks across the agent lifecycle
- 06Audit existing systems for trust debt - the accumulated cost of deferred trust design decisions
Essential Reading
Observatory Essays
Practice Frameworks
Key Vocabulary
Delegation Designer
Owns: Scope and Authority
The boundary between human intention and machine action
The Delegation Designer owns the structures through which humans grant, constrain, monitor, and revoke authority given to autonomous agents. This role determines what an agent is permitted to do, for how long, under what conditions, and what triggers human re-engagement. Delegation design is the most consequential design surface in agentic systems - it is the moment a human decides to hand over control. That decision must be designed, not assumed.
Core Design Question
"What is the agent permitted to do, and what must it not?"
Key Responsibilities
- 01Design delegation scope - what the agent is authorised to do and what remains human-only
- 02Define authority boundaries - spending limits, decision categories, temporal constraints
- 03Architect escalation conditions - when the agent must pause and seek human input
- 04Design revocation mechanisms - how humans withdraw delegated authority cleanly
- 05Create intent specification frameworks - how human goals are translated into agent constraints
- 06Audit delegation patterns for scope creep, authority drift, and constraint erosion
Essential Reading
Observatory Essays
Practice Frameworks
Key Vocabulary
Observability Lead
Owns: Agent Legibility
Making the invisible visible without demanding attention
The Observability Lead owns agent legibility - the design of how autonomous systems communicate their state, reasoning, actions, and confidence to human principals. This is not logging. It is not dashboards. It is the design of how an agent that operates while the human is absent makes its behaviour readable when the human returns or needs to intervene. The Observability Lead works at the intersection of transparency and trust, ensuring that agents are legible without being intrusive.
Core Design Question
"Can the human understand what the agent did and why?"
Key Responsibilities
- 01Design agent state communication - how the agent reports what it is doing and why
- 02Architect confidence signalling - how the agent communicates certainty and uncertainty
- 03Create action audit trails - legible records of what the agent did during absence
- 04Design signal clarity frameworks - ensuring agent communications are interpretable, not just available
- 05Define observability standards across multi-agent systems and agent-to-agent interactions
- 06Audit systems for observability gaps - where agent behaviour is opaque to human principals
Essential Reading
Observatory Essays
Practice Frameworks
Key Vocabulary
Re-engagement Designer
Owns: Human Return Paths
The design of how humans come back into the loop
The Re-engagement Designer owns the moments when humans return to systems that have been operating autonomously. This is the most underestimated role in agentic design. Every agentic system has a return state - the moment the human re-enters the relationship after a period of absence. That return must be designed, not defaulted. The Re-engagement Designer ensures that when a human comes back, they find a coherent narrative of what happened, clear options for what to do next, and a calibrated sense of whether their trust was warranted.
Core Design Question
"When the human returns, do they find clarity or chaos?"
Key Responsibilities
- 01Design return-state experiences - what the human sees when they come back after absence
- 02Architect interrupt patterns - when and how the agent requests human attention
- 03Create recovery flows - how the system helps humans regain control after agent failures
- 04Design re-engagement triggers - the conditions that bring humans back into the loop
- 05Define interrupt frequency calibration - balancing agent autonomy with human awareness
- 06Audit systems for return-state neglect - where humans come back to confusion rather than clarity
Essential Reading
Observatory Essays
Practice Frameworks
Key Vocabulary
The Interaction Model
Four Roles, One Relationship
The four AXD roles are not independent silos. They form a continuous design surface where each role's output becomes another role's input. The Trust Architect establishes the structural foundation - the trust model that determines what level of delegation is appropriate. The Delegation Designer takes that trust context and designs the specific authority boundaries, constraints, and escalation conditions. The Observability Lead ensures that everything the agent does within those boundaries is legible to the human principal. And the Re-engagement Designer owns what happens when the loop closes - when the human returns, when trust needs recalibration, when the delegation scope needs adjustment.
Consider an agentic shopping agent that manages a household's grocery purchases. The Trust Architect designs the trust formation process - how the agent earns the right to make purchasing decisions. The Delegation Designer defines the spending limits, brand preferences, substitution rules, and the conditions that require human approval. The Observability Lead designs how the agent communicates what it bought and why - not as a transaction log, but as a legible narrative. The Re-engagement Designer owns the return experience - what the human sees when they open the app after a week of autonomous purchasing.
If any one of these roles is absent, the system fails. Without trust architecture, the human never delegates. Without delegation design, the agent has no principled boundaries. Without observability, the human cannot assess whether their trust was warranted. Without re-engagement design, the human returns to confusion rather than clarity. The four roles are the minimum viable design team for any agentic system.
Trust context determines delegation scope
Authority boundaries define what must be observable
Legibility data shapes the return experience
Return outcomes feed trust recalibration
Career Pathways
Where These Roles Come From
The four AXD practitioner roles draw from existing professional disciplines but require new competencies that none of those disciplines currently teach. The Trust Architect may come from security architecture, risk management, or systems design - but must learn to treat trust as a designed material rather than a compliance requirement. The Delegation Designer may come from product management, policy design, or governance - but must learn to design authority structures for autonomous systems rather than human workflows.
The Observability Lead may come from DevOps, data engineering, or UX research - but must learn to design legibility for human principals rather than engineering dashboards. The Re-engagement Designer may come from UX design, service design, or customer experience - but must learn to design for the return state rather than the active-use state.
The AXD Academy provides structured learning pathways that build these competencies from first principles. The founding cohort programme offers the first professional credential in Agentic Experience Design - preparing practitioners for roles that do not yet exist in most organisations but will be essential within the next three years.
Frequently Asked Questions
About the AXD Practitioner Roles
What are the four AXD practitioner roles?
The four AXD practitioner roles are Trust Architect (owns trust architecture across the system), Delegation Designer (owns scope, authority, and constraint design), Observability Lead (owns agent legibility and transparency), and Re-engagement Designer (owns human return paths and recovery experiences). Together, these four roles cover the complete surface area of Agentic Experience Design.
Are these roles full-time positions or responsibilities within existing teams?
In most organisations today, these are responsibilities distributed across existing design, product, and engineering teams. As agentic systems mature, they will increasingly become dedicated roles. The AXD Institute recommends that any organisation deploying autonomous agents should have at least one person explicitly accountable for each domain - even if they hold other responsibilities alongside it.
How do the four AXD roles relate to traditional UX roles?
The four AXD roles are parallel to, not replacements for, traditional UX roles. A UX designer works in attention, affordance, and screen-based interaction. AXD practitioners work in trust, delegation, observability, and recovery - the design materials of autonomous systems. Some UX professionals will transition into AXD roles. Others will continue in UX while collaborating with AXD practitioners on the agentic layer of their products.
Which role should an organisation hire first?
The Trust Architect. Trust architecture is the foundational layer of every agentic system. Without it, delegation has no structural basis, observability has no trust context, and re-engagement has no recovery framework. The Trust Architect establishes the structural foundation on which the other three roles build.
Where can I train for these roles?
The AXD Academy offers structured learning pathways that cover the knowledge domains of all four practitioner roles. The founding cohort programme provides the first professional credential in Agentic Experience Design - the AXD Practitioner Certification.