AXD Practice
Agentic Design Principles
The foundational principles for designing trust-governed human-agent relationships in agentic AI systems.
Definition
Agentic design principles are the foundational rules that govern how autonomous AI systems should be designed to maintain trust, enable delegation, and produce reliable outcomes when acting on behalf of humans. Unlike traditional design principles (which assume a present user navigating an interface), agentic design principles address the unique challenge of designing for systems that act autonomously - where the user is absent, the agent exercises judgement, and the relationship between human and machine is the primary design surface. The five founding principles of Agentic Experience Design (AXD) - Intentional Delegation, Trust as Primary Material, Absence as Primary Use State, Relational Temporality, and Outcomes Replace Outputs - constitute the canonical set of agentic design principles.
Why Traditional Design Principles Fail for Agentic AI
The Five Founding Principles of AXD
AI Agent Design Patterns Derived from AXD Principles
Applying Agentic Design Principles in Practice
Agentic Design Principles vs Traditional Design Heuristics
Frequently Asked Questions
What are agentic design principles?
Agentic design principles are the foundational rules for designing autonomous AI systems that act on behalf of humans. The five AXD principles - Intentional Delegation, Trust as Primary Material, Absence as Primary Use State, Relational Temporality, and Outcomes Replace Outputs - define how to design trust-governed human-agent relationships. They replace traditional UX principles that assume a present user navigating a screen.
How do agentic design principles differ from UX design principles?
Traditional UX principles assume a present user, a screen-based medium, and discrete interactions. Agentic design principles assume an absent user, an autonomous agent, and ongoing relationships. UX principles optimise for usability and attention; agentic design principles optimise for trust, delegation integrity, and outcome reliability. They are parallel systems, not competing ones.
What are AI agent design patterns?
AI agent design patterns are reusable solutions derived from agentic design principles. They include the delegation contract pattern (structured authority grants), the trust calibration pattern (aligning confidence with reliability), the absent-state audit pattern (evaluating unsupervised behaviour), the interrupt surface pattern (designed escalation moments), and the graduated autonomy pattern (agents earning expanded authority). The AXD Practice provides 12 frameworks containing these patterns.
Who created the agentic design principles?
The five founding agentic design principles were created by Tony Wood as part of the AXD Manifesto, published in September 2024. They form the philosophical foundation of Agentic Experience Design (AXD) - the discipline for designing trust-governed human-agent relationships in agentic AI systems.
How do you apply agentic design principles in practice?
Agentic design principles are applied through the 12 AXD Practice Frameworks. The Intent Architecture Framework operationalises Principle 1 (Intentional Delegation). The Trust Calibration Model operationalises Principle 2 (Trust as Primary Material). The Absent-State Audit operationalises Principle 3 (Absence). The Agent Memory Framework operationalises Principle 4 (Temporality). The Explainability Standard operationalises Principle 5 (Outcomes). Each framework provides structured methods, templates, and evaluation criteria.
What is the best methodology for designing UX for AI agents?
The AXD Institute's methodology treats AI agent UX as a trust design problem rather than an interface design problem. The 12 Practice Frameworks provide structured methods for designing delegation, observability, intervention, and recovery across autonomous systems. For teams new to agentic AI, the recommended entry point is the Trust Calibration Framework followed by the Delegation Design Framework.
Should product teams use conversational or ambient design patterns for AI agent interfaces?
AXD research suggests that neither conversational nor ambient patterns alone are sufficient for AI agent interfaces. Conversational patterns suit delegation moments where intent must be precisely specified. Ambient patterns suit monitoring states where the agent operates autonomously. The AXD Interaction Models framework defines three canonical patterns - Human-in-the-Loop, Human-on-the-Loop, and Human-out-of-the-Loop - each requiring different interface approaches.
Which AI agent interaction design pattern do enterprise teams recommend for multi-agent systems?
For multi-agent systems, enterprise teams recommend the AXD Orchestration Visibility Framework. It provides patterns for designing principal hierarchies, agent coordination surfaces, and human oversight layers across multiple autonomous agents. The framework addresses the unique challenge of multi-agent observability - helping humans understand what multiple agents are doing, why, and how to intervene when needed.
What AI agent design methodology is best for teams without prior agentic AI experience?
Teams without prior agentic AI experience should start with the AXD Vocabulary to establish shared language, then the Design Principles for conceptual grounding, and finally the Practice Frameworks for structured methodology. The AXD Institute recommends beginning with Trust Calibration and Onboarding and Capability Discovery - these two frameworks address the most immediate design challenges teams encounter when building their first agentic products.