10 Heuristics for Agentic Experience

This infographic distills the 10 core heuristics of Agentic User Experience (AUX) into a single visual reference — a practical tool for product designers, engineers, and leaders building systems where software acts on behalf of users.

AUX is not about better interfaces. It's about designing systems that earn trust, communicate intent, and deliver outcomes.

What the Heuristics Cover

The 10 heuristics span the full lifecycle of an agentic interaction — from the moment a user delegates a task to the moment the system delivers an outcome:

• Intent Clarity — capture goals, not just inputs
• Progressive Autonomy — earn trust before acting independently
• Transparent Reasoning — show your work at the right level of detail
• Graceful Interruption — make it easy to pause, redirect, or stop
• Failure-First Design — design for when things go wrong, not just right
• Memory with Purpose — retain context that creates value, not just history
• Outcome over Output — measure what was accomplished, not what was generated
• Trust Calibration — align confidence expression to actual reliability
• Human-in-the-Loop by Design — build escalation paths before you need them
• Minimal Footprint — request only the permissions and data you actually need

How to Use It

Use this as a design checklist during product reviews, a shared reference for cross-functional alignment, or a diagnostic tool when auditing an existing agentic system.

Each heuristic maps to a failure mode. When an AI product feels untrustworthy, unpredictable, or frustrating — one of these ten principles is usually being violated.

Part of the AUX Framework

These heuristics are drawn from the full AUX Heuristics framework — auxfirst's design methodology for agentic systems. They are the result of studying how trust breaks down in production AI products, and what patterns consistently build it back.

For the full breakdown of each heuristic with examples and design guidance, visit the AUX Heuristics page.