Human-Centred Agentic Intelligence (HAI): A Three-Layer Framework for Designing Human-Agent-System Journeys

As artificial intelligence becomes embedded in both product experiences and customer lifecycles, organisations face a dual design challenge: how to design discrete task interactions where agent behaviour must be precise, transparent, and failure-resilient; and how to design across a customer's full lifecycle where agent behaviour must adapt to evolving trust, context, and intent over time. This edition extends the framework with a Sovereign AI dimension, addressing a third design imperative: how to specify data residency, compute location, model provenance, and cross-border constraints at design time — before engineering begins. The Human-Centred Agentic Intelligence (HAI) framework unifies these perspectives into a single, coherent Dual-Mode Matrix — the User Journey Matrix (Step-Based) for product and task design, and the Customer Journey Matrix (Stage-Based) for lifecycle and CX design — both grounded in the same three layers: what the human experiences (Human Layer), how the agent reasons and acts (Agent Layer), and what data and systems make agent behaviour reliable (System Layer). The System Layer now includes a Sovereignty Boundary dimension per step and stage, capturing data residency, compute location, model provenance, and cross-border data flow constraints alongside protocol, memory, and regulatory specifications. The framework introduces four original contributions: (1) the HAI concept, which fuses human-centred AI design principles with agentic AI applied to journey design; (2) the Unified Three-Layer Journey Matrix as a structured design tool; (3) the Dual-Mode framework architecture accommodating both step-based product design and stage-based lifecycle design; and (4) the Agent Mode taxonomy (Assistive / Advisory / Autonomous) as applied to journey-phase design. Complemented by a Goal Failure Response Protocol with a seven-type failure taxonomy, a Logical Handover Framework, Multi-Agent Role Taxonomy, Measurement Framework, Regulatory and Sovereignty compliance integration — EU AI Act risk classification and Sovereignty Boundary specification per step and stage — and a 28-template library spanning SaaS, fintech, healthcare, legal, B2B, and more, HAI provides a complete design language for human-agent-system journeys — from the precise trigger condition of a single agent action to the long-arc trust evolution of a customer across years of engagement.