{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreif5xnvdq7jslhs5wasfaczc6nn3fzeshyxo4s4fuo3pg2yeblhel4",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mjrkewhea2l2"
  },
  "path": "/t/human-centred-agentic-intelligence-hai-a-three-layer-framework-for-designing-human-agent-system-journeys/175362#post_1",
  "publishedAt": "2026-04-18T12:56:27.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "Human-Centred Agentic Intelligence (HAI): A Three-Layer Framework for Designing Human-Agent-System Journeys"
  ],
  "textContent": "Hi all\n\nI’m Anandakumar, an independent researcher from London. I’ve just published\nan open-access whitepaper on a design framework for agentic AI systems.\n\nFull paper (free PDF): Human-Centred Agentic Intelligence (HAI): A Three-Layer Framework for Designing Human-Agent-System Journeys\n\n* * *\n\n**THE PROBLEM**\n\nTeams building agentic AI face two different design challenges — and most\nframeworks conflate them:\n\n• Task-level: How does the agent behave at each STEP of a product interaction?\n• Lifecycle-level: How does agent behaviour evolve across a customer’s full\nRELATIONSHIP (Awareness → Purchase → Retention → Advocacy)?\n\nThese need different vocabularies — but share the same underlying infrastructure.\n\n* * *\n\n**THE HAI FRAMEWORK**\n\nA Dual-Mode, Three-Layer design matrix:\n\nThree shared layers:\n→ Human Layer — what the person sees, does, and feels\n→ Agent Layer — how the agent reasons, acts, and handles failure\n→ System Layer — data, infra, protocols, and governance\n\nTwo modes:\n→ User Journey Matrix (Step-Based) — for product/task design\n→ Customer Journey Matrix (Stage-Based) — for lifecycle/CX design\n\nAgent Mode Taxonomy per phase:\n→ Assistive (human controls) / Advisory (agent suggests) / Autonomous (agent acts)\nAutonomy should increase with trust — not be set globally.\n\n* * *\n\n**ALSO INCLUDES**\n\n• 7-type Goal Failure Response Protocol with mode-specific recovery paths\n• Logical Handover Framework (5-state model + 9-element context package)\n• Multi-Agent Role Taxonomy (Orchestrator / Specialist / Verifier / Liaison)\n• MCP + A2A Protocol Layer integration per step/stage\n• EU AI Act risk classification at step and stage level\n• Excel template with dropdowns — free download in the Zenodo record\n\n* * *\n\n**QUESTIONS FOR THE COMMUNITY**\n\n  1. Does the step-based vs. stage-based distinction match challenges you’ve faced?\n  2. Any gaps in the Assistive/Advisory/Autonomous taxonomy from your experience?\n  3. How are your teams handling EU AI Act compliance at the step/stage level?\n\n\n\nHappy to discuss — especially on failure handling and Agent Mode design.\n\nHuman-Centred Agentic Intelligence (HAI): A Three-Layer Framework for Designing Human-Agent-System Journeys\n\nLicense: CC BY-NC-ND 4.0 | Anandakumar Muniasamy Pothiraj, London",
  "title": "Human-Centred Agentic Intelligence (HAI): A Three-Layer Framework for Designing Human-Agent-System Journeys"
}