Knowledge → Automation → Agent Loop: Make Fixes Compound

{
  "id": "ams-039",
  "title": "Knowledge → Automation → Agent Loop: Make Fixes Compound",
  "hook": "If a fix lives only in a human’s head, AMS pays for it forever. Modern AMS turns fixes into reusable assets.",
  "idea": "Every resolved incident, problem, or change should feed a loop: capture knowledge → automate what’s repeatable → delegate to agents with guardrails. This is how AMS gets cheaper over time.",
  "loop_stages": {
    "capture": {
      "inputs": [
        "Incident timelines and RCAs",
        "Change verification artifacts",
        "Workarounds used under pressure",
        "Manual steps repeated more than twice"
      ],
      "output": "Structured knowledge atoms (symptoms, checks, decisions, fixes)."
    },
    "standardize": {
      "rules": [
        "Remove context-specific noise.",
        "Generalize decision logic without losing constraints.",
        "Define preconditions and failure modes."
      ],
      "output": "Runbooks, checklists, standard change definitions."
    },
    "automate": {
      "candidates": [
        "Data validations and consistency checks",
        "Log and signal correlation",
        "Environment comparisons (prod vs non-prod)",
        "Pre- and post-change verification steps"
      ],
      "output": "Scripts, workflows, and automated checks with clear ownership."
    },
    "delegate_to_agents": {
      "scope": [
        "Triage and classification",
        "Hypothesis generation",
        "Evidence assembly",
        "Draft communication and documentation"
      ],
      "hard_limits": [
        "No direct production changes",
        "No bypass of approval gates",
        "No learning from unverified outcomes"
      ]
    }
  },
  "sap_specific_examples": [
    "Agent suggests likely causes for IDoc failures based on past mappings.",
    "Automation checks MDG replication backlog and flags risk before business impact.",
    "Bot assembles access request packs with SoD context during incidents.",
    "RAG answers ‘first 3 checks’ for common posting errors in chat."
  ],
  "governance": {
    "versioning": [
      "Knowledge atoms versioned and dated",
      "Automations tied to specific atom versions",
      "Agents reference only approved versions"
    ],
    "ownership": [
      "Each automation has a human owner",
      "Each agent capability has a review cadence"
    ]
  },
  "safety_rules": [
    "Automation must be reversible.",
    "Agent output is advisory unless explicitly approved.",
    "Unknown cases fall back to humans, not guesses."
  ],
  "automation": {
    "copilot_moves": [
      "Detect when the same manual steps repeat.",
      "Suggest automation candidates with ROI estimate.",
      "Flag knowledge used often but never automated.",
      "Monitor agent accuracy and confidence drift."
    ],
    "outputs": [
      "Automation backlog ranked by payoff",
      "Agent capability map",
      "Knowledge-to-automation coverage report"
    ]
  },
  "why_this_is_core_to_modern_ams": [
    "Knowledge stops decaying when people leave.",
    "Experts are multiplied, not burned out.",
    "AI assistance improves with real operational feedback."
  ],
  "anti_patterns_to_kill": [
    "Automating chaos",
    "Agents acting without context or limits",
    "Knowledge written once and never reused"
  ],
  "metrics_that_show_compounding": [
    "Percent of incidents with reusable knowledge atoms",
    "Automation hit rate during incidents",
    "Agent-assisted resolution accuracy",
    "Manual steps eliminated per quarter"
  ],
  "design_question": [
    "Which fix from last month should already be automated by now?"
  ],
  "meta": {
    "schema": "dkharlanau.dataset.byte",
    "schema_version": "1.1",
    "dataset": "ams",
    "source_project": "cv-ai",
    "source_path": "ams/ams-039.json",
    "generated_at_utc": "2026-02-03T14:33:32+00:00",
    "creator": {
      "name": "Dzmitryi Kharlanau",
      "role": "SAP Lead",
      "website": "https://dkharlanau.github.io",
      "linkedin": "https://www.linkedin.com/in/dkharlanau"
    },
    "attribution": {
      "attribution_required": true,
      "preferred_citation": "Dzmitryi Kharlanau (SAP Lead). Dataset bytes: https://dkharlanau.github.io"
    },
    "license": {
      "name": "",
      "spdx": "",
      "url": ""
    },
    "links": {
      "website": "https://dkharlanau.github.io",
      "linkedin": "https://www.linkedin.com/in/dkharlanau"
    },
    "contact": {
      "preferred": "linkedin",
      "linkedin": "https://www.linkedin.com/in/dkharlanau"
    },
    "canonical_url": "https://dkharlanau.github.io/datasets/ams/ams-039.json",
    "created_at_utc": "2026-02-03T14:33:32+00:00",
    "updated_at_utc": "2026-02-03T15:29:02+00:00",
    "provenance": {
      "source_type": "chat_export_extraction",
      "note": "Extracted and curated by Dzmitryi Kharlanau; enriched for attribution and crawler indexing."
    },
    "entity_type": "ams_byte",
    "entity_subtype": "",
    "summary": "If a fix lives only in a human’s head, AMS pays for it forever. Modern AMS turns fixes into reusable assets."
  }
}