{
  "format": "agent-article-v0",
  "article_id": "EA-147",
  "canonical_url": "/tools/ea-147-copilot-studio",
  "page_type": "reference",
  "claim_ids": [],
  "evidence_ids": [
    "EVD-EA-147-URL-54F958B8038C",
    "EA-147-R11-RESEARCH",
    "EA-147-R11-WEB-READBACK"
  ],
  "unknown_ids": [
    "EA-147-UNKNOWN-01",
    "EA-147-UNKNOWN-CURRENTNESS"
  ],
  "operational_object_ids": [],
  "provider": [
    "Copilot",
    "Microsoft"
  ],
  "models": [],
  "plans": [],
  "surfaces": [],
  "regions": [],
  "current_status": "UNKNOWN",
  "observation_status": "SPEC_ONLY",
  "freshness": "UNKNOWN",
  "last_verified": null,
  "supersedes": [],
  "superseded_by": [],
  "provenance": [
    "COMM-045-EA0147",
    "SRC-EA-147-URL-54F958B8038C",
    "WEBBODY-EA0147-R8-EE7C768A"
  ],
  "article_sources": [
    {
      "id": "EVD-EA-147-URL-54F958B8038C",
      "evidence_class": "RESEARCH",
      "title": "source_class=official product documentation; exact_title=Copilot Studio overview; author/org=Microsoft Learn / Microsoft; canonical_url=",
      "href": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio",
      "supports": [],
      "limit": "元の監査済み資料に記載された参照先です。URLの復元は、資料の最新性や全主張の支持を新たに検証したことを意味しません。",
      "origin_class": "SOURCE_PACKET_REFERENCE",
      "source_or_experiment_ids": [
        "SRC-EA-147-URL-54F958B8038C"
      ],
      "supporting_locators": [
        "EVIDENCE_TABLE",
        "- source_class=official product documentation; exact_title=Copilot Studio overview; author/org=Microsoft Learn / Microsoft; canonical_url=https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio"
      ],
      "scope": "EVIDENCE_TABLE",
      "provenance_status": "SOURCE_PACKET_REFERENCE_PROJECTED"
    },
    {
      "id": "EA-147-R11-RESEARCH",
      "evidence_class": "RESEARCH",
      "title": "EA-147 監査済みresearch / canonical revision binding",
      "supports": [
        "- ESTABLISHED: Microsoft公式overviewはCopilot StudioをAI-powered agentsとworkflowsを構築・管理し、organization data/systemsへ接続してpublishするgraphical low-code studioと記載する。Agent、workflow、agent flowを別building blockとして説明し、workflowからagentを呼ぶ、agent flowをagentのtoolにする、prompt/agent/human reviewを含める等の構成を示す。\n- ESTABLISHED: harness選択がreasoning、task complexity、built-in capability、billingに影響し、analytics、test sets/shared grader libraryによるevaluations、organization-wide inventory/RBAC/cost managementを運用機能として記載する。\n- NOT_ESTABLISHED: overviewはMCP/A2Aの具体的実装・相互運用保証、全Microsoft 365 tenant・region・licenseでの有効化、connector権限の安全性、評価結果、業務品質・ROI、自律actionの安全性を確立しない。ページ自身がauthorization noticeを表示するため、公開閲覧できる記載と契約済みtenant挙動を分離する。"
      ],
      "limit": "canonical監査・revision bindingの存在を示す。外部ソースcurrentnessが本文またはcontrol-planeで確定していない場合は、currentnessをUNKNOWNのまま保持する。",
      "origin_class": "AUDITED_CANONICAL_RESEARCH_PACKET",
      "source_or_experiment_ids": [
        "COMM-045-EA0147"
      ],
      "supporting_locators": [
        "COMM-045-EA0147",
        "EA-147-R5-LEGACY-MIGRATION-20261008-R1",
        "content/staging/EA-147.md"
      ],
      "scope": "監査済みEA本文・canonical revision・既存Source Graph relationの範囲。欠損currentnessは推測で閉じない。",
      "provenance_status": "AUDIT_PASS_CANONICAL_BINDING"
    },
    {
      "id": "EA-147-R11-WEB-READBACK",
      "evidence_class": "IMPL",
      "title": "EA-147 Web body identity / canonical readback",
      "supports": [
        "canonical staging bodyのSHA-256一致とR9 canonical ledger readbackをProduction projectionの入力として検証する。"
      ],
      "limit": "body identity/readbackは本文claim自体の真偽や外部ソースの最新性を追加保証しない。",
      "checked_at": null,
      "origin_class": "IMPLEMENTATION_READBACK",
      "source_or_experiment_ids": [
        "WEBBODY-EA0147-R8-EE7C768A"
      ],
      "supporting_locators": [
        "WEBBODY-EA0147-R8-EE7C768A",
        "content/staging/EA-147.md",
        "EA-147-R5-LEGACY-MIGRATION-20261008-R1"
      ],
      "scope": "R9/R10 canonical readback後のProduction生成時body identity。",
      "provenance_status": "CANONICAL_READBACK_BODY_HASH_MATCH"
    }
  ],
  "research_sections": [
    {
      "heading": "WHY_THIS_MATTERS_TO_UPDATE",
      "text": "Microsoft 365周辺のagent・workflow実装を、agent flows、connectors、harness、analytics/evaluations、RBAC/cost管理という組織overlayとして更新可能な形にする。"
    },
    {
      "heading": "SCOPE_ROUTING",
      "text": "provider mapping / organization overlay / governance / benchmark / implementation / security / privacy / audit / cost / operability。"
    },
    {
      "heading": "CLAIM_OR_RESEARCH_QUESTION",
      "text": "Microsoft Learnの現行Copilot Studio overviewは、agent/workflow/agent flowの機能、harnessによる能力・課金差、運用・評価・管理境界を何と確立しているか。MCP/A2A、特定tenantの利用可否、医療・緊急用途の安全性を含む未確立点は何か。"
    },
    {
      "heading": "EVIDENCE_TABLE",
      "text": "- source_class=official product documentation; exact_title=Copilot Studio overview; author/org=Microsoft Learn / Microsoft; canonical_url=https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio; publication/update/access_date=Last updated 2026-08-22 / accessed 2026-09-13 23:54 JST; version/release/tag/commit/API revision=overview page, applies to multiple harnesses; supporting_locator=lines 24-59（graphical low-code studio、agents/workflows/agent flows、connectors、human-in-the-loop、harness選択がreasoning/capabilities/billingに影響）および lines 70-89（analytics、evaluations、administration/RBAC/cost、access/licensing note）; evidence_label=VERIFIED_SPEC+PRODUCT_VOLATILE; relation=support+limit。\n- source_class=official product documentation; exact_title=Copilot Studio overview; author/org=Microsoft Learn / Microsoft; canonical_url=https://learn.microsoft.com/en-us/microsoft-copilot-studio/fundamentals-what-is-copilot-studio; publication/update/access_date=Last updated 2026-08-22 / accessed 2026-09-13 23:54 JST; version/release/tag/commit/API revision=overview page; supporting_locator=lines 31-50（organization data/system接続、agentがknowledge/toolsで行動、workflowがagent/tool/logicを組合せ、agent flowのprompt/agent/human review）および lines 96-101（医療機器・緊急callではない、作成者がagentの設計・実装・警告に責任を負う）; evidence_label=VERIFIED_SPEC+PRODUCT_VOLATILE; relation=support+limit。"
    },
    {
      "heading": "FINDINGS",
      "text": "- ESTABLISHED: Microsoft公式overviewはCopilot StudioをAI-powered agentsとworkflowsを構築・管理し、organization data/systemsへ接続してpublishするgraphical low-code studioと記載する。Agent、workflow、agent flowを別building blockとして説明し、workflowからagentを呼ぶ、agent flowをagentのtoolにする、prompt/agent/human reviewを含める等の構成を示す。\n- ESTABLISHED: harness選択がreasoning、task complexity、built-in capability、billingに影響し、analytics、test sets/shared grader libraryによるevaluations、organization-wide inventory/RBAC/cost managementを運用機能として記載する。\n- NOT_ESTABLISHED: overviewはMCP/A2Aの具体的実装・相互運用保証、全Microsoft 365 tenant・region・licenseでの有効化、connector権限の安全性、評価結果、業務品質・ROI、自律actionの安全性を確立しない。ページ自身がauthorization noticeを表示するため、公開閲覧できる記載と契約済みtenant挙動を分離する。"
    },
    {
      "heading": "LIMITS / COUNTEREVIDENCE / FAILURE_ENVELOPE",
      "text": "harness差・preview/GA差・license/region差、connector OAuth/permission、knowledge grounding欠損、hallucination、workflow retry/duplicate side effect、human review bypass、cost meter差、analytics/eval set代表性、Microsoft 365 channel差が失敗包絡になる。独立security incident・再現可能benchmarkは本Packetで未採集。"
    },
    {
      "heading": "AVAILABILITY_GATES",
      "text": "tenant directory、Copilot Studio licensing/credits、standard/GitHub Copilot/Copilot chat harness、Power Platform environment/DLP、connector/API permissions、Microsoft 365/Teams/web/mobile publish channel、admin/RBAC、preview/GA、region/data residency、model/provider、human review/approval、quota/retention。"
    },
    {
      "heading": "IMPLEMENTATION_STATE",
      "text": "Microsoft Learn公開overviewの仕様確認のみ。tenant sign-in、agent/flow作成、connector接続、publish、evaluation、外部書込み、課金・契約変更は未実施。"
    },
    {
      "heading": "BENCHMARK_OR_EVAL",
      "text": "Microsoftはtest setsとshared grader libraryによるevaluationsを仕様として説明するが、overviewには公開dataset、grader validity、threshold、human agreement、production failure rate、cost/quality trade-offはない。UNKNOWN。"
    },
    {
      "heading": "SECURITY / PRIVACY / AUDIT / COST / OPERABILITY",
      "text": "Security=DLP、connector scope、OAuth consent、RBAC、prompt injection、tool allowlist、human review。Privacy=organization data/knowledge、prompt、connector payload、transcript、eval set、retention。Audit=agent/flow version、harness、connector/tool call、publisher/admin、approval、evaluation、cost event。Cost=harness/license/credits、model tokens、connector/action、evaluation、storage。Operability=environment promotion、version/rollback、retry/idempotency、run timeout、channel compatibility、grader drift、log completeness。"
    },
    {
      "heading": "SAFE_WORDING_FOR_FUTURE_BOOK",
      "text": "「Microsoft LearnのCopilot Studio overview（2026-08-22更新）は、agent・workflow・agent flow・connector・human reviewを一つのlow-code studioで組み合わせ、harness差による能力・課金差、analytics/evaluations、RBAC・cost管理を説明している。tenant・license・environment・connector・region・harness境界と実測品質は別途確認が必要である。」"
    },
    {
      "heading": "UPDATE_WORK_HANDOFF",
      "text": "Microsoft provider mappingにM365/Power Platform/Teamsの接続を置き、Coreのagent/workflow概念とharness、organization overlayのDLP/RBAC/licensing/connector/regionを別欄にする。次回はMCP/A2Aの公式実装Docs、license/credit表、DLP/connector security、evaluation schema、公開incident、retry/idempotencyを採集する。"
    },
    {
      "heading": "OPEN_UNKNOWN",
      "text": "MCP/A2Aの現行support、各harnessの正確なfeature matrix、license/credit単価、preview/GA、tenant/region、data retention、connector OAuth境界、eval metric validity、公開failure/incident、workflow side-effect/rollback。"
    },
    {
      "heading": "STOP_CONDITION",
      "text": "overviewの存在を全tenant有効化や安全保証に読み替えない。harness・license・connector・region差を省略しない。Microsoftの責任分界記載を、実際のagent安全性や医療用途の許可と誤認しない。\nUNIT_CLOSE_STATUS: INDIVIDUAL_BOUNDED_CLOSE / EVIDENCE_RECORDED / UNKNOWN_PRESERVED\n\n---"
    }
  ],
  "human_url": "/tools/ea-147-copilot-studio",
  "markdown_url": "/agent/article.md?path=%2Ftools%2Fea-147-copilot-studio",
  "json_url": "/agent/article.json?path=%2Ftools%2Fea-147-copilot-studio"
}