<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/"><channel><title>文章 · ObjectOS</title><link>https://www.objectos.ai/zh-Hant/blog/</link><atom:link href="https://www.objectos.ai/zh-Hant/rss.xml" rel="self" type="application/rss+xml" /><description>44 篇關於構建 AI-native 軟體的文章。</description><language>zh-Hant</language><lastBuildDate>Wed, 24 Jun 2026 00:00:00 GMT</lastBuildDate><generator>Astro</generator><item><title>AI 智慧體刪除生產資料：為什麼執行時護欄比提示詞可靠</title><link>https://www.objectos.ai/zh-Hant/blog/when-ai-agent-deletes-production-database/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/when-ai-agent-deletes-production-database/</guid><pubDate>Wed, 24 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>公開記錄中的 Replit 事故提醒我們：智慧體的影響半徑不能只靠提示詞收窄。生產資料、破壞性操作和恢復證據，都需要由執行時許可權、審批和審計來約束。</description><category>AI 與智慧體</category><category>IT / CIO</category><category>Replit</category><category>AI 智慧體</category><category>執行時護欄</category><category>治理</category><category>審計</category></item><item><title>Retool 與受治理 AI 應用平臺：能否審查業務許可權</title><link>https://www.objectos.ai/zh-Hant/blog/retool-vs-ai-native-app-platform/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/retool-vs-ai-native-app-platform/</guid><pubDate>Wed, 24 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>Retool 的 RBAC、審計日誌、SSO 和自託管能力很強。真正要比較的是業務許可權層：退款、審批、寫入動作如果散落在 JavaScript 繫結裡，AI 改動就很難被業務負責人用 diff 審查。</description><category>應用搭建</category><category>開發者</category><category>Retool</category><category>低程式碼</category><category>內部工具</category><category>治理</category><category>AI 原生</category></item><item><title>Power Platform 鎖定：Dataverse、Azure 與自託管取捨</title><link>https://www.objectos.ai/zh-Hant/blog/power-platform-lock-in-dataverse-self-host/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/power-platform-lock-in-dataverse-self-host/</guid><pubDate>Wed, 24 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>Power Platform 的優勢是真實的：身份、Teams、Dynamics 和賬單都在同一個租戶裡。但若你關心主權、規模化成本或 AI 改動審查，Dataverse 和 Azure 執行時就是必須提前算清的邊界。</description><category>安全與治理</category><category>IT / CIO</category><category>Power Platform</category><category>Power Apps</category><category>Dataverse</category><category>自託管</category><category>資料主權</category></item><item><title>Lovable 上生產安全嗎：訪問控制審查問題</title><link>https://www.objectos.ai/zh-Hant/blog/is-lovable-safe-for-production/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/is-lovable-safe-for-production/</guid><pubDate>Wed, 24 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>Lovable 很適合快速原型，但生產系統的問題不是能不能跑，而是誰能審查訪問控制。RLS、前端過濾和安全掃描都重要；真正的邊界必須在服務端可見、可強制、可簽字。</description><category>安全與治理</category><category>開發者</category><category>Lovable</category><category>Vibe Coding</category><category>安全</category><category>訪問控制</category><category>AI-Native</category></item><item><title>Airtable Omni 與受治理 AI 應用平臺：為什麼審閱 diff 比撤銷更重要</title><link>https://www.objectos.ai/zh-Hant/blog/airtable-omni-vs-governed-ai-app-platform/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/airtable-omni-vs-governed-ai-app-platform/</guid><pubDate>Wed, 24 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>Airtable Omni 把自然語言和表格式應用結合得很強。記錄系統真正要比較的，不是能不能生成應用，而是 AI 改動許可權、欄位和流程時，是否能在上線前形成可審閱的 diff。</description><category>應用搭建</category><category>業務決策者</category><category>Airtable</category><category>Airtable Omni</category><category>AI Builder</category><category>治理</category><category>審計</category></item><item><title>AI 寫完應用之後：你敢審查 diff 並點 Merge 嗎？</title><link>https://www.objectos.ai/zh-Hant/blog/ai-wrote-your-app-dare-to-merge/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-wrote-your-app-dare-to-merge/</guid><pubDate>Tue, 16 Jun 2026 10:30:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 可以很快生成能跑的應用，CI 也可能全綠。真正的問題是：那份幾千行、沒人完整理解的 PR，誰敢負責合併？當寫程式碼被自動化，瓶頸就從“寫”轉向“審查與簽字”。</description><category>應用搭建</category><category>開發者</category><category>AI 寫程式碼</category><category>程式碼審查</category><category>後設資料</category><category>治理</category><category>Vibe Coding</category><category>趨勢觀點</category></item><item><title>Agent 規則檔案怎麼寫：讓 AI 生成可治理應用</title><link>https://www.objectos.ai/zh-Hant/blog/give-your-agent-rules-for-governable-apps/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/give-your-agent-rules-for-governable-apps/</guid><pubDate>Tue, 16 Jun 2026 10:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>編碼 agent 的 AGENTS.md、.cursor/rules 或 CLAUDE.md 不該只管程式碼風格。把許可權、審批、審計和目標後設資料格式寫進去，AI 生成的應用才更容易被審查和簽字。</description><category>AI 與智慧體</category><category>開發者</category><category>Agent 規則</category><category>AI 寫程式碼</category><category>MCP</category><category>開放協議</category><category>治理</category><category>趨勢觀點</category></item><item><title>開放與封閉企業本體：誰擁有業務語義層</title><link>https://www.objectos.ai/zh-Hant/blog/enterprise-ontology-race-open-vs-closed/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/enterprise-ontology-race-open-vs-closed/</guid><pubDate>Tue, 16 Jun 2026 09:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>企業 AI 需要機器可讀的業務定義層。但如果“客戶、訂單、裝置”的定義分別鎖在不同平臺裡，agent 看到的就不是同一家公司。語義層應當歸企業自己，而不是歸某個套件。</description><category>安全與治理</category><category>業務決策者</category><category>企業本體</category><category>語義層</category><category>Palantir</category><category>微軟 Fabric</category><category>開放協議</category><category>趨勢觀點</category></item><item><title>Vibe Coding 技術債：為什麼 AI 生成的應用後來難改</title><link>https://www.objectos.ai/zh-Hant/blog/vibe-coding-technical-debt-2026/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/vibe-coding-technical-debt-2026/</guid><pubDate>Tue, 16 Jun 2026 08:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 生成程式碼能讓原型很快上線，但長期系統的問題在半年後出現：沒人完整讀過那一萬多行實現，也沒人能解釋當初的業務取捨。企業應用需要生成可審查的定義，而不是難維護的黑箱程式碼。</description><category>應用搭建</category><category>IT / CIO</category><category>Vibe Coding</category><category>技術債</category><category>程式碼生成</category><category>後設資料</category><category>AI 治理</category><category>趨勢觀點</category></item><item><title>EU AI Act 審計準備：你的 AI 執行時能交出證據嗎</title><link>https://www.objectos.ai/zh-Hant/blog/eu-ai-act-runtime-audit/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/eu-ai-act-runtime-audit/</guid><pubDate>Tue, 16 Jun 2026 07:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI Act 的多數規則按官方時間線將在 2026 年 8 月 2 日適用。審計真正要看的不是模型多強，而是執行時能否交出誰授權、動了什麼、證據在哪。</description><category>安全與治理</category><category>IT / CIO</category><category>EU AI Act</category><category>CADA</category><category>合規</category><category>資料主權</category><category>自託管</category><category>AI 治理</category></item><item><title>MCP 安全：為什麼協議還需要受治理的工具層</title><link>https://www.objectos.ai/zh-Hant/blog/mcp-governed-tool-layer/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/mcp-governed-tool-layer/</guid><pubDate>Tue, 16 Jun 2026 06:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>MCP 和 A2A 讓 agent 連線工具與其他 agent 變得更容易，但連線不等於授權。企業缺的不是再包一層介面，而是每次呼叫都帶身份、強制許可權、留下審計的工具層。</description><category>AI 與智慧體</category><category>IT / CIO</category><category>MCP</category><category>A2A</category><category>Agent 互操作</category><category>工具層</category><category>AI 治理</category><category>趨勢觀點</category></item><item><title>為什麼 AI Agent 試點進不了生產：缺的是四層執行基礎</title><link>https://www.objectos.ai/zh-Hant/blog/why-ai-agent-pilots-fail-four-layers/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/why-ai-agent-pilots-fail-four-layers/</guid><pubDate>Tue, 16 Jun 2026 05:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>一個 agent 演示可以很精彩，生產評審卻只問一件事：你怎麼證明它不會越權、會等審批、能交出審計證據？試點失敗通常不是模型不夠強，而是缺語義、許可權、審批和審計四層。</description><category>AI 與智慧體</category><category>業務決策者</category><category>AI Agent</category><category>落地</category><category>ROI</category><category>治理</category><category>執行時</category><category>趨勢觀點</category></item><item><title>AI Agent 定價：按動作計費與自託管執行時成本</title><link>https://www.objectos.ai/zh-Hant/blog/enterprise-agent-true-cost/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/enterprise-agent-true-cost/</guid><pubDate>Tue, 16 Jun 2026 04:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>按動作或按 token 計費看起來靈活，但 agent 越自主，中間工具呼叫越多。真正的成本賬要同時看用量曲線、資料出域、平臺鎖定，以及什麼時候自託管才划算。</description><category>安全與治理</category><category>業務決策者</category><category>成本</category><category>ROI</category><category>按動作計費</category><category>自託管</category><category>Agentforce</category><category>趨勢觀點</category></item><item><title>Agentforce、Copilot Studio 之外：何時選擇開放自託管執行時</title><link>https://www.objectos.ai/zh-Hant/blog/beyond-agentforce-copilot-open-runtime/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/beyond-agentforce-copilot-open-runtime/</guid><pubDate>Tue, 16 Jun 2026 03:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>Agentforce、Copilot Studio、ServiceNow AI Agents 在各自生態內都很強。問題是你的資料、流程和許可權是否也在同一個生態內；若不是，開放自託管執行時或混合架構可能更穩。</description><category>AI 與智慧體</category><category>IT / CIO</category><category>Agentforce</category><category>Copilot Studio</category><category>ServiceNow</category><category>開放平臺</category><category>自託管</category><category>趨勢觀點</category></item><item><title>企業 AI Ontology：為什麼業務語義層應該是開放協議</title><link>https://www.objectos.ai/zh-Hant/blog/ai-ontology-open-protocol/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-ontology-open-protocol/</guid><pubDate>Fri, 12 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 進入企業需要受治理的業務語義層，這個判斷已經越來越清楚。真正要重新思考的是形態：業務定義應當開放、可審查、可遷移；執行時可以收費並承擔執行責任。</description><category>安全與治理</category><category>業務決策者</category><category>Ontology</category><category>Palantir</category><category>AI 治理</category><category>開放協議</category><category>趨勢觀點</category></item><item><title>跨系統自動化：聯結器和 Webhook 如何進入受治理流程</title><link>https://www.objectos.ai/zh-Hant/blog/automation-cross-system-flows/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/automation-cross-system-flows/</guid><pubDate>Fri, 05 Jun 2026 04:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>CRM、ERP、合同、財務和外部服務都在流程裡。可靠的自動化引擎不是多接幾個 API，而是把外部呼叫、失敗補救和審計放進同一條業務流程。</description><category>整合與資料</category><category>業務決策者</category><category>ObjectOS</category><category>自動化引擎</category><category>聯結器</category><category>Webhook</category></item><item><title>自然語言改流程：自動化引擎如何治理變更</title><link>https://www.objectos.ai/zh-Hant/blog/automation-governed-flow-changes/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/automation-governed-flow-changes/</guid><pubDate>Fri, 05 Jun 2026 03:50:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>自然語言可以降低流程修改門檻，但不能讓流程直接上線。可靠的自動化引擎要把每次變更變成可審查、可校驗、可釋出、可回滾的流程後設資料。</description><category>應用搭建</category><category>業務決策者</category><category>ObjectOS</category><category>自動化引擎</category><category>流程治理</category><category>自然語言搭建</category></item><item><title>自動化觸發模型：資料變化、定時、按鈕和 API 如何進入同一流程</title><link>https://www.objectos.ai/zh-Hant/blog/automation-trigger-model/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/automation-trigger-model/</guid><pubDate>Fri, 05 Jun 2026 03:40:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>流程可以由記錄變化、定時計劃、人工按鈕、Webhook 或 API 呼叫觸發。成熟的自動化引擎要統一入口後的業務規則，避免同一流程在多個地方重複實現。</description><category>應用搭建</category><category>業務決策者</category><category>ObjectOS</category><category>自動化引擎</category><category>觸發器</category><category>業務流程</category></item><item><title>審批與暫停恢復：自動化引擎為什麼必須會等待</title><link>https://www.objectos.ai/zh-Hant/blog/automation-pause-resume-approvals/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/automation-pause-resume-approvals/</guid><pubDate>Fri, 05 Jun 2026 03:30:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>企業流程經常要等人審批、等使用者補材料、等時間到達或等外部結果。自動化引擎只有支援暫停和恢復，才能把這些等待放進同一條完整流程。</description><category>應用搭建</category><category>業務決策者</category><category>ObjectOS</category><category>自動化引擎</category><category>審批</category><category>暫停恢復</category></item><item><title>AI 自動化流程：如何判斷、等待並留下證據</title><link>https://www.objectos.ai/zh-Hant/blog/objectos-automation-engine/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/objectos-automation-engine/</guid><pubDate>Fri, 05 Jun 2026 03:20:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>企業需要的自動化不是簡單觸發器，而是能把業務規則變成可審批、可等待、可恢復、可審計的流程後設資料。ObjectOS 讓 AI 生成的流程進入業務執行時。</description><category>應用搭建</category><category>業務決策者</category><category>ObjectOS</category><category>AI 自動化</category><category>業務流程</category><category>後設資料驅動</category></item><item><title>AI 如何觸發業務動作：Action 後設資料如何保證受控執行</title><link>https://www.objectos.ai/zh-Hant/blog/objectos-action-tools/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/objectos-action-tools/</guid><pubDate>Fri, 05 Jun 2026 03:10:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI agent 的價值不只是回答問題，而是幫助使用者更新記錄、建立任務、發起審批和推動流程。ObjectOS 用 Action 後設資料把按鈕、流程和審批開放給 AI，同時保留許可權、確認和審計。</description><category>應用搭建</category><category>業務決策者</category><category>ObjectOS</category><category>AI Agent</category><category>業務動作</category><category>對話式應用</category></item><item><title>AI Agent 許可權邊界：資料和動作如何受執行時約束</title><link>https://www.objectos.ai/zh-Hant/blog/objectos-agent-permission-boundaries/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/objectos-agent-permission-boundaries/</guid><pubDate>Fri, 05 Jun 2026 03:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>企業 AI agent 可以查資料、發起動作，但必須繼承使用者許可權，遵守記錄範圍、欄位級安全、動作確認和審計要求，不能成為系統後門。</description><category>安全與治理</category><category>業務決策者</category><category>ObjectOS</category><category>AI Agent</category><category>許可權治理</category><category>企業安全</category></item><item><title>後設資料，不是程式碼生成：AI 應用為什麼可治理</title><link>https://www.objectos.ai/zh-Hant/blog/metadata-not-code-generation/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/metadata-not-code-generation/</guid><pubDate>Fri, 05 Jun 2026 02:40:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>程式碼生成能讓原型變快，但企業應用真正需要的是物件、欄位、關係、檢視、許可權、流程、動作和 agent 工具共同受控的後設資料執行時。</description><category>應用搭建</category><category>業務決策者</category><category>AI Builder</category><category>後設資料</category><category>應用搭建</category><category>程式碼生成</category></item><item><title>從自然語言到應用後設資料：AI Builder 如何生成物件和許可權</title><link>https://www.objectos.ai/zh-Hant/blog/natural-language-to-app-metadata/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/natural-language-to-app-metadata/</guid><pubDate>Fri, 05 Jun 2026 02:30:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI Builder 真正重要的不是把一句話變成頁面，而是把業務需求拆成物件、欄位、檢視、流程、許可權、自動化和 agent 工具。</description><category>應用搭建</category><category>業務決策者</category><category>AI Builder</category><category>自然語言搭建</category><category>後設資料</category><category>Agent</category></item><item><title>Airtable 式 AI Builder：用表格理解應用，用對話修改應用</title><link>https://www.objectos.ai/zh-Hant/blog/airtable-style-ai-builder/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/airtable-style-ai-builder/</guid><pubDate>Fri, 05 Jun 2026 02:20:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>最好的 AI Builder 把表格式搭建、對話式修改和受治理的後設資料結合起來：物件、欄位、檢視、許可權和自動化都可見，也都能被審查。</description><category>應用搭建</category><category>業務決策者</category><category>AI Builder</category><category>Airtable</category><category>無程式碼</category><category>自然語言互動</category></item><item><title>對話式應用迭代：加欄位、改流程、生成檢視和自動化</title><link>https://www.objectos.ai/zh-Hant/blog/conversational-app-iteration/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/conversational-app-iteration/</guid><pubDate>Fri, 05 Jun 2026 02:10:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI Builder 的真正價值不是“聊一句就改”，而是把欄位、流程、檢視、許可權和自動化都落到可審查、可回滾的後設資料層。</description><category>應用搭建</category><category>業務決策者</category><category>對話式應用搭建</category><category>AI Builder</category><category>自動化</category><category>許可權</category></item><item><title>AI 工單中樞：讓客服系統讀懂客戶問題</title><link>https://www.objectos.ai/zh-Hant/blog/ai-ticket-hub/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-ticket-hub/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>客服工單不應該只負責排隊。把物件、佇列、SLA、知識庫和受控工具做成後設資料後，AI 才能在許可權邊界內理解問題、推薦回覆並推動流轉。</description><category>應用搭建</category><category>業務決策者</category><category>案件管理</category><category>客戶門戶與自助</category><category>AI工單</category><category>客服系統</category><category>自然語言搭建</category><category>後設資料驅動</category></item><item><title>AI 銷售助理：如何更新 CRM 並建議下一步</title><link>https://www.objectos.ai/zh-Hant/blog/ai-sales-assistant/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-sales-assistant/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 銷售助理不只是自動填表，而是把客戶、聯絡人、商機、跟進和任務做成可治理的後設資料，讓銷售用對話推進商機，同時保留許可權和審計。</description><category>應用搭建</category><category>業務決策者</category><category>CRM 客戶管理</category><category>AI銷售</category><category>CRM</category><category>自然語言搭建</category><category>Agent</category></item><item><title>AI 專案管理助手：從進度更新裡發現風險</title><link>https://www.objectos.ai/zh-Hant/blog/ai-project-risk-assistant/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-project-risk-assistant/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 專案助手的價值不是再做一個看板，而是把專案、任務、會議、風險、變更和行動計劃做成物件，從日常更新中識別延期和阻塞。</description><category>應用搭建</category><category>業務決策者</category><category>AI專案管理</category><category>風險識別</category><category>會議紀要</category><category>自然語言搭建</category></item><item><title>AI 採購風險：下單前看見供應商風險</title><link>https://www.objectos.ai/zh-Hant/blog/ai-procurement-risk/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-procurement-risk/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 採購應用不只是給供應商打分，而是把供應商、資質、報價、合同、履約和風險做成後設資料，讓採購在下單前看見依據。</description><category>應用搭建</category><category>業務決策者</category><category>供應鏈與採購</category><category>製造</category><category>AI採購</category><category>供應商風險</category><category>供應鏈</category><category>採購決策</category></item><item><title>AI 報銷稽核：理解政策，而不只是識別發票</title><link>https://www.objectos.ai/zh-Hant/blog/ai-expense-audit/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-expense-audit/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 報銷稽核不止識別發票，而是把費用政策、預算、專案、審批、異常模式和審計記錄做成後設資料，讓財務能解釋和處理風險。</description><category>應用搭建</category><category>業務決策者</category><category>金融</category><category>AI財務</category><category>報銷稽核</category><category>費用控制</category><category>審計</category></item><item><title>AI 員工服務中心：從 HR/IT 工單到可執行服務</title><link>https://www.objectos.ai/zh-Hant/blog/ai-employee-service-center/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-employee-service-center/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 企業服務中心不是把 HR、IT、行政工單換成聊天框，而是把服務目錄、知識庫、申請物件、審批流和受控動作做成可執行服務。</description><category>應用搭建</category><category>業務決策者</category><category>人事與內部應用</category><category>案件管理</category><category>AI員工服務</category><category>企業服務中心</category><category>HR</category><category>IT服務</category></item><item><title>AI 合同審查：先標風險，再讓法務簽字</title><link>https://www.objectos.ai/zh-Hant/blog/ai-contract-risk/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-contract-risk/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 合同應用不只是總結合同，而是把合同型別、條款、義務、風險規則、審批和審計做成後設資料，讓 AI 標風險、法務確認責任。</description><category>應用搭建</category><category>業務決策者</category><category>金融</category><category>AI合同審查</category><category>法務科技</category><category>風險管理</category><category>後設資料驅動</category></item><item><title>AI 內容工作臺：從選題到釋出與覆盤</title><link>https://www.objectos.ai/zh-Hant/blog/ai-content-workbench/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-content-workbench/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>內容團隊需要的不是孤立寫作框，而是把選題、素材、草稿、稽核、釋出和覆盤做成物件，讓 AI 參與完整內容運營流程。</description><category>應用搭建</category><category>業務決策者</category><category>電信媒體</category><category>AI內容運營</category><category>內容工作臺</category><category>自然語言流程</category><category>應用搭建</category></item><item><title>AI 內控檢查：把制度變成可追溯控制</title><link>https://www.objectos.ai/zh-Hant/blog/ai-compliance-control/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-compliance-control/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>AI 內控應用不是制度問答，而是把制度條款、控制點、證據、缺口、整改和審計做成物件；AI 發現缺口，人確認責任，執行時保留證據鏈。</description><category>應用搭建</category><category>IT / CIO</category><category>金融</category><category>AI合規</category><category>內控檢查</category><category>審計</category><category>治理</category></item><item><title>AI Agent 工作臺：讓 agent 在業務系統內受控執行</title><link>https://www.objectos.ai/zh-Hant/blog/ai-agent-workbench/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-agent-workbench/</guid><pubDate>Fri, 05 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>企業 agent 不能只是會聊天，而要能在業務物件、工具、許可權、審批和審計邊界內執行任務。工作臺的關鍵是受控執行，而不是萬能許可權。</description><category>應用搭建</category><category>IT / CIO</category><category>AI Agent</category><category>業務工作臺</category><category>工具呼叫</category><category>許可權治理</category></item><item><title>自託管 AI 應用平臺：為什麼執行時屬於你</title><link>https://www.objectos.ai/zh-Hant/blog/self-hosted-ai-app-platform/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/self-hosted-ai-app-platform/</guid><pubDate>Thu, 04 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>當 AI 開始讀取業務資料、觸發流程、生成應用和呼叫工具，企業真正要控制的不只是模型，而是承載物件、許可權、工具、審批和審計的執行時。</description><category>安全與治理</category><category>IT / CIO</category><category>自託管</category><category>私有化部署</category><category>資料安全</category><category>AI治理</category></item><item><title>從需求到可審查應用：AI 如何生成 ObjectStack 後設資料</title><link>https://www.objectos.ai/zh-Hant/blog/from-requirement-to-app/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/from-requirement-to-app/</guid><pubDate>Thu, 04 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>用裝置報修場景拆開 AI Builder 的生成過程：物件、欄位、關係、檢視、許可權、動作、流程、API 和 agent 工具如何由同一份後設資料驅動。</description><category>應用搭建</category><category>開發者</category><category>AI Builder</category><category>應用搭建</category><category>後設資料</category><category>物件建模</category></item><item><title>AI Agent 資料安全邊界：如何在企業許可權內工作</title><link>https://www.objectos.ai/zh-Hant/blog/ai-agent-business-data-security-boundaries/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/ai-agent-business-data-security-boundaries/</guid><pubDate>Thu, 04 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>企業不是不想讓 AI agent 使用業務資料，而是不允許它繞過身份、許可權、審批和審計。真正可上線的 agent，必須像一個受控使用者，而不是影子管理員。</description><category>安全與治理</category><category>IT / CIO</category><category>AI Agent</category><category>許可權治理</category><category>資料安全</category><category>審計</category></item><item><title>製造業既有系統怎麼接 AI：先從報表和工單開始</title><link>https://www.objectos.ai/zh-Hant/blog/manufacturing-legacy-systems-ai/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/manufacturing-legacy-systems-ai/</guid><pubDate>Wed, 03 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>製造業系統鏈路複雜，ERP、MES、WMS、裝置臺賬和工單系統都不能輕易替換。更務實的 AI 路線，是先連線現有系統，從報表、工單和異常分析切入。</description><category>整合與資料</category><category>IT / CIO</category><category>案件管理</category><category>供應鏈與採購</category><category>製造</category><category>製造業</category><category>ERP</category><category>工單</category><category>AI落地</category></item><item><title>低程式碼 vs AI 原生應用平臺：複雜業務卡在哪裡</title><link>https://www.objectos.ai/zh-Hant/blog/low-code-vs-ai-native-app-platform/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/low-code-vs-ai-native-app-platform/</guid><pubDate>Wed, 03 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>低程式碼解決的是更快搭頁面和流程；複雜業務真正卡住的是物件、許可權、整合、變更和可維護性。AI 原生平臺要把這些變成可審查的執行時後設資料。</description><category>應用搭建</category><category>IT / CIO</category><category>人事與內部應用</category><category>案件管理</category><category>低程式碼</category><category>AI-native</category><category>應用平臺</category><category>系統架構</category></item><item><title>CRM AI：讓 agent 在許可權內讀取客戶和商機</title><link>https://www.objectos.ai/zh-Hant/blog/crm-ai-understands-customers/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/crm-ai-understands-customers/</guid><pubDate>Wed, 03 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>很多公司的 CRM 裡已有客戶、商機、聯絡人和跟進記錄。真正有價值的做法不是匯出資料問一次，而是讓 agent 在許可權之下讀懂這些業務物件。</description><category>整合與資料</category><category>業務決策者</category><category>CRM 客戶管理</category><category>CRM</category><category>銷售管理</category><category>AI落地</category><category>客戶資料</category></item><item><title>定製系統技術債：為什麼半年後沒人敢改</title><link>https://www.objectos.ai/zh-Hant/blog/why-custom-systems-die/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/why-custom-systems-die/</guid><pubDate>Tue, 02 Jun 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>很多內部系統不是越用越順，而是越用越僵。根因往往不在程式設計師，而在業務規則被埋進程式碼裡，後續既難審查，也難讓 AI 安全修改。</description><category>系統現代化</category><category>業務決策者</category><category>AI落地</category><category>數字化轉型</category><category>趨勢觀點</category></item><item><title>給現有系統加 AI：連線資料庫，而不是先遷移</title><link>https://www.objectos.ai/zh-Hant/blog/extend-existing-systems-with-ai/</link><guid isPermaLink="true">https://www.objectos.ai/zh-Hant/blog/extend-existing-systems-with-ai/</guid><pubDate>Sat, 30 May 2026 00:00:00 GMT</pubDate><dc:creator>ObjectStack Team</dc:creator><description>把 ObjectOS 連到已經執行的資料庫，讓 agent 把關鍵資料表建模為物件，再在你的許可權和伺服器邊界內疊加 AI 能力；原系統繼續執行，AI 走受控物件層。</description><category>整合與資料</category><category>通用</category><category>Data Sources</category><category>AI-Native</category><category>Architecture</category></item></channel></rss>