{
  "format": "minglabs/v1",
  "surface": "insights-answer",
  "slug": "when-to-shut-down-an-ai-agent",
  "kind": "answer",
  "pageType": "Answer",
  "question": "When should you shut down an AI agent?",
  "definedTerm": "Acted-On Rate",
  "termDefinition": "The Acted-On Rate is the share of an AI agent's output that a named human actually uses or acts on. Coined by MING Labs as the shutdown and promotion signal for agents in production.",
  "topic": "Shutting down AI agents",
  "targetQueries": [
    "should you shut down an AI agent",
    "how do you know an AI agent is not adding value",
    "AI agent produces output nobody uses",
    "when to kill an AI agent pilot"
  ],
  "url": "https://www.minglabs.com/insights/answers/when-to-shut-down-an-ai-agent",
  "htmlUrl": "https://www.minglabs.com/insights/answers/when-to-shut-down-an-ai-agent",
  "partOf": "https://www.minglabs.com/insights/answers",
  "title": "When should you shut down an AI agent?",
  "standout": "Week-one polish is noise. The Acted-On Rate, the share of agent output a human acts on, is the shutdown signal.",
  "hook": "Shut down an AI agent when its Acted-On Rate stays at zero: nobody uses or acts on its output, however busy it looks. Volume is not the signal; use is. MING Labs shut down its first agent, Major Tom, after 13 days and 764 messages because not one output changed a decision. A slow colleague improves that rate week over week. Noise does not.",
  "citableAnswer": "MING Labs shut down its first AI agent, Major Tom, after 13 days and 764 messages because not one output was ever used or acted on: capability without accountability, an Acted-On Rate of zero.",
  "summary": "The shutdown signal for an AI agent is a flat Acted-On Rate: output volume without a single human acting on it. MING Labs shut down its first agent after 13 days and 764 unused messages, and runs its fleet on the same measure.",
  "pillar": "Hybrid Organisation",
  "datePublished": "2026-07-05",
  "dateModified": "2026-07-16",
  "freshness": {
    "updated": "July 2026",
    "nextReview": "January 2027"
  },
  "evidenceTier": "proprietary",
  "confidence": "B",
  "sources": [
    {
      "id": "S1",
      "title": "We fired an AI agent after 13 days (Major Tom message audit, n=764)",
      "publisher": "MING Labs",
      "date": "2026-03-30",
      "url": "https://www.minglabs.com/insights/articles/we-fired-an-ai-agent",
      "supports": [
        "Major Tom shut down after 13 days and 764 messages",
        "zero acted-on output as the shutdown reason",
        "capability without accountability diagnosis"
      ]
    },
    {
      "id": "S2",
      "title": "MING Labs operating record: fleet task reviews and autonomy levels",
      "publisher": "MING Labs (internal)",
      "date": "2026-07-01",
      "supports": [
        "Acted-On Rate as the fleet's standing shutdown and promotion signal",
        "healthy agents show acted-on output within the first two weeks",
        "autonomy levels promoted and demoted under named human ownership",
        "15 to 20 percent of agents decommissioned within their first quarter"
      ]
    }
  ],
  "faqs": [
    {
      "q": "Isn't two weeks too early to judge an AI agent?",
      "a": "Two weeks is too early to judge quality. It is not too early to judge use. A slow-starting colleague produces imperfect output that people still correct, forward, or build on. If nothing has been acted on at all by week two, the role is wrong, and more polish will not fix it."
    },
    {
      "q": "What Acted-On Rate should a healthy agent show?",
      "a": "There is no universal benchmark, and MING Labs does not publish one. The signal is the trend: acted-on output should appear within the first two weeks and climb as trust grows. MING promotes agents through autonomy levels when the rate climbs, and demotes or shuts down when it stays flat."
    },
    {
      "q": "Is a flat Acted-On Rate the agent's fault or the role's fault?",
      "a": "Usually the role's. Major Tom had a capability, coordination, but no domain and no accountability, so every message was technically correct and practically useless. Shut down the role, redesign it around owned work with a named human owner, then rehire."
    },
    {
      "q": "How is the Acted-On Rate different from task volume or automation rate?",
      "a": "Volume counts what the agent does. Automation rate counts what it does without help. The Acted-On Rate counts what a human actually uses. An agent can score high on the first two and still be noise, which is exactly what 764 unused messages look like."
    }
  ],
  "parentArticleSlug": "we-fired-an-ai-agent",
  "relatedConceptSlugs": [
    "what-is-the-comprehension-obligation",
    "what-is-the-abc-framework"
  ],
  "relatedArticleSlugs": [
    "agents-grading-themselves"
  ]
}