{
  "slug": "manus",
  "client": "Manus",
  "title": "Creative research and automated brand review",
  "period": "Review activity: May 2026. Research example: March 4, 2026.",
  "lede": "Scaling creator content also scales the work of deciding what to film, checking each video and coordinating revisions. Amy AI Agent connects that work inside the Vizzy platform.",
  "metrics": [
    [
      "411",
      "submitted video versions",
      "May 2026"
    ],
    [
      "325",
      "completed AI brand reviews",
      "May 2026"
    ],
    [
      "42 sec",
      "median AI review turnaround",
      "365 completed April-May runs"
    ],
    [
      "305",
      "videos in one research corpus",
      "March 4 gameplan"
    ]
  ],
  "sections": [
    [
      "How the workflow works",
      "George Gui designed the Amy AI Agent workflow within Vizzy: creative research becomes creator-ready briefs, multimodal review checks speech, on-screen text and visual content, and structured feedback guides revisions. Persistent brand memory retains requirements and feedback across the execution and review loop."
    ],
    [
      "Recorded activity",
      "May's 411 video versions comprised 304 first submissions and 107 resubmissions, from 30 active creators. The workflow completed 325 AI brand reviews and posted 325 automatic review comments. The 42-second median measures review creation to completion across 365 completed April-May runs, including queue and processing time."
    ],
    [
      "Research that creators can use",
      "The March 4 gameplan analyzed a 30-day corpus of 305 videos and over 150 comments to develop five briefs. A separate March 20 gameplan contained 11 briefs and 70 hook options; the April 25 plan contained 20 creative concepts. These give creators concrete directions while leaving room for their delivery style."
    ],
    [
      "Review workload calculation",
      "At May's activity level, the estimated equivalent manual review workload is 41-82 hours: 411 versions x 5-10 minutes / 60 x 1.20. The design record supplies the 5-10 minute full-review baseline; 20% is a coordination allowance. The estimate covers full review and review-related coordination at May's volume. The 325 AI-reviewed versions correspond to 32.5-65 hours under the same model."
    ]
  ],
  "sources": [
    [
      "Review design and baseline",
      "https://fe9fyzyw00x.sg.larksuite.com/docx/ChMEdTJCMorL5Vx1C67l5W8ngAe"
    ],
    [
      "March 4 research and five briefs",
      "https://fe9fyzyw00x.sg.larksuite.com/docx/LvQQd0rlWoiy4QxAl59lL8Kjg1b"
    ],
    [
      "March 20 briefs and hooks",
      "https://fe9fyzyw00x.sg.larksuite.com/docx/X4NkdVmznoCqjTx3xUJlF32LgsW"
    ],
    [
      "April 25 creative concepts",
      "https://fe9fyzyw00x.sg.larksuite.com/docx/Frk3dA3TpogC8CxvoXYlUzw9gZg"
    ]
  ],
  "method": "Vizzy submission, resubmission, AI review and feedback records extracted September 29, 2026. Monthly boundaries use UTC. A video version is a distinct video ID in a submission or resubmission event; a revised video counts as another version. Completed AI reviews are VIO brand-review runs within the Amy workflow. Research batches and monthly review activity are reported separately.",
  "rows": [
    [
      "2026-04",
      330,
      188,
      142
    ],
    [
      "2026-05",
      411,
      304,
      107
    ]
  ],
  "columns": [
    "month_utc",
    "distinct_submitted_versions",
    "first_submissions",
    "resubmissions"
  ]
}