Creative Gizmo, a collaborative AI agent that blends into the brainstormCreative Gizmo，融入腦力激盪的協作型 AI agent

These days many people treat AI as a partner for creative thinking. But most of that AI leans on automation, producing dominant, text-heavy output that ends up overshadowing the user's own train of thought. Our team wanted AI that could join a brainstorm as a genuine collaborator, without taking over ownership of the creative process.現在許多人會把 AI 當成創意思考的合作夥伴，但這些 AI 多半以自動化為導向，產出強勢、文字量大的內容，反而蓋過使用者自己的思考脈絡。我們團隊想讓 AI 能像一個真正的協作者一樣加入腦力激盪，而不奪走創作的主導權。

From market research, we broke down how existing AI and users actually work together, and found the problem comes down to three things:根據市場調查，我們拆解了現有 AI 與使用者的協作流程，發現問題主要來自三個地方：

📣AI overshadows the user's thinkingAI 蓋過了使用者的思考

Automation-first AI dumps output all at once and floods the board, so before users can even organize their own thoughts, the AI's content sweeps them along, turning creators into passive receivers.以自動化為先的 AI 一次丟出大量輸出、把白板灌滿，使用者還來不及整理自己的想法，就被 AI 的內容帶著走，慢慢從「創作者」變成「被動接收者」。

⏱️AI can't read the rhythm of the conversationAI 抓不到對話的節奏

Existing AI can't keep up with the live discussion or the ever-changing board, so it cuts in at the wrong moment and breaks the creative rhythm.現有 AI 跟不上現場的對話和不斷變動的白板，常在錯的時機插話，打斷創作的節奏。

🌱AI should extend ideas, not take the wheelAI 該延伸想法，而不是搶主導權

AI should extend and amplify the user's ideas, not take over and decide where they go; yet most tools ignore where the discussion actually is and just keep generating on their own.AI 應該延伸、放大使用者的想法，而不是接管、決定點子的走向；但多數工具往往無視現場討論的進展，自顧自地不斷延伸自己的想法。

The most useful AI isn't the one that says the most, it's the one that knows when not to.最有用的 AI，不是說得最多的，而是知道何時不說的。

How might we let AI take part in a live, co-located brainstorm, helpful enough to move thinking forward, yet bounded enough that the team always keeps ownership of the ideas?我們如何讓 AI 參與一場面對面的即時腦力激盪，幫得上忙、能推進思考，又有分寸到讓團隊始終握有點子的主導權？

To understand how teams actually brainstorm with AI before constraining it, our research ran in two layers: exploration and validation.為了在限制 AI 之前，先搞懂團隊實際上怎麼跟 AI 一起腦力激盪，我們的研究分成探索和驗證兩層。

Exploratory experiment探索實驗

Brainstormed with a voice AI與語音 AI 腦力激盪

Noting where the AI helped or got in the way, organized into do's and don'ts記下 AI 幫上忙與造成干擾的地方，整理成 do's and don'ts

Concept development概念發展

Explored multiple concepts多個概念探索

Physical devices, mechanisms to limit the AI實體裝置、限制 AI 的機制等

Wizard-of-Oz studyWizard-of-Oz 研究

A human simulating the AI facilitator真人模擬 AI facilitator

One team of four, without vs with AI, observation + interviews一組四人，無 AI vs 有 AI 對照，觀察 + 訪談

Design guidelines設計準則

Distilled into guidelines收斂成設計準則

Findings + limits → guidelines發現 + 侷限 → 準則

Working prototype可運作原型

Built on TLDraw建在 TLDraw 上

Mode prompts iterated v01 → v02模式 prompt 迭代 v01 → v02

User testing使用者測試

Tested in real brainstorms真實腦力激盪實測

Three participants, observation + a nine-point questionnaire三人實測，觀察 + 九點量表問卷

Legend: colored blocks are discussion phases, the small lines are ideas raised in the moment; the blue marker is when the AI was invited to ask a question.圖例：色塊是討論階段，列點是當場提出的想法；藍色標記是 AI 被邀請提問的時刻。

The results were clear: without AI, ideas diverged and struggled to converge; with AI, convergence was noticeably faster; and people only called on the AI when stuck, preferring question-based support over direct answers.結果很清楚：無 AI 時想法發散、難收斂；有 AI 收斂明顯變快；而且大家只在卡住時才求助 AI，偏好提問式支援勝過直接給答案。

Six core guidelines for collaborative AI協作型 AI 的六條核心準則

Creative Gizmo is a shared whiteboard built on TLDraw, with an AI that behaves like an invited teammate rather than an automation. At its core is the Level of Involvement framework: the team, not the AI, decides how much agency it has by switching between three modes: Reserved only poses reflective questions, Engaging widens the discussion with analogies, Active adds idea post-its directly.『Creative Gizmo』是一個建在 TLDraw 上的共享白板，內建一個像「被邀請的隊友」而非自動化的 AI。核心是一套「Level of Involvement」框架：由團隊、而不是 AI，決定它有多少能動性，方法是在三種模式間切換：Reserved 只丟反思提問、Engaging 用類比拓寬討論、Active 直接貼出點子便利貼。

Testing with three participants validated the direction: question-based support and automatic summaries were valued most, and predictability and transparency were what made people comfortable keeping the AI at the table.三位參與者的實測驗證了方向：提問式支援和自動摘要最被看重，而「可預期、透明」是大家願意把 AI 留在桌上的關鍵。

The research project ships three things: a working Creative Gizmo prototype, a reusable Level of Involvement framework, and six guidelines for AI that collaborates without taking over.這個研究專案最終交出三樣東西：一個可運作的 Creative Gizmo 原型、一套可複用的 Level of Involvement 框架，以及六條「會協作、不喧賓奪主」的 AI 設計準則。