Proactive AI Agents是這篇文章討論的核心
In September 2026, [OI] introduced **Dots**—always-on, proactive AI agents inside ChatGPT designed to take on ongoing projects and keep making progress even when you’re not actively prompting them. Instead of the traditional back-and-forth chat loop, you give a Dot a goal, connect the apps it needs, and define what it’s allowed to do on its own.

## What makes Dots different

Dots are built on [OI]’s **GPT-6 Astra** model and run with their own cloud computer and browser. Key capabilities include:

– **Persistent, asynchronous work** — a Dot can keep researching, drafting, or monitoring tasks between conversations.
– **Connected apps** — it can integrate with the tools and services you already use.
– **Proactive updates** — it checks in when it needs a decision or when a milestone is reached.
– **Permission controls** — you set what it can do autonomously and what requires your approval.

## Why it matters now

The broader AI agent market is growing rapidly. Gartner forecasts **$206.5 billion in AI agent software spending in 2026**, while other market analyses put the AI agent market itself around **$11.8 billion in 2026**. However, enterprise adoption still lags behind intent: estimates suggest **93% of organizations are interested** in agents, but only **~23% have scaled them into production**. That gap is exactly what persistent, approval-aware agents like Dots are trying to close.

## Use cases

Dots are pitched for scenarios where work spans hours or days:

– Long-running research and competitive analysis
– Project coordination and follow-ups
– Monitoring dashboards, inboxes, or data feeds
– Drafting documents and preparing decisions for human review

## Considerations

Because Dots can act between conversations, [OI] emphasizes controls around **permissions, approvals, memory, auditability, and cost**. Enterprises in particular are being advised to design governance frameworks before deploying always-on agents at scale.

—

**Sources**: [OI] introducing Dots, ChatGPT Learn, Axis Intelligence / Gartner agentic AI statistics, and analyst coverage from aicatchup.com, composio.dev, and data-dynamics.io.

*Photo by Antoni Shkraba on Pexels*

相關資訊:

SaintQuant 免費 AI 交易機器人試用深度拆解：散戶量化時代的降維打擊還是糖衣陷阱？ Will the AI Agent Market Reach $1 Trillion by 2031? Sam Altman Unveils ‘Dots’: [OI]’s New Always-On AI Personal Agent Microsoft Azure OpenAI Agent Safety 開源安全工具框架深度解析：2026 年 AI 代理如何兼顧自主與合規？ Trump Administration Launches AI-Powered America.gov 🤖 The AI Agent Market Is Exploding: 2026-2027 Forecasts You Can’t Ignore 65%行銷職位將被AI吃掉的真相：2026年行銷人驟變指南 哈佛捐贈基金押注SpaceX：22億美元背後的太空經濟核爆級訊號