OpenAI was first to develop artificial intelligence capable of operating browsers and other applications on behalf of users, with its ChatGPT app also bringing generative AI to mainstream consumers. Now that competitors like SpaceX and Meta keep pushing technological boundaries with personal AI agents that handle multi-step tasks, OpenAI is working to catch up. According to sources familiar with the projects, OpenAI is developing new features that directly target the Grok Bot launched recently by SpaceX. That product allows users to create always-online AI "collaborative partners" that manage daily affairs, operate software, and execute coordinated work in the background automatically. Another insider said OpenAI is also discussing building a personal AI assistant to compete with a similar product Meta unveiled this month. In developing these tools, OpenAI will reuse and repackage existing agentic technology from Codex and ChatGPT. Some OpenAI app users already connect the product to email, calendars, and other applications to automate tasks such as scheduling appointments and contacting businesses via email, with some operations spanning multiple days. OpenAI has stated that since the release of GPT-5 last year, each successive model iteration has been designed for long-horizon tasks, and these capabilities have been integrated into related applications. However, many OpenAI users are unaware of these features. In recent weeks, SpaceX and Meta have captured market attention with their respective AI agent products. Grok Bot, launched in August, aims to help users quickly build bots tailored to different scenarios, such as conducting market research or planning advertising campaigns. Users can grant bots access to Gmail, Google Calendar, Notion, Figma, or various office tools like accounting, finance, and AI code assistants, allowing the bots to complete tasks directly within those applications. Users can also let the AI observe their operations across software and learn to replicate those workflows autonomously. Bots proactively report back when tasks are complete or when they need information from the user. Users can also set permissions requiring bot approvals before actions like sending emails, initiating purchases, transferring funds, or modifying data in other applications. Similarly, Meta launched Muse this month, which automatically handles common mobile tasks such as contacting merchants, replying to emails, and online shopping. Before Muse's debut, the personal assistant app Instinct had already gained traction, currently operating on an invitation-only basis but surpassing 100,000 users. Although consumers have grown accustomed to using AI chatbots like ChatGPT and Google Gemini for information retrieval, advice, or project assistance, these conversational AI tools have consistently underperformed when handling more complex personal matters such as online appointment booking. It remains unclear when OpenAI will roll out features that rival Grok Bot and Muse. The company must balance numerous priorities, including ensuring the safety and alignment of its next-generation models and the agent clusters those models can power. But OpenAI also cannot afford to watch rivals claim the leading position in the AI agent space.