OpenAI has started deploying a novel attribute on ChatGPT as of Tuesday, allowing individuals to organize reminders and actions for the AI (artificial intelligence) assistant. Known as Tasks, this fresh functionality is currently accessible to all premium subscribers of the AI service, with the company disclosing intentions to extend it to the complimentary tier within the year. This component also seems to signify OpenAI’s initial move towards incorporating agentic properties in its renowned assistant.

OpenAI Introduces Tasks Feature for ChatGPT

Through a series of publications on X (previously identified as Twitter), OpenAI has unveiled Tasks, presently accessible in beta form to ChatGPT Plus, Teams, and Pro audiences. Expansion to users on the AI service’s free tier is scheduled for later in the year. At this time, the Tasks page is accessible solely via the platform’s web interface.

The organization expressed that it seeks to utilize the beta phase to gain a better understanding of how individuals engage with the feature and to fine-tune it before making it available to free-tier users. Furthermore, users are currently unable to utilize the Advanced Voice Mode for task setting.

Tasks manager page in ChatGPT web
Photo Credit: OpenAI

Tasks represent fundamentally a planning feature. Users can access ChatGPT web to either input a prompt in the text area or navigate to the Tasks page through the profile menu to establish a new task. These chores can be reminders set either as a one-off on a chosen date and time or as repeated occurrences. Users may also direct the assistant to execute a task at the predetermined time.

This attribute is applicable for obtaining a morning news digest or a motivational note right before exercise. Since prompts may configure Tasks, users can just input a natural language command such as “Send me an overview of the leading tech news of the day at 7:30pm on weekdays,” and it will configure the timetable.

While the organizing aspect of this feature is not particularly revolutionary (every smartphone alarm app permits users to establish reminders), the “action execution” at a later moment is particularly intriguing. Although technically not agentic as no activity occurs external to the assistant platform, it nonetheless requires the AI to utilize memory and sequential execution to create text or imagery at a designated moment. Such functionalities combined are typically linked with AI agents.

There has been speculation about OpenAI developing its inaugural AI agent, named Operator, for some time, and Tasks might be the preliminary stride towards that objective.