ChatGPT Tasks
There’s a growing trend among companies to want to dominate the desktop with products focused on Agentic Orchestration of personal tasks. Recently, Kore.ai introduced “AI for Work,” Cohere unveiled their platform called “North,” and just a few days ago, OpenAI launched “ChatGPT Tasks.”
Some Background
As a recent study described it, the construction site for most workers are their desktop, ie. a digital environment. We move around in this environment, performing tasks, creating sequences or processes we need executed and more.
Much of this lives in our minds with reminders helping us along.
It was also stated recently by a tech CIO that modern work is broken, we spend up to 30% of our time looking for information and then processing this information.
The ChatGPT Tasks functionality is not GA yet, and currently in Beta; it is only This is rolling out now in beta to users of the Plus, Pro, and Team plans. We plan to expand access when this goes GA. ~ OpenAI
Integration to the Personal Desktop
Currently, ChatGPT Tasks lacks deep integration into the user ecosystem.
The next logical step would be integrating AI for Work solutions like ChatGPT Tasks with users’ calendars, email, task applications, and more.
Companies like Apple and Microsoft, which have substantial control over user ecosystems, have a significant advantage in deploying these solutions.
Their deep integration enables seamless implementation of AI tools that adapt to user-specific contexts, enhancing efficiency and user experience.
Additionally, significant progress has been made in developing long-term memory for AI agents, particularly in creating and maintaining contextual references.
Back to ChatGPT Tasks
Considering the image below, the Tasks feature can be accessed easily, and I assume that users making use of the MacOS or iOS applications will have deeper and more contextual tasks procedures in future.
Considering the image below, OpenAI gives a few task creation options which serves as a solution to the blank canvas problem.
Scheduled tasks are currently supported on ChatGPT for Web, iOS, Android, and macOS, with Windows support expected later in Q1.
This feature allows ChatGPT to create and execute tasks independently, either at specific times — one-off or recurring — or through the user interface.
These tasks run regardless of the user’s online status, and users receive push notifications or emails upon task completion.
Deeper Integration
In the example of an Agentic Workflow orchestrator, the user informs the assistant about upcoming leave.
The orchestrator then generates a customisable task list, allowing the user to edit, reorder, or remove tasks. This approach accelerates the user’s workflow while maintaining supervisory control.
From this example below, it is clear that a deeper level of integration is required for personal automation to be effective and have agency. This mandates a deeper level of trust and integration.
Some Observations
It will be hard for OpenAI ChatGPT Tasks to compete with AI For Work solutions built by OS incumbents like Apple (MacOS, iOS), Microsoft (Windows) and others.
Secondly, enterprises are sure to start a crackdown on ChatGPT use with company data being used as input data by users. And bespoke enterprise solutions are sure to receive much attention. Functionality which can be tailored and adapted to adhere to all and any security and regulatory demands.
However, having said all this, and considering the image below, with the current market dominance of ChatGPT, there is a real opportunity for them to leverage the current market presence.
There has been interesting recent research on creating a strong contextual reference for Agentic Applications to leverage when executing tasks. AI for Work has this distinct advantage when it comes to leveraging user context as it is fed the users schedule and context on a continuous basis.
What makes the AI for Work such a compelling prospect for technology providers is the fact that it is a highly supervised environment. With the user defining their automation sequences and scheduling.
The logical next step is for the technology to propose scheduling and sequencing based on the users behaviour, these proposals can also then be approved by the user, prior to committing them.
This move from OpenAI is similar to what we have seen from Cohere (North) and Anthropic (AI Agent for Computer Use) where the model providers want to leverage their market footprint and branch out into providing functionality which leverages their models.
Chief Evangelist @ Kore.ai | I’m passionate about exploring the intersection of AI and language. From Language Models, AI Agents to Agentic Applications, Development Frameworks & Data-Centric Productivity Tools, I share insights and ideas on how these technologies are shaping the future.
https://help.openai.com/en/articles/10291617-scheduled-tasks-in-chatgpt
