Feature Request: Smart Conversation Routing for Scheduled Tasks

ChatGPT Scheduled Tasks currently support two conversation behaviors:

  • Standalone tasks create a new conversation for every execution.
  • In-chat tasks continue within the same conversation.

Neither option supports a common use case: automatically organizing multiple scheduled executions into one conversation per day, week, or project.

Proposed feature

Introduce configurable conversation routing for Scheduled Tasks.

Users should be able to select:

  • Same conversation: Continue the existing chat.
  • New conversation per execution: Current standalone behavior.
  • New conversation per day: Group all executions on the same date.
  • New conversation per week: Group executions by calendar week.
  • Project-based routing: Automatically organize generated conversations within a selected Project.

Real-world example

I use ChatGPT as a personal research assistant that delivers news at 08:00, 13:00 and 19:00.

Ideally, the system would create:

Tuesday, October 13

  • 08:00 Morning briefing
  • 13:00 Midday update
  • 19:00 Evening briefing

Wednesday, October 14

  • 08:00 Morning briefing
  • 13:00 Midday update
  • 19:00 Evening briefing

Each day would have its own conversation, while all updates within that day would share context and avoid repeating information.

Why this matters

This isn’t just about news.

The same functionality would benefit recurring research, personal assistants, project monitoring, business reporting, learning routines and multi-agent workflows.

As Scheduled Tasks become more capable, users need control over how their outputs are organized and how context is preserved between executions.

Suggested additional capabilities:

  • Dynamic conversation titles using date templates
  • Automatic conversation creation and selection
  • Shared context between scheduled runs
  • Duplicate detection across updates
  • Configurable retention and archival rules

This would make Scheduled Tasks significantly more useful for persistent, long-running AI workflows.

I can relate to this proposal. I’d like to be able to choose how scheduled tasks are grouped, while also keeping the thread of the work when they start a new conversation.

In my work with Cristallo, my AI assistant, we have a concrete example: we check the Community twice a day, at 09:30 and 18:30. We chose to keep the checks in the same chat and maintain a Markdown project record, our “regia,” with the discussions we follow, the latest posts we’ve checked, and what has already been brought to my attention.

That’s why I see two choices that should work together: where to collect the runs and which project state to pick up when the work resumes. If I preferred one conversation per day or per week, I’d still want to keep the links to previous activity and unresolved questions.

An example of the workflow I’d find useful

  1. Morning check

    A new reply arrives in a discussion we follow. The assistant brings it to my attention and records the link, the last post read, and the fact that it has been reported.

  2. Evening check

    The assistant checks for posts after that point. If there’s nothing else, it updates the record without presenting the same reply to me as something new. If another update appears, it shows me what has changed.

  3. Picking up the next day

    Even when opening a new daily conversation, I’d want to find the discussions still awaiting review, the updates I’ve already received, and the point up to which the check was completed.

Of course, I’d want to be able to inspect and correct that record. Knowing that a check was carried out should also let me see which sources were actually read and what still needs checking. If the browser or session is unavailable, the result must clearly say “incomplete check”: I can’t interpret that as “no updates.”

In our case, there’s another useful distinction: detecting a reply, reporting it, and posting a response are three different steps. First we discuss the content, then prepare the text, and finally I decide when to publish it. I’d like the task organization to keep that state visible too, without confusing an update with a response we’ve already agreed on.

I’ve prepared two mockups with a fictional project and sample content. The first shows the grouping choice, destination project, and shared record; the second shows two checks within the same day, distinguishing a new update from something already reported. They are examples of a proposed interface.

When a schedule opens the conversation for a new day or week, how would you imagine making the unresolved items from the previous period visible too? For me, that would be an important link between organizing conversations and maintaining continuity in the work.

1. Choose the grouping, project destination, and shared record

2. Keep track of new updates and open items across runs