What Are Event-Triggered Work Tasks in ChatGPT in 2026?
An event-triggered Work task is a webhook-based job that ChatGPT runs when something happens in a connected app, rather than at a time you pick. OpenAI states that eligible Plus, Pro, Business, Enterprise, Edu and ChatGPT for Healthcare users can ask Work to run a task when a supported event occurs. The supported events in 2026 are new Gmail messages, new Slack channel messages, and GitHub pull request activity in an authorised repository.
This sits alongside the older option. A Work task can run once, repeat on a schedule, or monitor for changes through Scheduled Tasks. The event-triggered form is the one that closes the gap between an assistant you remember to ask and a process that runs whether anyone remembers or not.
Explain It Like I Am Five
Imagine a doorbell with a note taped under it. The note says: when someone rings, check whether they are the postman, and if they are, take the parcel to the kitchen. You do not have to sit by the door waiting. The bell is the trigger, the "is it the postman" part is the condition, and "take the parcel to the kitchen" is the instruction. An event-triggered task in ChatGPT is exactly that note, except the doorbell is a new email, a new Slack message, or a change to a code request, and the kitchen is wherever you told it to put the result. You write the note once and then go and do something else.
| Task type | When it runs | Best for in 2026 |
|---|---|---|
| Run once | Immediately, as a single job | A one-off deliverable you hand over and collect. |
| Scheduled | On a repeating schedule you set | Weekly reports, monthly summaries, recurring checks. |
| Monitoring | Watches for changes | Keeping an eye on a source that updates unpredictably. |
| Event-triggered | When a supported event fires in a connected app | Inbox and channel triage, pull request reactions, anything where latency matters. |
Why Do Event Triggers Matter for Business Teams in 2026?
Because most operational work is reactive, and a schedule is a poor fit for reactive work. A daily digest of enquiries that arrived yesterday is worth less than a triaged enquiry three minutes after it landed. Event triggers move ChatGPT from something a person opens to something that responds to the systems a team already lives in, which in 2026 means email, chat and the code review queue.
- Latency becomes a design choice: a trigger fires on the event, so the response time is set by the work rather than by the next scheduled run.
- Three integration points that cover most teams: Gmail for inbound, Slack for internal signal, GitHub pull requests for delivery work.
- Conditions keep the noise out: the task's Condition field decides whether the prompt runs at all, so a busy channel does not become a busy task log.
- Sharing distributes the pattern: an eligible task can be shared, and a recipient can schedule their own separate copy, so a working recipe can spread across a team.
- Approvals stay in the loop: connected-app permissions, workspace controls and approval requirements continue to apply, so a trigger does not silently widen what ChatGPT can do.
How Do You Create an Event-Triggered Task in 2026?
You ask Work for it in plain language, then inspect the three parts it produces. Open the task to review Trigger, Condition and Prompt, and go to Scheduled to manage it afterwards. Those three fields are the whole model, and they are worth writing deliberately rather than accepting whatever the first attempt produced.
| Field | What it controls | How to write it well in 2026 |
|---|---|---|
| Trigger | Which event in which connected app starts the task | Pick the narrowest source that still catches what you need. One Slack channel beats a workspace-wide watch. |
| Condition | Whether the prompt should run for this particular event | This is your noise filter. State explicitly what should be ignored, not only what should pass. |
| Prompt | What Work actually does when the condition passes | Name the deliverable and where it goes. Include the review criteria, as with any Work task. |
Where you can and cannot create them
- Create or edit on web, or in supported mobile apps. That is where trigger conditions can be set.
- The desktop app can display existing tasks but cannot create or edit their trigger conditions. This catches out desktop-first users regularly.
- For a Slack-triggered task, add @ChatGPT to every Slack channel the task monitors. A task pointed at a channel ChatGPT was never invited to will not fire.
- Admins must switch it on for managed workspaces. Enterprise, Edu and ChatGPT for Healthcare admins must enable Allow event-triggered scheduled tasks before members can create them.
Who Can Use Event-Triggered Tasks, and Who Cannot, in 2026?
Eligibility is specific and the exclusions are absolute rather than gradual. OpenAI lists eligible Plus, Pro, Business, Enterprise, Edu and ChatGPT for Healthcare users. Event-triggered tasks are not available on Free or Go, and not available in FedRAMP workspaces.
| Plan or workspace | Event-triggered tasks in 2026 | What to do |
|---|---|---|
| Plus, Pro | Eligible | Create on web or supported mobile. |
| Business | Eligible | Check workspace controls and connected-app permissions. |
| Enterprise, Edu | Eligible, admin gate applies | An admin must enable Allow event-triggered scheduled tasks first. |
| ChatGPT for Healthcare | Eligible, admin gate applies, and not covered by a BAA | Do not use these tasks to transmit, store or process protected health information. |
| Free, Go | Not available | Use scheduled or one-off Work tasks instead where the plan allows. |
| FedRAMP workspaces | Not available | No workaround is documented. |
OpenAI states that in ChatGPT for Healthcare, event-triggered tasks are not covered under a Business Associate Agreement, and that they must not be used to transmit, store or process protected health information. If you market for or work with a healthcare provider, that sentence is the difference between a useful automation and a compliance problem. Build the trigger around non-PHI signals, or do not build it.
How Does Sharing a Scheduled Task Work in 2026?
Where scheduled task sharing is available for your account, you can share an eligible cloud-based scheduled task created in Work. An eligible recipient can review the shared task and schedule a separate copy, and that copy can retain its Work mode and selected model when those options are available to the recipient. Event-triggered tasks can also be shared and scheduled by eligible recipients.
What sharing does not do is the more important half, because it determines what a recipient must set up themselves. The recipient needs their own access to Work, any required model, and any connected apps the task uses.
| Sharing a task does transfer | Sharing a task does not transfer |
|---|---|
| The task instructions, which are included in the shared link | Your chat history |
| The ability to schedule a separate copy | Local files, folders and device access |
| Work mode and selected model, where available to the recipient | Connected app credentials |
| Saved memories and custom instructions | |
| Workspace permissions |
Two consequences follow. First, because the task instructions travel inside the shared link, read them before you share: anything you wrote into the prompt, including internal context or client names, goes with it. Second, tasks that run only on your computer cannot be scheduled from a shared link, so a local-only recipe is not a distributable one.
How Should Marketing and Growth Teams Use Triggered Tasks in 2026?
Point them at high-volume inbound where the first few minutes matter and the judgement is bounded. The three supported events map cleanly onto three different team functions, and the honest fit is triage and preparation rather than autonomous action on a customer.
| Workflow | Trigger to use | Human checkpoint |
|---|---|---|
| Inbound enquiry triage and summarising | New Gmail message, with a condition that filters to the enquiry alias | A person reads the summary before anyone replies to the prospect. |
| Lead qualification prep before a call | New Gmail message from a form notification | Verify the research; never let it reach the client unread. |
| Escalation watch on a client channel | New Slack message in that channel, with @ChatGPT added | Route to an owner. The task flags, the human decides. |
| Standup or weekly digest of channel activity | New Slack message, aggregated, or a schedule instead | Light. This is internal reporting. |
| Marketing site and tracking change review | GitHub pull request activity on the site repository | Engineer review stays mandatory. Treat output as a second opinion. |
| Campaign launch readiness checks | GitHub pull request activity on the landing page repo | Consent, tracking and claims review before traffic. |
| Anything that emails a customer automatically | Do not, without an approval step | Keep the approval requirement in place. Approvals still apply. |
The design rule that keeps this safe in 2026 is simple: let a trigger decide what a person looks at next, not what a customer receives next. Connected-app permissions, workspace controls and approval requirements continue to apply to triggered tasks, and those approvals are the reason a busy automation stays a help rather than a liability.
How this connects to the rest of ChatGPT in 2026
Triggered tasks are a Work feature, so everything that governs Work governs them: the agentic allowance, the workspace role permissions, and the model available to that role. Our guide to Chat, Work and Codex covers the plan and permission layer, and the GPT-6 Astra guide covers the model most of this work will run on. A project keeps related chats, files and instructions together, which is usually the right home for a trigger that needs standing context.
What Are the Common Mistakes With Triggered Tasks in 2026?
- Forgetting to invite ChatGPT to the Slack channel. Add @ChatGPT to every channel the task monitors, or it will not fire.
- Trying to build the trigger on desktop. The desktop app can display existing tasks but cannot create or edit trigger conditions. Use web or supported mobile.
- Leaving the Condition field vague. A broad condition on a busy inbox or channel turns one automation into a flood of runs.
- Assuming an admin gate is already open. Enterprise, Edu and Healthcare admins must enable event-triggered scheduled tasks first.
- Using triggers with protected health information. In ChatGPT for Healthcare these tasks are not covered under a BAA.
- Sharing a link without reading the prompt. Task instructions travel inside the shared link, including any internal context you wrote into them.
- Expecting a shared task to arrive working. The recipient needs their own Work access, model access and connected apps, and gets none of your credentials.
- Letting a trigger act on a customer. Keep approval requirements in place, and remember spoken approval is not supported on web or mobile.
Key Takeaways for 2026
Event-triggered Work tasks are the point where ChatGPT stops being something a person opens and becomes part of a team's plumbing, which makes the boring details, eligibility, conditions and approvals, the ones that decide whether it helps.
- Supported events in 2026 are new Gmail messages, new Slack channel messages, and GitHub pull request activity in an authorised repository.
- Every task is Trigger, Condition and Prompt. The Condition field is the noise filter and deserves the most care.
- Eligible on Plus, Pro, Business, Enterprise, Edu and ChatGPT for Healthcare. Not available on Free or Go, or in FedRAMP workspaces.
- Enterprise, Edu and Healthcare admins must enable Allow event-triggered scheduled tasks before members can create them.
- In ChatGPT for Healthcare these tasks are not covered by a BAA and must not touch protected health information.
- Create and edit on web or supported mobile. Desktop can display tasks but not edit trigger conditions.
- Sharing passes the instructions and the ability to copy the task, but never your credentials, memories, local files or permissions.
- Let triggers decide what a person sees next, not what a customer receives next.
Distk helps growth teams across India and internationally decide which inbound signals deserve an automated first pass, write the conditions that keep the volume sane, and place the approval steps so an automation never speaks to a customer unreviewed. If triggered tasks are on your 2026 roadmap, that design is where we start.
Sources
- OpenAI Help Center, ChatGPT Work and Codex. Every eligibility rule, supported event, admin requirement and sharing behaviour in this guide comes from that article and is OpenAI's own documentation.