ChatGPT Adds Teams and Shared Tasks With Schedules and Event Triggers

OpenAI made ChatGPT automation a team asset at DevDay 2026: new teams and shared tasks let coworkers organize shared work and jointly manage recurring instructions, instead of every automation living and dying with the individual who set it up. Tasks can run on schedules or respond to events — an arriving email, a Slack message — and company-managed connections let teams operate against shared external accounts rather than a patchwork of personal credentials.

This piece covers what shared tasks change about how AI automation survives inside organizations, how event triggers move ChatGPT from assistant to workflow engine, and what operations leads should set up first. It is written for the people who end up owning team tooling.

Teams and shared tasks: what launched

The unit of organization is the team: a group that can schedule a task together, see it, and edit its recurring instructions jointly. That fixes the quiet failure mode of personal automations — the report that stops generating when its creator changes roles, the monitoring prompt nobody else can modify. Shared ownership makes the automation an organizational object with more than one keyholder.

Scheduling covers the familiar ground of recurring runs, but event triggers are the larger addition: a task can fire when something happens — OpenAI names emails and Slack messages as examples — which converts ChatGPT tasks from a cron-style reporter into a reactive system that meets work as it arrives.

Company-managed connections: the credential fix

Automation in teams has always tripped over accounts: personal connections mean the workflow breaks when its owner offboards, and shared passwords are a security anti-pattern. Company-managed connections let an organization provision the external-account access a team's tasks use, centrally — so the connection to the CRM or the data source belongs to the company, survives personnel changes, and is governed like other corporate access.

For admins this is the difference between tolerating ChatGPT automation and standardizing on it. Access is provisioned once, auditable, and revocable; tasks built on it are portable across the team that shares it. It also gives security teams a single place to answer the question that shared automations otherwise scatter across personal accounts: what external systems can our AI workflows actually reach, and under which identity.

How this fits the DevDay agent stack

Shared tasks slot between two other launches from the same keynote. Below them, dots — always-on agents — handle open-ended, ongoing responsibilities; shared tasks are the structured counterpart, explicit instructions on explicit triggers with team visibility. Above them, ChatGPT Space gives the team a shared workspace where the output of scheduled and event-driven work can land as pages, charts, and updates the whole group sees.

The event triggers also connect to the platform side: DevDay's MCP Events support lets connected apps notify ChatGPT of changes — new tasks, messages, content updates — through webhooks. Together they sketch a pipeline: external systems emit events, tasks react, agents execute, and results land in shared surfaces.

Who benefits, and where to start

The immediate audience is every team running operational cadences through individual accounts today — weekly summaries, inbound triage, monitoring digests. Converting those into shared tasks with company-managed connections upgrades them from personal scripts to durable process. Teams without any AI automation get a low-friction entry: describing a recurring instruction is the entire setup.

A sensible rollout order: inventory existing individual automations first, then recreate the load-bearing ones as shared tasks under managed connections, then experiment with event triggers on one high-volume stream — inbound email triage is the canonical candidate — before wiring anything customer-facing. Keep the first event-triggered tasks read-and-summarize rather than act-and-send until the team has watched a few weeks of runs.

Governance, audit and metering: what to watch

OpenAI announced the mechanics without exhausting the governance detail: how granular task permissions get inside a team, what audit trail event-triggered runs leave, and which trigger sources beyond email and Slack arrive next are the open questions that determine enterprise depth. Watch also how shared tasks meter — tasks run through ChatGPT Work count against plan usage, which the same day's plan restructuring repriced.

The strategic read is straightforward: automation that survives personnel changes is infrastructure, and infrastructure gets budgeted. Teams that map their recurring work onto shared tasks early will find out quickly which of their processes were really just one person's prompt.