What Is ChatGPT Work? OpenAI's New Agent, Explained (2026)
ChatGPT Work is OpenAI’s agent inside ChatGPT, launched July 9, 2026. Instead of prompting it message by message, you give it an outcome — “build a competitive analysis of these five vendors” — and it gathers context from your connected apps and files, breaks the job into steps, works independently for hours, and delivers finished spreadsheets, slides, documents, or interactive web apps. It runs on GPT-5.6, the new flagship model OpenAI released the same day, and it’s usage-metered against your plan allowance.
Here’s what ChatGPT Work actually is, how the agent works, what it costs, and the one structural limit worth understanding before you reorganize your workflow around it.
How the ChatGPT Work agent operates
The loop is different from a normal ChatGPT conversation:
- You describe an outcome, not a prompt. “Turn last quarter’s sales calls into a win/loss report” rather than a series of back-and-forth messages.
- It gathers context. ChatGPT Work pulls from your connected apps and uploaded files. The launch directory covers over 1,400 connectable apps, and you can @-mention a specific app in your request to pull context from it — @HubSpot for deal data, @Slack for a channel’s discussion.
- Plan mode shows you the steps. Before the agent starts, it lays out its plan so you can correct course early instead of discovering a wrong turn three hours in.
- It works independently — for hours. This is the headline capability. The agent researches, drafts, iterates, and assembles without you in the loop for every step. Check-ins and action approvals gate the riskier moves.
- It ships an artifact. The output isn’t a chat reply. It’s a finished spreadsheet, slide deck, document, or interactive web app.
Early testers wired it into HubSpot, Gong, Slack, email, and project management tools. Zapier said an automated lead-triage workflow “helped us identify and hand off seven figures in pipeline every month” — a vendor-supplied launch-day number, but it shows the intended shape: multi-app, multi-hour jobs that used to be a person’s afternoon.
How it’s priced: metered, not flat
ChatGPT Work is not a separate subscription — it’s part of your ChatGPT plan, and tasks consume a variable slice of your plan’s allowance depending on complexity. It’s the same structure Codex uses. There is no published per-task price, so a simple summarization job and a four-hour research-and-build job draw down the same pool at very different rates, and you find out how fast by using it. Enterprise admins get spend controls; individual users get a meter.
If the metering model is the part you care about, we broke it down in detail in ChatGPT Work limits.
Who has it right now
Rollout started July 9 for Pro, Enterprise, and Edu on web and mobile, with Plus and Business following over the next few days. It also anchors OpenAI’s new macOS desktop app, which puts Chat, Work, and Codex side by side and is available on every plan, including Free — the standalone Codex app is merging into it, and OpenAI is beginning to sunset its Atlas browser as part of the same consolidation.
ChatGPT Work vs. prompting ChatGPT normally
| Regular ChatGPT | ChatGPT Work | |
|---|---|---|
| Input | A prompt, refined turn by turn | An outcome |
| Context | What you paste or attach | Connected apps + files, @-mentioned |
| Duration | Seconds per reply | Hours per task |
| Output | Chat text | Finished spreadsheets, slides, docs, web apps |
| Your role | Drive every step | Approve the plan, answer check-ins |
| Cost | Standard plan usage | Metered slice of allowance, varies by task |
The practical difference: regular ChatGPT is a collaborator you steer continuously; ChatGPT Work is a delegate you brief once and review at checkpoints.
What it’s good at — and where it stops
Good at: bounded projects with a clear finish line. Competitive research, data cleanup and analysis, report and deck assembly, turning scattered context (calls, threads, docs) into one structured artifact. The 1,400-connector directory means the context-gathering step is genuinely broad.
Where it stops: ChatGPT Work is session-based and task-based. You kick off each job. It works for hours on the thing you assigned — and then it’s done. It doesn’t sit in the background watching your inbox, your CRM, or your calendar, and it doesn’t fire on its own when something happens. There’s no “every Friday at 4pm” and no “when a deal closes.” Each run also draws down a metered allowance, which makes it expensive to treat as an always-on process even if you were willing to manually restart it forever.
Connector depth varies too: the Gmail connector can now send email, but only on web, without attachments, not in the EU/UK, and with per-message approval — and the Google Calendar connector is read-only. Impressive for a chat product; thin for an operations layer.
For the always-on half of the job: Carly
ChatGPT Work answers “do this big thing for me.” It doesn’t answer “keep doing this thing whenever it needs doing.” That second category — triage every inbound lead, chase unanswered emails after three days, prep a brief before each meeting, send the Friday pipeline report — is trigger-based work, and it’s what Carly is built for.
Carly is an AI executive assistant that runs 24/7 in the cloud on triggers: a schedule, an incoming email, a CRM change, a calendar event. You set it up by conversation — describe the workflow in plain English and Carly builds it, no code. It connects to 200+ tools natively and anything else via your own API key (paste it at dashboard.carlyassistant.com/integrations), and it actually sends email on both Gmail and Outlook, updates your CRM, and manages your calendar rather than stopping at drafts. AI agents start at $35/month, and steps in a workflow that don’t use AI run free and unlimited — so a workflow that mostly moves data and only occasionally thinks stays cheap and predictable.
The two aren’t substitutes: brief ChatGPT Work when you have a one-off project with a deadline; hand Carly the recurring work you never want to kick off again. If you’re comparing the full field, see the best ChatGPT Work alternatives.
Frequently Asked Questions
What is ChatGPT Work in one sentence?
It’s OpenAI’s agent inside ChatGPT (launched July 9, 2026, on GPT-5.6) that takes an outcome instead of a prompt, pulls context from 1,400+ connectable apps, works independently for hours, and delivers finished spreadsheets, slides, documents, or web apps.
How much does ChatGPT Work cost?
There’s no separate price — tasks consume a variable slice of your existing plan’s allowance depending on complexity, the same metering structure Codex uses. OpenAI hasn’t published a per-task rate. Enterprise admins get spend controls. Details in ChatGPT Work limits.
Is ChatGPT Work available on my plan?
Pro, Enterprise, and Edu got it first (web and mobile) on July 9, with Plus and Business rolling out over the following days. The new macOS desktop app — which includes Work alongside Chat and Codex — is available on all plans, including Free.
Can ChatGPT Work run automatically on a schedule or trigger?
No. You start each task, it runs for hours, and it finishes. It doesn’t monitor your apps or fire when something happens. For work that should run on triggers 24/7 — inbox triage, follow-ups, recurring reports — an always-on assistant like Carly is the right shape.
More: ChatGPT Work limits · Best ChatGPT Work alternatives · AI news, July 9 · ChatGPT agent mode · ChatGPT Gmail integration
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
