If you want practical ChatGPT Work tips for agentic AI projects, the place to start is understanding what makes it different from the chat tool you already know: Work does not answer questions, it completes tasks. Launched on 9 July 2026, OpenAI describes it as an agent that can take action across apps and files, stay with a project for hours, and turn a goal into finished work, with outputs that include documents, spreadsheets, presentations, reports, and interactive web apps.
That distinction matters. ChatGPT answers. ChatGPT Work acts. It can open a browser, navigate the web visually (not just retrieve HTML), create and move files, and execute a long sequence of steps without pausing for your approval at every turn. The platform runs on GPT-5.6 and has Codex technology built in, according to Digital Applied.
How ChatGPT Work Grew Out of Coding Tools
The tool did not emerge from nowhere. Agentic products like Claude Cowork and ChatGPT Work evolved directly from coding agents that non-programmers started using for general tasks. Following that usage surge, both Anthropic and OpenAI formalised agentic work products for a broader audience.
The consolidation on OpenAI’s side came quickly. On 9 July 2026, Codex merged into a new unified ChatGPT desktop app, at which point the existing standalone app was renamed ChatGPT Classic, according to OpenAI’s launch announcement. The super-app strategy was overseen by Fidji Simo, OpenAI’s chief executive of applications, with assistance from president Greg Brockman, CNBC reported. The merger had the side-effect of removing some features from the original desktop app that have not returned.
Since launch, Work has continued to gain capabilities. On 25 August 2026, OpenAI added webhook-triggered scheduled tasks, shared tasks, flexible limits for free users, and expanded Work’s browser to complete tasks on signed-in websites, with password manager support and confirmation prompts before consequential actions, according to Releasebot, which aggregates OpenAI release notes. Then, on 22 September 2026, OpenAI introduced GPT-6 Sol and GPT-6 Luna in Work and Codex; both models are rolling out to Plus, Pro, Business, Enterprise, and Edu users, while Free and Go users can access Luna in the desktop app, per OpenAI’s release notes.
Voice control has also arrived. ChatGPT Voice is now available in Work and Codex in the desktop app, letting users speak naturally, interrupt mid-task, and ask Voice to start or coordinate work using whatever tools and permissions are active in the selected mode, according to the OpenAI Help Centre release notes. TechCrunch reports a September 2026 mobile rollout extends similar voice-based agentic features to mobile users.
Which Plan You Need: Free to Pro 500
ChatGPT Work is available across all subscription tiers, but capacity varies sharply. Agentic operations consume significant compute, and free-tier access is intentionally limited. The ChatGPT Plus plan at $20 per month offers what the author of the original ZDNET piece describes as fairly capable regular usage: roughly five hours of intense Codex work per session, with a monthly capacity that resets and supports around 20 hours of agentic coding.
Above that sits a structured set of Pro tiers. According to the OpenAI Help Centre, these run at $100, $200, and $500 per month. The $100 plan provides five times more Codex usage than the $20 Plus level, CNBC reported when it launched on 9 April 2026. The $500 plan includes Ultrafast computing for GPT-6 Astra in Work and Codex, an 8x speed increase in Codex, and a usage allowance 10 times that of the Plus plan, Business Insider reported. Note that as of 10 September 2026, OpenAI temporarily paused new sign-ups and upgrades to the $200 Pro plan; existing subscribers at that tier were not affected.
The scale of the underlying platform gives some context for why capacity is managed this way. As of April 2026, OpenAI’s enterprise update reported that Codex had reached 3 million weekly active users, APIs were processing more than 15 billion tokens per minute, and enterprise revenue had grown to more than 40% of OpenAI’s total revenue.
Ten ChatGPT Work Tips to Get You Started
These ChatGPT Work tips draw from direct experience with the tool and reflect how its permission model and capacity limits work in practice.
1. Feed it small tasks one at a time. Agentic AI performs better when it is not trying to hold an entire complex project in one prompt.
2. Ask for a plan before it executes. For longer projects, get the AI to outline its steps first. You can catch misunderstandings before they cost you capacity.
3. Check your usage before a big job. Go to Settings and then Usage. If capacity is low, the AI will stop mid-run and leave a project half-finished.
4. Never share passwords. Full stop. No exceptions.
5. Back up before any destructive task. If you ask Work to reorganise a folder or rename files, ensure backups exist somewhere the AI cannot reach.
6. Double-check everything. The AI can be wrong. Verify its outputs before acting on them.
7. Expect failure sometimes. Accept imperfection and pace the AI carefully rather than running long autonomous sessions without review.
8. Delegate what you dread. Long, repetitive, tedious research tasks, such as trawling product specs across dozens of Amazon listings, are exactly where Work earns its keep. A task that would take a human an hour or more can finish in around 10 minutes while you make coffee.
9. Use draft mode for email. If you connect Gmail, have Work create drafts rather than send directly. Review every draft before it leaves your outbox.
10. Grant the minimum permissions. Read-only access is relatively safe. Anything involving writing, deleting, sending, or taking action on your behalf carries real risk if the AI misunderstands the task.
Work’s permission behaviour is one area still worth watching. The tool does not always ask for confirmation as often as cautious users might prefer, which makes the folder-scoping feature, restricting a project to a specific directory, an important safeguard to use by default.
With GPT-6 models now rolling into Work and voice control extending to mobile, the capability gap between the free and paid tiers is widening fast. The $20 Plus plan remains the practical entry point for regular use; the question of whether the $500 tier’s Ultrafast speed justifies its price will depend on how much of your working day you are ready to hand over to the agent.