OpenAI on Thursday launched ChatGPT Work, a brand new AI agent embedded inside its flagship chatbot that goals to rework ChatGPT from a question-and-answer instrument into an autonomous work platform able to executing advanced, multi-step duties throughout customers' e-mail, calendars, code repositories, and messaging apps.
The product is powered by OpenAI's newest flagship mannequin, GPT-5.6, and is designed to go far past producing textual content. ChatGPT Work can collect context from related apps, recordsdata, and workflows to provide completed paperwork, spreadsheets, shows, stories, and web sites. The agent takes a acknowledged end result, breaks it into smaller steps, and stays with advanced initiatives for hours, finishing them independently.
The launch marks OpenAI's clearest try but to reposition ChatGPT as a office platform fairly than a chatbot — and it arrives at a second of extraordinary monetary significance for the corporate. Final month, OpenAI confidentially submitted a draft S-1 registration assertion to the SEC, initiating what might turn into one of many largest know-how IPOs in historical past, with reported valuations clustering between $730 billion and $852 billion and annualized income that has blown previous $25 billion.
In a brief demonstration and dialog with VentureBeat on Friday, Ty Geri, a product supervisor at OpenAI who helped construct ChatGPT Work, stated the product's mission is to democratize the sort of agentic AI capabilities that OpenAI's inside engineering instrument, Codex, has already demonstrated. "What's actually thrilling is we've seen how a lot Codex has been capable of push the frontier of what we will get performed with these AI instruments, versus simply getting info or solutions or steerage," Geri stated. "Our inside adoption of Codex is actually an exponential curve throughout each single product operate and each single use case."
Why OpenAI constructed a persistent digital machine that works from the seashore
The core architectural guess behind ChatGPT Work is a persistent cloud-based digital machine that runs on OpenAI's servers, all the time out there to the consumer no matter which gadget they occur to be on. That marks a deliberate departure from opponents whose brokers require an area machine to stay powered on and related.
"What's actually thrilling about ChatGPT Work is that it's a digital machine within the cloud that's all the time on for you, and that is out there throughout all of our paid tiers," Geri stated. "All Plus customers are getting this. I believe that's a really distinctive facet of this."
The mobile-first facet of the launch is one thing Geri described as "lacking from the market." He pointed to the flexibility to create a web site on a telephone and share it with collaborators as a very novel functionality. "Websites are new on the whole to Codex. They launched in Codex a few week and a half in the past, however now we're launching additionally in internet and cell. You possibly can create a web site in your telephone on the seashore and share it with your folks," he stated.
ChatGPT Work will roll out starting with Professional, Enterprise, and Edu customers, and can broaden to Plus and Enterprise customers over the following few days. Within the interview, Geri emphasised that the supply of the product to Plus subscribers — not simply premium tiers — is central to OpenAI's technique. "It's accessible to all paid plans, together with Plus customers, which in my view is a extremely huge feat, and actually a part of that OpenAI mission, which is about bringing all this energy to as many individuals," he stated.
How MCP plugins join ChatGPT Work to Slack, Gmail, and GitHub
The product depends on MCP-based plugins to hook up with exterior providers like Gmail, Google Calendar, Slack, and GitHub. When requested whether or not the plugin structure relies on the Mannequin Context Protocol commonplace, Geri confirmed: "These are all based mostly on MCP." He added that connecting a number of Gmail accounts — a frequent consumer request — "is certainly on the roadmap."
The expertise is designed to be action-oriented from the primary interplay. ChatGPT Work affords a customized onboarding circulate that surfaces totally different steered use instances relying on the consumer's position. Geri demonstrated how the system, detecting his position as a product supervisor, instantly steered duties like evaluating AI programs, constructing analysis artifacts, and managing his calendar. "You can begin with a easy process like catch me up on Slack or Groups or learn in the present day's calendar," Geri stated. He described a situation the place the system reviewed his calendar, recognized scheduling conflicts, flagged conferences requiring preparation, after which — on his instruction — declined, accepted, or rescheduled occasions instantly.
Customers may also customise the agent by educating it their writing model, organizing outputs into initiatives, and — in a lighter contact — selecting a digital pet that accompanies them within the interface. The interface additionally introduces a hosted web site function that permits customers to construct and share interactive websites instantly by ChatGPT Work, turning what would sometimes be a static slide deck right into a dynamic, collaborative artifact. "Now we instantly have a collaborative interface that's truly extra thrilling and extra accessible than a slide deck, which has all these formatting restrictions," Geri stated.
Scheduling 10 bug bashes without delay: what agentic productiveness appears like in follow
Geri's personal utilization of ChatGPT Work illustrates the breadth of duties the system can deal with. Within the run-up to the product's launch, he wanted to prepare pre-release testing classes — identified internally as "bug bashes" — throughout dozens of options and staff members.
"I simply come to ChatGPT Work and say, 'Arrange a bug bash for all of the distinct options in ChatGPT Work. Add all of the those who labored on that function,' and it may examine Slack, it may examine GitHub, it may examine Docs, and discover a time that works for the 4 highest contributors to that function," Geri stated. "It went and scheduled 10 bug bashes, all coordinated throughout all these totally different folks. That may have taken me half-hour at the very least."
However Geri pushed again in opposition to the characterization that ChatGPT Work is restricted to rote administrative work. He described utilizing it for analytically advanced duties like figuring out the most important causes of consumer churn for particular product options and producing product options — work he stated would beforehand have taken months. "Issues that we’d have spent three months doing, we will now spend every week doing — and do way more, and make a a lot better product," Geri stated. "Bugs that we’d have discovered three or 4 weeks from now, we will now discover inside two days and repair for our customers."
He additionally described handing off the tedium of product testing itself. "It was that although like essentially the most fascinating a part of my job is like what to check, I’d truly find yourself having to spend most of my job doing the testing, which is like me taking a mouse and like clicking on the identical factor over and over, like 5 instances," Geri stated. "As an alternative, now I can outline what will we need to check, and ChatGPT Work or Codex can truly go check it for me, ship me that bug report, after which we will work on fixing that bug."
What OpenAI says about information privateness when AI reads your Slack and e-mail
When pressed on information privateness issues — provided that ChatGPT Work pulls delicate info from office instruments like Slack, Google Drive, and e-mail — Geri stated privateness "is extremely essential, and a very powerful a part of that is it's all the time within the consumer's management."
He pointed to OpenAI's present enterprise safety infrastructure, noting that "enterprise accounts have ZDR, and customers can all the time choose out of letting their conversations assist enhance future fashions, which many customers do." The remark aligns with assurances OpenAI made when it first launched ChatGPT Enterprise in August 2023, when the corporate wrote in a weblog submit that it does "not prepare on your enterprise information or conversations."
The privateness query carries extra weight now due to the sheer quantity of delicate office information ChatGPT Work is designed to entry. In contrast to a chatbot session the place a consumer voluntarily pastes textual content right into a immediate, ChatGPT Work actively reaches into related programs — studying Slack messages, scanning calendar invites, pulling GitHub commit histories — to assemble context for its duties. That represents a basically totally different information floor space than something OpenAI has provided earlier than, and one which enterprise safety groups will scrutinize fastidiously earlier than granting entry.
ChatGPT Work enters a three-way arms race with Anthropic and Microsoft
ChatGPT Work lands squarely in the midst of what has turn into the defining aggressive battlefield in enterprise AI: the race to construct autonomous office brokers that may transcend producing textual content and really execute duties.
The product arrives months after Anthropic took Claude Cowork out of preview and into basic availability in April, bringing its AI agent to internet and cell platforms aimed toward serving to enterprise customers monitor and handle long-running AI-driven duties from wherever. In the meantime, Microsoft made Copilot Cowork usually out there worldwide on June 16, inbuilt partnership with Anthropic to maneuver past chat and into execution. The three merchandise — ChatGPT Work, Claude Cowork, and Microsoft Copilot Cowork — now compete instantly for the eye of enterprise IT departments and particular person data employees alike.
The convergence is putting. All three merchandise share a remarkably comparable imaginative and prescient: a persistent AI agent working within the cloud that may break advanced duties into steps, connect with office instruments through plugins, and produce completed outputs fairly than simply conversational replies. All three work throughout desktop, internet, and cell.
What distinguishes OpenAI's method is its uncooked shopper distribution benefit. ChatGPT has reached 900 million weekly energetic customers, and OpenAI now has 50 million paying subscribers. Greater than 9 million paying enterprise customers depend on ChatGPT for work, and 92% of Fortune 500 firms now use ChatGPT. By making ChatGPT Work out there to Plus subscribers at $20 a month — not simply Enterprise or Professional prospects — OpenAI is betting that broad accessibility will drive adoption sooner than any competitor can match.
OpenAI's product supervisor says AI is a accomplice, not a substitute — with a caveat
When requested concerning the potential impression on the labor market, Geri was cautious together with his framing. He declined to talk broadly about workforce disruption however provided his private expertise as a product supervisor whose day-to-day work has been considerably reshaped by the instrument.
"My job is to not schedule bug bashes and discover out who contributed to a selected function. That's a process I do in my job, however that's not my job," Geri stated. "My job is to make an incredible product." He described ChatGPT Work as "a accomplice" and "an extension of me, actually not a substitute," including: "All people feels much more productive than earlier than, however can also be virtually working more durable than earlier than, since you get to work on all of the stuff you need to work on versus the drudgery round it."
However Geri was additionally cautious to not decrease the sophistication of the work the agent can deal with. "I additionally don't need to say that it's solely doing mundane duties as a result of, like one thing like hill climbing retention curves on a given function just isn’t mundane. It's truly actually exhausting to do," he stated. The excellence issues. If ChatGPT Work have been merely automating calendar invites and expense stories, it could be a comfort instrument. The truth that Geri describes it compressing three months of analytical product work right into a single week suggests one thing with far larger implications for a way groups are structured and staffed.
An IPO-bound firm wants ChatGPT Work to show enterprise AI can generate income
The timing of ChatGPT Work's launch is unattainable to separate from OpenAI's IPO trajectory. The corporate must exhibit that it may convert its huge shopper consumer base into sturdy enterprise income — a story that turns into considerably extra compelling with a product explicitly designed round skilled workflows.
OpenAI stated it’s producing $2 billion in income per 30 days, rising 4 instances sooner than Alphabet and Meta did at comparable phases, with enterprise now making up greater than 40% of income and on monitor to achieve parity with shopper by the tip of 2026. However OpenAI stays closely loss-making, and the corporate doesn’t count on to achieve profitability till round 2030, with inside projections suggesting losses of $14 billion in 2026 alone.
The aggressive dynamics are unprecedented. Anthropic filed for its personal IPO on June 1 at a $965 billion valuation, organising simultaneous public listings from the 2 most distinguished AI startups in historical past. Whether or not each can maintain their lofty valuations underneath the scrutiny of public market buyers will rely largely on whether or not merchandise like ChatGPT Work and Claude Cowork ship measurable productiveness positive factors to paying enterprise prospects.
The launch additionally caps a product trajectory that started with ChatGPT Enterprise in August 2023, accelerated by the discharge of OpenAI's Operator agent in January 2025, and continued by Operator's deprecation and shutdown on August 31, 2025, when its capabilities have been folded into the ChatGPT agent framework. ChatGPT Work is the consolidation of these efforts right into a single, unified product — one which pairs GPT-5.6's three mannequin variants (Sol for energy, Luna for pace, and Terra for balanced on a regular basis use) with a persistent cloud setting and an increasing library of MCP plugins.
The way forward for work might already be working within the cloud
When requested whether or not ChatGPT Work alerts a shift towards a brand new sort of working system — one the place customers work together with their computer systems primarily by an AI agent fairly than by conventional mouse-and-keyboard interfaces — Geri stopped wanting making sweeping predictions. However he hinted on the path OpenAI sees forward.
"Anyone who has labored with Codex or now ChatGPT Work will notice how thrilling it’s to work together together with your setting and your laptop through the agent," he stated. "Particularly within the desktop app, the place the mannequin has entry to your complete machine and may work together with web sites in your behalf — it's actually capable of be an extension of you and an actual accomplice, and that actually looks like the long run."
On the finish of the interview, Geri circled again to one thing private. "I've by no means loved work as a lot as I’ve within the final month utilizing ChatGPT Work and Codex," he stated — a putting admission from a product supervisor who, till lately, spent a significant share of his days clicking by the identical interface 5 instances in a row simply to see if it could break. OpenAI is now asking 900 million customers to consider that feeling scales. For an organization weeks away from one of many largest public choices in historical past, the reply to that query is value roughly $850 billion.
