Uber Puts Agentic AI on Every Desktop: ChatGPT Work Rolls Out to All 30,000 Employees
On October 22, 2026, Uber finished what was reported to be the largest rollout of an AI agent system by a single company in history. The company gave more than 30,000 employees around the world access to ChatGPT Work. This AI system can complete tasks that involve several steps and different apps without needing a person to guide it through every step. Uber first tested the system with 1,000 employees. Within a few months, it expanded access to about 30 times as many people. This shows other businesses what it looks like to use AI agents across an entire company. Instead of being a small experiment, the system is now part of Uber's everyday work tools. It uses OpenAI's newest model and is available to employees who mainly work with information, documents, and other office tasks.
For company technology leaders thinking about when to introduce AI agents, Uber offers a real example of a very large rollout. It also tests whether the safety rules and management controls that make AI agents different from ordinary chatbots can still work when employees throughout a global company use the system.
From a Test With 1,000 People to a Company-Wide System
Before the October 22 announcement, Uber tested ChatGPT Work with about 1,000 employees. This smaller test gave the company time to check how well the system worked, make sure it connected properly with other apps, and prepare help documents for employees before making it available worldwide.
Expanding the system to 30 times as many people follows a pattern that became common among large businesses using AI in 2026. Companies first run small tests to see whether AI helps employees get more work done. Once they see enough benefits, they quickly make the tools available across the company instead of introducing them slowly over several years. Uber's rollout stood out because it moved from a small test to worldwide use in less than a year. The company also wanted employees in many different departments, including operations, finance, marketing, and product development, to use the same AI tools that were previously available only to the test group.
Uber's chief technology officer, Praveen Neppalli Naga, has helped lead the company's engineering and science plans since joining Uber in 2015. He also spoke about expanding the benefits of AI at HumanX 2026 in San Francisco. Naga described the rollout as an important part of Uber's future, not just another way to improve productivity. In OpenAI's announcement, he said that as AI changes how people work, Uber wants everyone to have tools that help them create new ideas, solve problems, and shape the company's future. He called bringing ChatGPT Work to the whole company an exciting step toward that goal.
How ChatGPT Work Completes Tasks With Several Steps
ChatGPT Work does more than answer questions or search the internet. It is a system that can organize and carry out work across several connected apps. A person tells it what they want to accomplish. The AI breaks that goal into smaller steps, completes those steps using the connected apps, and gives the person finished work. It can do all of this in one session without needing someone to manage every step. OpenAI describes the process in much the same way: give the system a goal, and it plans the steps, works with your connected apps and files, and returns the completed result.
For Uber employees, the connected apps include Google Drive and Slack, along with Gmail and Google Calendar. For example, an employee might ask ChatGPT Work to prepare notes for a Monday operations meeting. The system could check the calendar, find useful documents in Google Drive, read recent Slack conversations to identify decisions, and put everything into one document. It could then return that document ready to use. The employee would not have to open each app and collect the information by hand.
The AI model doing this work is GPT-6.1 Sol, which OpenAI confirmed is the default model for Uber's system. OpenAI introduced GPT-6.1 Sol at DevDay on September 29, 2026. It described the model as an improved version of GPT-6 Sol, especially for difficult coding tasks, using computers, and professional work. The model can handle a context window of up to 1,050,000 tokens. Tokens are small pieces of text that an AI reads and processes. On October 11, 2026, OpenAI increased the default context window from 272,000 to 500,000 tokens. It also stopped charging subscribers extra fees for using long contexts. The full 1.05 million-token limit is still available when needed. That amount is roughly equal to 700 pages of text. In practical terms, the AI can keep an entire project's documents, messages, and data in view while working on one task, without needing to split everything into smaller pieces or repeatedly search for information.
OpenAI offers GPT-6.1 Sol to Plus, Pro, Business, Enterprise, and Edu users. Developers can also use it through the API, which lets other software connect to the model. API pricing is $2 for every million input tokens and $10 for every million output tokens. Uber uses the model through ChatGPT Enterprise, where prices are set through a custom business contract.
What Can an AI Agent Do That a Normal Chatbot Cannot?
Research company Gartner makes an important distinction between real AI agents and what it calls "agent washing." Agent washing happens when companies take ordinary chatbots, automated software, or digital assistants and market them as AI agents even though they cannot truly act on their own. In a June 2025 study, Gartner found that only about 130 of the thousands of companies selling agentic AI products had real abilities to act independently, according to Gartner. That means their systems could actually take meaningful actions instead of only writing text about what should be done.
ChatGPT Work has three important features that help separate it from ordinary chatbots. The first is automatic review. Advanced AI models check important actions involving connected tools before those actions happen. For example, if an AI agent is about to send an email or change a shared Google Drive document, the system can stop and show the planned action for review when it crosses a set confidence threshold. This was described by Forkast News in its analysis of ChatGPT Work. The second feature lets company administrators see what the AI did, when it did it, and which information it used. This creates a record that can be checked later. The third feature gives administrators controls for each workspace, allowing them to decide which plugins, information sources, and AI abilities different employees or teams can use.
These controls matter because a Gartner prediction from May 2026 says that 40% of companies will reduce the responsibilities of their independent AI agents or stop using them by 2027 because of problems with oversight and control. Checking important actions before they happen, keeping records for administrators, and limiting access are ways to address the exact problem Gartner warns about: an AI agent doing something nobody wanted, with no clear way to find out what happened.
Uber is also giving different AI tools to different groups of employees. Software developers and engineers will keep using Codex for programming and software development. ChatGPT Work will mainly support employees doing other kinds of office work, including operations, finance, marketing, and product development. The two tools were built for different jobs. Codex is designed specifically for coding. It can work in separate cloud environments and understand information from software projects. ChatGPT Work is designed to complete tasks across different apps used in everyday office work. Uber's vice president of engineering, Sumanth Sukumar, confirmed that most employees in the original test were not engineers. They used the system for tasks such as studying data, spotting trends, and finding better ways to complete work.
Can Other Companies Do What Uber Has Done?
How quickly a company can move from testing an AI agent to using it everywhere depends mainly on whether the test shows clear, measurable benefits. Those benefits must be strong enough to justify the money and effort needed to set up proper safety rules and oversight. Uber's rollout gives other companies an example, but Uber and OpenAI have not yet shared exact numbers showing how much the system improves productivity. They plan to release information about how many employees use it and what results it produces as the rollout continues.
Uber is not the first large company to introduce AI agents on a major scale in 2026. Commonwealth Bank of Australia reportedly gave about 50,000 employees access to ChatGPT Enterprise. Morgan Stanley says more than 98% of its financial advisor teams use an internal AI assistant built with OpenAI models. Along with Uber's rollout, these examples show that using AI across very large businesses became much more common in 2026. It is no longer something only a few early users are trying. Major employers in different industries are now making AI part of their regular operations.
OpenAI now earns more revenue from business customers than from individual consumers. The company's chief financial officer, Sarah Friar, told investors about this change on August 14, 2026. It happened roughly two quarters earlier than OpenAI had expected. Revenue from business customers grew by 32% in July 2026 alone. According to Friar's announcement, OpenAI had reached an annualized revenue rate of about $40 billion by August 2026. This means its revenue at that point was growing at a pace equal to about $40 billion over a full year. Later reports suggested that strong growth continued through September.
Why OpenAI Wants Large Business Customers Like Uber
Uber's rollout gives OpenAI more than a large customer contract. It also gives OpenAI a well-known example of a company using its technology with 30,000 employees. The announcement came during the same week that OpenAI's biggest competitors were making claims about their own business AI products. Anthropic introduced Claude Cowork in January 2026. Microsoft's Copilot Cowork became generally available on July 9, the same day ChatGPT Work launched. These products compete for the money companies spend on AI agents, according to Forkast's analysis of the market.
Uber's rollout gives OpenAI a public example of a large-scale deployment that neither competing product has publicly claimed to match. For company teams deciding which of the three platforms to buy, seeing a system used by 30,000 office workers provides a different kind of evidence from watching a demonstration or looking at test scores.
OpenAI president and co-founder Greg Brockman explained why the deal matters in the announcement. He said Uber is putting AI to work throughout its business, and giving tens of thousands of employees ChatGPT Work is an important step toward making advanced AI tools more widely available. He added that teams across Uber can now use OpenAI's advanced models and agents to handle harder tasks, connect the tools they use every day, and accomplish more work.
What Gartner's 40% Warning Means for Uber
Gartner senior director analyst Anushree Verma warned in her June 2025 analysis that most AI agent projects were still early experiments or small tests. She said many were driven by excitement about the technology and were often used for the wrong purposes. Gartner predicted that more than 40% of current AI agent projects would be canceled before the end of 2027 because costs were rising, business benefits were unclear, and safety controls were not strong enough. A second Gartner prediction from May 2026 said 40% of companies using independent AI agents would reduce their responsibilities or stop using them after problems with oversight appeared during real-world use.
Uber's rollout is happening under the kinds of pressures Gartner warned about. According to media reports, Uber employees used up the company's entire AI tools budget for 2026 in just four months. In September 2026, Uber announced it would lay off about 3,300 employees. That was roughly 10% of its workforce and its biggest round of job cuts since the COVID-19 pandemic. Uber said it wanted fewer layers of management and a simpler company structure. CEO Dara Khosrowshahi did not say AI caused those layoffs. However, Uber had already cut about 10% of its customer service jobs in July 2026 during a restructuring that reports connected to the company's growing use of AI.
Uber has not said that ChatGPT Work will directly replace any more jobs. The rollout also includes an internal collection of AI learning materials, approved connections to other apps, and special training for employees. These resources are meant to help people learn and use the new system rather than suggest they will be replaced. An important question remains: Can Uber show a clear financial return from giving AI agents to employees across the company? This is exactly the kind of result Gartner says many companies have trouble proving. Uber and OpenAI's planned reports about employee use and results may eventually provide an answer.
For technology leaders at other companies, the most useful part of Uber's story may be the situation surrounding the rollout, not simply the fact that it happened. Uber is introducing AI agents while also changing the size and structure of its workforce, dealing with pressure on its AI budget, and competing with companies developing self-driving vehicles that could one day threaten its main business. These are the kinds of pressures companies face when making real decisions about investing in AI. Uber's choice to introduce the technology on such a large scale, despite these challenges, is a strong sign of how far business adoption of AI has progressed by October 2026.
Frequently Asked Questions
What Is ChatGPT Work, and How Is It Different From ChatGPT Enterprise?
ChatGPT Enterprise is the business environment where employees use OpenAI's tools. It includes privacy protections, security rules, administrator controls, and systems to help companies follow required standards. ChatGPT Work is the AI agent system that works inside that environment. A person gives it a goal, and it carries out the necessary steps using connected apps such as Google Drive, Gmail, Google Calendar, and Slack. ChatGPT Enterprise provides the protected and managed workspace. ChatGPT Work is the tool that actually completes tasks inside it. Company administrators can decide which apps the AI may use, which actions need a person's approval, and what records of the AI's actions administrators can see.
What Does GPT-6.1 Sol Offer Employees Who Work With Information?
One of GPT-6.1 Sol's most useful features for office work is its 1,050,000-token context window. That is large enough to keep roughly 700 pages of text in view during a single task. For example, the model could read a full folder of project documents, a long email conversation, past calendar entries, and messages from a Slack channel before preparing its answer. It would not need to divide the information into smaller groups or repeatedly collect it in separate steps. OpenAI introduced GPT-6.1 Sol at DevDay on September 29, 2026. It placed the model below its leading GPT-6 Astra model as a less expensive option. On October 11, 2026, OpenAI increased the default context window from 272,000 to 500,000 tokens and removed extra charges for subscribers using long contexts.
Does Gartner Think Most AI Agent Projects Will Succeed?
No. Gartner's June 2025 research predicts that more than 40% of current AI agent projects will be canceled by the end of 2027. The reasons include rising costs, unclear financial benefits, and weak safety controls. Another Gartner prediction from May 2026 says 40% of companies using independent AI agents will reduce what those agents are allowed to do or stop using them after problems with oversight appear during real-world use. Gartner's analysts also said that only about 130 of the thousands of companies selling products described as AI agents had systems that could truly act independently. One reason Uber's rollout stands out is that ChatGPT Work includes controls meant to prevent these problems. It can pause important actions for review, give administrators records of what the AI has done, and let administrators control access to tools and information. These are the kinds of protections that projects described in Gartner's warnings often did not have.
How Are Uber's Recent Layoffs Connected to the AI Rollout?
The layoffs and the AI rollout happened one after the other, but Uber has described them as separate decisions. In September 2026, Uber announced plans to cut about 3,300 jobs, or roughly 10% of its workforce. These were its largest layoffs since the pandemic. CEO Dara Khosrowshahi said the goal was to reduce management layers and make the company less complicated. He did not blame AI for the cuts. On October 22, 2026, Uber introduced ChatGPT Work to employees doing office and information-based work across its remaining workforce. The company has not said the system will replace more jobs. It is also providing employee training and an internal library of AI learning materials. Even so, the timing matters. Uber used its entire 2026 AI budget in four months, then reorganized its workforce, and then committed to giving AI agents to office workers across the company. The unanswered question is whether these decisions will lead to clear, measurable improvements in productivity. Uber and OpenAI have said they will share results as employees continue using the system.