Uber rolls out ChatGPT Work to more than 30,000 employees with GPT-6.1 Sol and enterprise AI agents
Uber and OpenAI announced a company-wide deployment of ChatGPT Work on October 22, 2026, extending access to more than 30,000 employees following a 1,000-person pilot. The rollout places agentic task execution within ChatGPT Enterprise across Uber's knowledge-work functions.
The companies identify GPT-6.1 Sol as the default model, alongside approved integrations with Google Drive, Gmail, Google Calendar and Slack, multimodal features, employee training and an internal AI skills library. Uber's software engineers continue using Codex.
Neither company disclosed the contract value, per-employee pricing, actual activation rate or quantified pilot productivity outcomes. Rollout availability must not be confused with demonstrated adoption or return on investment.
··········
THE OCTOBER 22 ANNOUNCEMENT AND THE SCALE OF THE ROLLOUT.
Both companies announced the rollout on October 22, 2026. Its intended audience exceeds 30,000 employees, compared with 1,000 people in the pilot. This is more than a thirtyfold expansion of potential access, not a measured increase in active usage.
The companies do not provide a final activation date, department-level utilization, pilot engagement rate or identical permission profile for every role.
........
Metric | Confirmed detail |
Announcement | October 22, 2026 |
Pilot | 1,000 employees |
Rollout | >30,000 employees |
Default model | GPT-6.1 Sol |
Contract value | Not disclosed |
........
··········
WHAT CHATGPT WORK DOES INSIDE CHATGPT ENTERPRISE.
ChatGPT Work is an agentic work environment that can coordinate multi-step tasks involving information retrieval, analysis, connected applications and production of finished documents, spreadsheets or presentations. Its cloud computer and browser can support longer workflows where enabled.
A finance user might ask an agent to assemble a variance report from authorized documents, while an operations employee might request a meeting brief combining project files and discussions. These are representative workflows, not claims that Uber has activated every example.
Enterprise permissions and administrator policies govern access and available actions; an agent's ability to propose an action does not itself authorize execution.
........
Capability | ChatGPT Work |
Multi-step tasks | Agentic workflows |
Connected tools | Authorized integrations |
Deliverables | Documents, analysis, presentations |
Governance | Enterprise controls |
........
··········
GPT-6.1 SOL AS THE DEFAULT MODEL.
OpenAI and Uber identify GPT-6.1 Sol as the default model for ChatGPT Work in the rollout. OpenAI's documentation describes a default 500,000-token context window for GPT-6.1 Sol in Work and Codex.
Large context capacity can support long documents and extended tasks, but does not guarantee accurate reasoning or retrieval. No Uber-specific comparative model benchmarks, average token usage or model-level operating costs were disclosed.
........
Characteristic | Disclosure |
Default model | GPT-6.1 Sol |
Context window | 500,000 tokens |
Uber-specific benchmarks | Not published |
Per-task cost | Not disclosed |
........
··········
GOOGLE DRIVE, GMAIL, GOOGLE CALENDAR AND SLACK INTEGRATIONS.
The named integrations allow authorized retrieval of files, email, scheduling context and team discussions. Combining these sources can reduce manual collection of context before creating deliverables.
The companies do not disclose the full connector inventory, permissions for every employee, or which connectors support write operations. Access depends on user rights, connector configuration and organizational policies.
........
Service | Potential use |
Google Drive | Documents and files |
Gmail | Relevant correspondence |
Google Calendar | Meeting context |
Slack | Team discussions |
........
··········
UBER'S INTERNAL AI SKILLS LIBRARY.
Uber is developing an internal AI skills library alongside approved integrations and dedicated employee training. Reusable skills can standardize common task procedures and reduce variability in instructions.
The announcement does not disclose the number of skills, complete architecture, approval process or specific production workflows. The library's existence is confirmed, but detailed implementation is not public.
··········
FINANCE, OPERATIONS, MARKETING AND PRODUCT WORKFLOWS.
Uber names finance, operations, marketing and product as functions expected to benefit. Potential tasks include organizing financial reporting, investigating operational patterns, preparing campaign briefs and consolidating product requirements.
These examples illustrate product capabilities, not verified claims about production processes at Uber. No department-level savings or automation rates were reported, and consequential decisions still require appropriate review and approval.
··········
CODEX REMAINS THE SPECIALIZED TOOL FOR SOFTWARE ENGINEERING.
Uber's engineering teams will continue using Codex for code-oriented work. ChatGPT Work addresses broader knowledge tasks, while Codex specializes in software development, debugging and repository workflows.
The announcement does not indicate that Work replaces Codex, nor does it disclose Uber's engineering adoption figures or productivity benchmarks.
··········
MULTIMODAL CAPABILITIES: VOICE AND IMAGE GENERATION.
The rollout includes access to multimodal capabilities such as voice and image generation. These can support alternative input methods, visual deliverables and selected marketing workflows.
The announcements do not specify access levels by department, complete modality coverage, review rules for generated assets or multimodal usage volumes.
··········
ENTERPRISE SECURITY, GOVERNANCE AND ADMINISTRATIVE CONTROLS.
ChatGPT Work operates within ChatGPT Enterprise, which offers centralized user management, controls for connected applications and tools, and compliance visibility. OpenAI documents administrative restrictions for browser, network and sensitive tool actions and describes review mechanisms for certain actions.
The security posture of an agent depends on granted permissions and the customer's implementation. Uber's detailed settings, full connector policies, retention configuration and deployment architecture are not publicly disclosed.
........
Control | Platform capability |
User access | Centralized administration |
Integrations | Admin-managed permissions |
Sensitive actions | Restrictions and review |
Compliance | Compliance API |
Uber configuration | Not disclosed |
........
··········
DATA PRIVACY AND THE LIMITS OF CONNECTED AGENTS.
Agents working across documents, messages and calendars can encounter confidential business information and personal data. Appropriate controls include permission scoping, output review, auditing and policies for sensitive actions.
The Uber announcement does not publish contractual data-processing provisions, precise data-residency arrangements or all retention settings. General ChatGPT Enterprise security capabilities must not be represented as proof that Uber enabled every optional control.
··········
EMPLOYEE TRAINING AND ORGANIZATIONAL ADOPTION.
Uber confirms dedicated training to support the rollout. Employees need to understand how to specify tasks, verify sources, review outputs and recognize actions that require human approval.
The transition from 1,000 pilot participants to more than 30,000 eligible employees creates change-management challenges. The companies do not disclose training duration, completion targets or verified improvements attributable to training.
··········
PRICING, CONTRACT VALUE AND USAGE COSTS.
The financial terms are undisclosed. Neither company states a negotiated seat price, annual commitment, usage allowance, total contract value or quantified return on investment.
OpenAI documents that resource consumption can vary with task complexity and enterprise administrators can manage spending. Multiplying the announced headcount by a generic public price would not establish Uber's actual contract value.
........
Commercial metric | Disclosure |
Eligible employees | >30,000 |
Per-seat price | Not disclosed |
Total contract | Not disclosed |
Included usage | Not disclosed |
Pilot ROI | Not disclosed |
........
··········
HOW TO MEASURE THE ROLLOUT'S BUSINESS IMPACT.
Useful indicators would include activation rate, recurring usage, completed tasks, accuracy, review time, employee time saved and cost per accepted deliverable. Such metrics distinguish nominal access from effective deployment.
The companies indicate that future updates may address adoption and outcomes, but they have not published a quantified pilot ROI or productivity improvement.
··········
HOW THE DEPLOYMENT RELATES TO UBER'S EXISTING AI SYSTEMS.
Uber previously described OpenAI-powered experiences for drivers, couriers and customers, including an internal governance layer called AI Guard for selected applications. The October 22 announcement instead concerns employee-facing ChatGPT Work within ChatGPT Enterprise.
The official disclosures do not establish that ChatGPT Work shares Uber Assistant's orchestration stack or AI Guard implementation. The relationship is organizational rather than proof of a common technical architecture.
··········
LIMITATIONS AND UNCONFIRMED DETAILS.
The precise activation schedule, usage rates, pilot results, contract value, negotiated pricing, connector permissions, full security configuration and Uber-specific model benchmarks remain undisclosed.
The deployment establishes access to an agentic platform, not independently measured operational benefits. Agents can still produce incorrect results or encounter untrusted content, requiring verification and authorization for consequential actions.
··········
WHAT UBER'S DEPLOYMENT INDICATES ABOUT ENTERPRISE AI ADOPTION.
The announcement marks a shift from limited experimentation to broad availability of agentic systems across a large organization. It combines a frontier default model, enterprise integrations, reusable skills and training while preserving Codex for engineering.
The next meaningful evidence will be actual adoption, validated task quality, time savings, costs and sustained use. Until those outcomes are published, the rollout should be treated as a verified deployment initiative rather than proof of a quantified productivity gain.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org
