Uber rolls ChatGPT Work out to more than 30,000 employees

A pilot becomes a company-wide agent platform

Uber is making ChatGPT Work available to more than 30,000 employees after beginning with a pilot of 1,000 people. OpenAI describes it as Uber’s first company-wide deployment of an agentic platform designed for knowledge work, with support for complex tasks that span multiple steps rather than single conversational answers.

The move is notable for its scale and for the transition from an isolated trial to broad access. A successful pilot can establish demand and surface useful tasks, but enterprise rollout changes the risk profile. More users, departments and data types increase the chance of accidental disclosure, duplicated effort or an agent acting outside the context its designer anticipated.

Connected applications make the agent useful and consequential

ChatGPT Work can connect to Google Drive, Gmail, Google Calendar and Slack. OpenAI says employees can research internal information, prepare meeting briefs, create presentations, analyse data and complete workflows across those applications. The integration turns the product from a writing assistant into an operational layer over everyday company systems.

Each connector should preserve the permissions of the requesting employee and avoid broad service credentials that expose material the user could not normally see. Read access and write access also deserve separate approval. Drafting a brief from authorised documents is different from sending messages, editing shared files or changing calendar commitments.

GPT-6.1 Sol becomes the default capability

Uber will receive access to OpenAI’s portfolio of models, with GPT-6.1 Sol planned as the default in ChatGPT Work. The deployment also includes voice and image capabilities, allowing employees to choose different modes for different tasks.

A default simplifies adoption, but model choice should remain visible in logs and evaluation records. Upgrades can change reasoning, formatting, tool selection and refusal behaviour. Before a new snapshot reaches the whole workforce, Uber will need regression tests for its highest-volume and highest-consequence tasks, including cases where the safest response is to stop and ask.

Internal skills and training shape safe adoption

OpenAI says the rollout will be supported by Uber’s internal AI skills library, approved integrations and dedicated employee training. That combination can encode repeatable procedures and keep staff within supported routes instead of encouraging ad hoc prompt sharing or unreviewed third-party tools.

Skills should have named owners, version histories and retirement dates. Training needs to cover verification, confidentiality and escalation as well as productivity tips. Employees should understand that an agent-generated analysis or presentation may still contain omissions, stale context or invented details, even when the interface is connected to trusted company data.

The next test is measurable organisational value

Uber’s developers will continue using Codex for software work, while other functions including operations, finance, marketing and product gain ChatGPT Work. OpenAI and Uber say they plan to share further information about adoption, use cases and measurable outcomes.

Those measurements should include completed-task quality, time saved, rework, incidents and the distribution of benefits across teams. Licence activation alone is not evidence of transformation. The rollout will be most informative if Uber publishes which workflows improved, what controls were necessary and where employees still need traditional tools or specialist review.

Production checks

For implementation, teams should record the exact model, software, connector and policy versions behind every result. Representative evaluation must include normal work, boundary cases, deliberate misuse and component failure, with outcomes separated by user role, data sensitivity and consequence. A strong average should never hide a serious failure in a critical path. Changes should pass the same regression set before promotion, and operators need a documented fallback that preserves evidence rather than silently broadening access or lowering quality.

Procurement should account for the human work required to review exceptions, maintain controls and investigate incidents. Published performance or economic claims are useful starting points, but production confidence comes from evidence collected in the organisation’s own environment. A named owner, measurable acceptance threshold, cost model and tested shutdown path turn a promising release into a service that can be governed over time. Where decisions affect safety, rights, money or regulated records, accountable professional review remains essential.

Security review should map every identity, credential, network destination, stored artefact and administrative access path involved in the workflow. Logs need enough context to reconstruct who initiated an action, which evidence was consulted, what the system proposed and what ultimately executed. Sensitive prompts and traces should not be retained merely because they are useful for debugging. Retention, redaction and access must follow the underlying data classification, with a practical process for correcting or removing material when policy or law requires it.

Operational monitoring should measure quality and service behaviour together. Latency, throughput and availability matter, but so do unsupported claims, missed constraints, unsafe tool calls and the rate at which people overturn recommendations. Teams should rehearse dependency outages and partial failures before unattended operation. If a model, connector or external service becomes unavailable, the workflow should fail predictably, preserve its state and tell the user what remains unfinished. Clear degradation is safer than an agent improvising around a missing control.

Finally, organisations should publish the scope and limitations of the deployed capability to the people affected by it. Users need to know when automation is involved, how to challenge an outcome and where a human can intervene. That transparency improves feedback and prevents a pilot result from being interpreted as blanket approval for broader autonomy. Expansion should follow demonstrated evidence, not novelty or pressure to match a competitor announcement.