Goldman Sachs, State Farm and Uber have offered a glimpse into how AI is helping reshape their operations — from boosting internal productivity within a regulated bank, to supporting employee and agent support at a major insurer and streamlining engineering and product-development workflows across a global technology platform.
OpenAI’s recent blog details from a vendor standpoint how each customer shows no single model for how AI is being deployed in practice.
State Farm: early-stage plans within regulatory guardrails
In February earlier this year, State Farm said it is participating in OpenAI’s Frontier platform to strengthen the capabilities of State Farm agents and employees and to explore new ways to accelerate its technology strategy.
The company also said it intends to implement capabilities over time with “rigorous security and oversight” and that its standards for privacy, security and accountability remain unchanged.
That language lands in a market where regulators are already treating AI as part of core insurance supervision.
In July 2024, the New York State Department of Financial Services told insurers that the use of AI systems and external consumer data in underwriting and pricing must comply with existing insurance laws, and said insurers remain responsible for governance, testing and oversight even when third-party vendors are involved.
In a March 2026 issue brief, the National Association of Insurance Commissioners said AI is already used in underwriting, pricing, claims, fraud detection and utilization management, and that existing state insurance laws apply whether decisions are made by humans, algorithms or third-party vendors.
Uber: engineering infrastructure at production scale
OpenAI also named Uber among customers using Frontier to build, deploy and manage agents company-wide. Uber’s own engineering posts show AI already embedded in internal technical workflows.
In August 2025, Uber said its uReview code-review system analyzes more than 90% of the roughly 65,000 diffs the company generates each week, with engineers marking 75% of its comments as useful and addressing more than 65% of posted comments.
In April 2026, Uber also said AI prototyping was helping ideas that once required weeks of cross-functional coordination become tangible in hours, and said nearly 40% of submissions in a global Uber tech hackathon incorporated a prototyping tool.
Goldman: productivity and compliance in a regulated environment
OpenAI named Goldman Sachs as one of its new customers, which ties in with Goldman’s 2024 annual report, which said it had launched a three-year program to optimize operations, manage non-compensation expenses and increase automation and productivity, including through artificial intelligence.
The same report said employees already had access to a developer copilot and the natural-language GS AI assistant, and that Goldman intended to expand the use of those tools in day-to-day workflows through 2025.
Goldman’s public description of that work emphasizes control as much as capability. In a Goldman Sachs Exchanges interview, George Lee said the GS AI Assistant gives more people across the firm access to leading-edge models in a way that is “safer, more reliable, and more compliant,” language he tied directly to Goldman’s role as a regulated financial institution.
The bank’s public positioning therefore centers less on autonomous deployment than on productivity and workflow support inside a tightly governed environment.
What Frontier is being positioned as
OpenAI’s own description of Frontier helps explain why these customer accounts sit together even as the work differs. The company says Frontier is designed to connect enterprise systems and business context, run agents in production, improve them through evaluation and optimization, and govern them through permissions, controls and auditable actions.
That is broad enough to accommodate Goldman’s compliance-oriented tooling, State Farm’s employee and agent support, and Uber’s engineering workflows without reducing them to a single use case. What the customer disclosures show is not one enterprise AI playbook, but three distinct patterns of AI adoption and exploration developing under the same broader platform push.