From Short Sessions to Persistent Enterprise AI Agents
OpenAI’s acquisition of Ona marks a shift from short-lived, device-tied assistants toward enterprise AI agents that run securely in the cloud for hours or days, automating complex workflows while remaining under customer control and governance. OpenAI plans to fold Ona’s cloud execution and orchestration stack into Codex, turning it from a session-bound coding assistant into an engine for long-running AI workflow automation. Today, many agents stop working when a laptop sleeps or a browser tab closes; with Ona, Codex agents gain persistent cloud environments that keep running in the background. OpenAI reports that weekly Codex usage has passed five million people, a 400% increase from earlier in 2026, driven by software development, research, analysis and workplace automation. That growth underlines demand for Codex cloud execution that can handle not only quick tasks but also sustained work across the software and knowledge lifecycle.
Why Secure Cloud Execution Is Now Central to Codex
Ona was built to move development from fragile local setups into reproducible cloud environments, and OpenAI is now applying that idea directly to enterprise AI agents. Instead of running agent code on a user’s laptop or inside opaque vendor infrastructure, Ona’s model gives each agent a persistent, isolated cloud workspace. In these environments, agents can run tests, resolve issues, modernize applications and complete multi-step workflows without relying on an open session. Sarah Friar, OpenAI’s CFO, wrote that bringing Ona into Codex will let agents work “over hours or days, not just minutes.” That capability matters as enterprises connect agents to production systems and critical data. Persistent environments give AI agents stable tools, logs and context so they can pause, resume and coordinate work in a reliable way that fits existing engineering and operations practices.
Customer-Controlled Infrastructure for Enterprise AI Agents
At the core of the OpenAI Ona acquisition is a customer-controlled execution model designed for cautious enterprise buyers. Rather than pushing all workloads into OpenAI’s own cloud, Codex agents will be able to run inside the customer’s environment, while OpenAI supplies the models and orchestration that drive the work. That split lets enterprises decide where agents run, which systems they may access, how credentials are scoped, what is logged and when tasks go to humans for review. OpenAI says Ona’s technology has already supported around two million developers moving workloads off local machines and into cloud environments, proving the model at engineering scale. For enterprises moving from proof-of-concept bots to production AI workflow automation, this control over infrastructure, security boundaries and compliance is as important as raw model accuracy, and it is where OpenAI is positioning Codex to compete.
Long-Running AI Workflow Automation and a New Competitive Arena
By pairing Codex with Ona’s orchestration stack, OpenAI is signaling that enterprise AI agents are expected to own entire workflows, not isolated prompts. The company highlights continuous tasks such as automated testing, vulnerability resolution and complex software lifecycle activities as natural fits for long-running agents operating in persistent cloud environments. Users will be able to check progress, refine instructions and review results while agents keep running in the background. That puts OpenAI into more direct competition with established workflow orchestration and automation platforms that already sit inside production pipelines. As enterprises standardize on Codex cloud execution for both code and knowledge work, OpenAI moves closer to offering a full operational stack for AI workflow automation—where secure, persistent, customer-controlled execution becomes the key differentiator rather than the model alone.
