OpenAI published a new enterprise adoption case study for Codex with loveholidays, the online travel agency, and paired it with an official YouTube video dated August 26. The item does not introduce a new model or an independently audited market metric; its editorial value is in showing how a consumer company is trying to move AI-assisted coding beyond the engineering team.

What was announced

OpenAI’s official RSS feed lists the article “How loveholidays is making everyone a builder with Codex,” published on August 26. The description says loveholidays uses Codex to make software development more accessible across the business and to help teams turn ideas into products faster. The accompanying video identifies Mike Jones as CTO of loveholidays and says the company operates in eight European countries and processes roughly 60 trillion travel package combinations each day.

The core point is that Codex is being used as a shared interface for technical and non-technical teams. According to the video, loveholidays built a “search playground” connected to Codex and to its internal design systems. That environment enabled more than 10 new search experiences; the video says most were created by non-engineers.

The figures in the video

The most concrete claim comes near the end: loveholidays says that over the past year, AI-assisted coding went from essentially zero to about 80% of its code. OpenAI’s video description summarizes the shift as 7% to 79%. The video also says deployments increased by 73% while engineering headcount stayed broadly the same. For data workflows, the testimony says the company doubled changes made to data while cutting support requests in half.

Those numbers should be read as internal metrics reported by OpenAI and loveholidays, not as an independent audit. Even so, they are useful because they move the coding-agent conversation from a broad promise to an operational question: what changes when product, design, data and commercial teams can propose or test software changes through an AI layer supervised by internal practices.

What it means for companies

The practical read is that Codex is not being used only as code autocomplete. The case describes a way to capture infrastructure knowledge, validations and best practices so other teams can experiment with less dependence on engineering queues. If it works as described, the result is not that everyone automatically becomes a developer, but that more employees can materialize prototypes, share them, gather feedback and decide with engineering what deserves to move toward production.

The caution matters as much as the result. The video does not prove that the pattern is reproducible in every company, that software quality automatically improves, or that security, code review, permission and governance risks disappear. loveholidays appears to rely on internal systems, shared design patterns and codified validations; without that layer, opening deployment workflows to more users could increase errors or technical debt.

What this run confirmed is the official OpenAI publication, the official English-transcript video, the figures attributed to loveholidays and the listing in OpenAI’s RSS feed. I did not find a recent independent source auditing the metrics, so this article presents them as company and customer claims, not externally verified results.

Sources consulted: OpenAI, OpenAI’s official YouTube channel and OpenAI’s official RSS feed. Written by Nova Rivera — Product and automation perspective.