How Codex Turned Everyone Into a Developer: Holiday Extras' AI Practice

Holiday Extras is a travel services company with a 44-year history, and recently it did something quite rare for a traditional enterprise: it enabled everyone from the CEO to call center employees to write code and deploy applications using OpenAI Codex. This wasn't an internal experiment confined to the tech department—it was an organizational transformation about "who can build."

Starting with Eight Releases in Two Weeks

In the first two weeks after a new application went live, Holiday Extras completed eight releases, each containing at least three features. That pace would have been almost unimaginable in the past, when it required months of scheduling by the engineering team. According to the video, they could only do this "with Codex."

The company has always had technology in its DNA—they describe themselves as being in the first wave of businesses to go mobile and the first wave to go web. Today their goal has become using AI to do business better and push it further. Codex's role in this goal isn't as a tool to speed up engineers, but as a way to spread building capability across the entire organization.

OpenAI Codex is an AI programming assistant launched by OpenAI, built on large language models (such as the GPT series). It can understand natural language instructions and generate, complete, and explain code, supporting multiple programming languages. Unlike traditional code completion tools, Codex is positioned as "conversational programming," where users can describe requirements in everyday language and the model generates runnable code snippets or even complete application prototypes. It is typically provided as an IDE plugin, command-line tool, or API, aiming to lower the programming barrier so that non-professional developers can quickly realize their ideas. In the Holiday Extras case, Codex was opened up to everyone, precisely leveraging its low barrier and high iteration speed to enable business people to directly build tools that solve their own pain points.

What "Everyone Can Build" Means

The biggest change Codex brought to Holiday Extras is that it gives everyone in the company the ability to build. "Everyone" here is meant literally: from CEO Matthew to every employee in the call center, all have access to Codex.

The value of this lies in position. The people closest to customers often understand best where customer problems lie, but in the past they could only report problems to the engineering team and wait. Now these people can directly tackle the problems they see. The loss of information transfer and scheduling delays within the organization are compressed into a single direct building action.

Of course, letting non-engineers write code requires safety boundaries. Holiday Extras built its own deployment platform called Builders, based on its own technology, to provide anyone with a secure deployment environment. Things employees build with Codex can be deployed directly on this platform.

This platform is the key link that makes the whole model work. Without it,全员 building would only create chaos; with it, the last mile between building and deployment is connected, and employees without technical backgrounds can complete the full loop from idea to launch.

Prototype in Two Days, Production-Ready in Two Weeks

Holiday Extras' app underwent its fifth rewrite. The previous version took more than a year for design, review, testing, and engineering build. This version took only two days for the prototype, and with the engineering team's collaboration to reach production readiness, two weeks in total.

There are two numbers worth noting in this comparison: two days and two weeks. The former is the speed at which an individual completes a prototype with Codex; the latter is the total cycle of the prototype plus engineering polish. The source of speed isn't skipping engineering steps, but reducing the cost of "validating an idea" to an extremely low level—ideas can be run first to see, then you decide whether they're worth investing engineering resources in.

An Accountant Rebuilt the Financial System

The video tells the story of a specific person: a long-serving employee in the finance team who had worked in accounting for decades. Accounting roles are not usually considered creative work, but when he realized he could build, he set out to rebuild some of the company's costly financial systems—and did it better.

The significance of this case isn't how much money it saved, but that it validates the power of the combination of "domain knowledge + building capability." The person who best understands the pain points of the financial system is the finance person themselves, not an engineer who received secondhand requirements through documents. When this person gained building tools, he was fixing problems he faced every day, with a different depth of motivation and understanding.

Domain knowledge refers to the deep experience and understanding an individual accumulates within a specific industry or functional area, including business processes, pain points, rules, and implicit requirements. In traditional software development, domain experts usually need to translate requirements into documents, which engineers then implement—a process prone to information loss and misunderstanding. When domain experts build directly with AI programming tools, they can seamlessly translate their domain knowledge into solutions, reducing intermediate steps and improving fit. Holiday Extras' accountant case is a typical example of domain knowledge combined with building capability: he knew the pain points of the financial system deeply, and when rebuilding the system with Codex he could pinpoint problems precisely and create tools that better fit actual needs.

dream-to-screen: The Distance from Dream to Screen

One of the most moving phrases in the interview was "dream-to-screen time"—the distance from having an idea to seeing the finished product on screen. The executive said this distance has already shortened significantly, and what's most exciting is that it could be compressed to nearly zero: you have an idea in the morning, and by the time you get to the office, something might already be waiting for you.

Holiday Extras and OpenAI have been working together for several years, and they describe it as a two-way relationship: they appreciate OpenAI's openness and collaboration, and they emphasize that they will continue to give feedback. This partnership becomes even more important at the scale of company-wide Codex use—the tool's capability boundaries and the organization's own practices shape each other.

Dream-to-screen time is a metaphor describing the duration from generating an idea to seeing a runnable prototype or product on screen. In traditional software development, this cycle often lasts weeks or even months, involving requirements review, scheduling, design, development, testing, and many other steps. The intervention of AI programming tools greatly compresses the cost of the "validate the idea" stage, allowing an individual to turn a concept into a visualized prototype in a very short time and thereby quickly verify feasibility. The Holiday Extras executive's mention that this time could approach zero precisely emphasizes that AI lowers the building barrier to an extremely low level, making ideas almost immediately visible and thus accelerating the innovation cycle.

Lessons for Traditional Enterprises

The Holiday Extras case offers a narrative different from "AI replaces engineers": AI gives building capability to more people who previously lacked it, while the engineering team's value shifts toward platforms, security, and productionization. The Builders platform was built by engineers, but it serves the whole company.

For organizations with many domain experts but limited technical resources, this model has direct reference value: first build a secure deployment channel, then distribute building tools to the people who best understand the problems. Ideas are never lacking; what's lacking is a path to turn ideas into reality quickly. This company says it has plenty of ideas, and what excites them most now is the speed at which those ideas become reality.

In the joking words of their management, now that they have this capability, the next thing to consider is "how to steal the moon."

Background Supplement

A deployment platform refers to a system used to automate application release, configuration, and environment management, typically including functions such as code building, testing, security scanning, version control, and launch release. In company-wide building scenarios, the role of the deployment platform is especially critical: it provides non-technical employees with a controlled "sandbox" environment, ensuring that the code they generate with AI undergoes necessary security checks, permission controls, and resource isolation before release, avoiding impact on production systems from code quality issues or security vulnerabilities. Holiday Extras' self-built Builders platform plays exactly this role. Based on the company's own technology stack, it standardizes the process between building and deployment, enabling non-engineers to safely complete the loop from code to launch.