AutoScout24 accelerated development 10x with OpenAI's Codex and ChatGPT

OpenAI published a case study on AutoScout24, in which the automotive marketplace shared how it integrated ChatGPT and Codex into daily operations. According to the company, for some projects this reduced the development cycle from 2–3 weeks to 2–3 days and helped scale AI far beyond a single engineering team.

Why AutoScout24 Needed AI

AutoScout24 Group operates a large automotive marketplace in Europe and Canada: over 30 million users per month, more than 2 million listings, and a network of 45,000 dealer partners. At this scale, any development delay quickly becomes a business problem. The team needs to simultaneously maintain multiple brands, develop products for buyers and dealers, and maintain the stability of systems that have already grown complex internal logic.

The company noted that previous point improvements to processes were no longer sufficient. Engineering teams faced pressure from migrations, legacy systems, and a growing demand for new features, while the cost of slow releases continued to rise. Therefore, AutoScout24 decided to use large language models not as a separate experiment, but as a way to rebuild the entire approach to development, testing, and internal interaction between product, data, and engineering teams. It was no longer about accelerating individual tasks, but about attempting to remove systemic friction from the entire product delivery pipeline.

How the Tools Were Implemented

The implementation was rolled out in two layers. First, ChatGPT was deployed across nearly the entire organization, providing access to approximately 2,000 employees and establishing a basic level of AI literacy beyond the engineering department. In parallel, Codex was integrated into the daily processes of those who directly build the product: engineering, product, and data teams. In total, this involved approximately 1,000 employees in product development roles, for whom AI became not a separate tab, but part of routine operations.

Before this, the company spent three months evaluating the tool across different teams and ultimately chose Codex for its convenience, compatibility with existing processes, and measurable impact on productivity and code quality. To ensure implementation didn't depend solely on top-down directives, AutoScout24 assembled a network of AI Champions—internal advocates from various departments. They gather feedback, help translate model capabilities into understandable scenarios, and ensure that AI is integrated into existing processes rather than operating as a separate pilot project.

"Codex has become one of the key tools in our engineering processes, delivering measurable impact on productivity, quality, and speed," — Frederick Kraus, CTO of AutoScout24 Group.

Where Results Were Visible

The company saw the fastest impact in areas where there was previously a lot of manual routine, repetitive operations, and lengthy approval cycles. AI is used not for abstract experiments, but for concrete tasks that daily slow down the release of changes. As a result, the tools help accelerate iterations, relieve experienced developers, and improve consistency across teams without a complete overhaul of architecture or processes. These are areas where the impact is easiest to measure in days, reviewer workload, and decision-making speed.

• Automated pull request reviews
• Large refactoring projects
• Preparation of technical documentation
• Post-incident analysis
• Prototyping ideas outside of engineering teams

According to AutoScout24, timelines were reduced by approximately 10-fold on some projects—from 2–3 weeks to 2–3 days. At the same time, the throughput of engineering teams increased, and code reviews and documentation became less dependent on manual work. The company notes a separate benefit among non-technical employees, who can now test hypotheses and create simple prototypes themselves.

The next step is to more deeply integrate AI into key internal systems and into customer-facing products for buyers and dealers.

What This Means

This case demonstrates that corporate AI is increasingly evaluated not by the number of licenses purchased, but by how much it reduces the path from idea to release. When a company succeeds in combining broad access, embedded processes, and clear implementation owners, ChatGPT and Codex transform from an experiment into a real layer of production infrastructure. This is what distinguishes a working implementation from a fashionable pilot. And the numbers here matter more than the slogans.