loveholidays scales internal development with OpenAI Codex

TL;DR

loveholidays reports that 79% of its code changes are now AI‑assisted with OpenAI Codex, allowing product managers, designers, and other non‑engineers to prototype, deploy, and maintain software, which has boosted deployment frequency by 73% while keeping headcount flat.

Democratizing Software Development

Key point: Codex lets non‑engineers directly modify code, infrastructure, and data pipelines, eliminating the traditional engineering queue.

Product managers, designers, and commercial teams can prototype customer experiences and make data‑infrastructure changes without waiting for engineering approval.
Dmitri Lerko, Head of Engineering, states that “Everybody is a builder” and that code changes are no longer an engineering‑only activity.
Mike Jones, CTO, frames this as part of building a “general intelligence for travel” that combines the company’s technology with its people’s expertise.

Search Playground: From Idea to Live Feature

Key point: The Search Playground prototype, built with Codex, enables rapid creation of new search experiences by non‑engineers.

Previously, any new customer‑experience idea required engineering prioritization, incurring opportunity cost.
Engineers built the Playground using the company’s design system and Codex, allowing any employee to turn an idea into a working prototype.
Over ten new search experiences have been created; at least three are live on the loveholidays website.
A marketing‑driven microsite for the “Crisps from Abroad” campaign was built in hours, avoiding external agency costs.

Codex‑Guided Data Platform and Infrastructure

Key point: Codex encodes best‑practice workflows, enabling employees to perform data‑platform changes with AI guidance.

Engineers codify validations and best practices into Codex‑driven workflows, so users receive step‑by‑step assistance.
Success rates for AI‑assisted Data Platform changes rose from 58 % to 93 % in one year.
The volume of successful Data Platform changes per support request increased fourfold.
Across broader self‑service workflows, success rates improved from 63 % to 90 %.

Scaling Software Output Without Adding Engineers

Key point: AI‑assisted development has dramatically increased output while headcount remains flat.

AI‑assisted code changes grew from 7 % to 79 % over the past year.
Deployments rose 73 %, yet engineering headcount stayed roughly constant.
Business‑focused metrics show concrete savings: cloud storage costs reduced by ≈ £36,000 / year and data‑processing waste savings of ≈ £100,000 / year.

Building a Single Control Plane for the Organization

Key point: Codex serves as a unified interface for engineers, data scientists, and business users.

Lerko describes Codex as “a single control plane” that removes the need to teach multiple tools across roles.
This unified interface supports a model where specialist expertise is accessible organization‑wide, and the barrier between idea generation and implementation diminishes.

Implications for the Travel Industry

Key point: By democratizing development, loveholidays can iterate faster on travel‑related features and reduce operational costs.

Faster prototyping leads to more ideas reaching production, potentially improving customer experience and conversion rates.
The reduction in engineering bottlenecks allows the company to focus engineering talent on higher‑impact platform improvements rather than routine tasks.
The approach illustrates a practical pathway toward the “general intelligence for travel” vision, where AI augments human expertise across the business.

All figures and quotations are taken directly from the OpenAI announcement dated 26 August 2026.