79% of Code Changes Are Now AI-Assisted! How loveholidays Became an "Everyone's a Builder" Company with Codex

Hi, it's me, Shiichan! Today I've got a story from OpenAI News about a whole company turning into a team of "builders." There are lots of numbers in this one, so let's dig in!

OpenAI News openai.com

What was announced?

OpenAI News published a customer story about how loveholidays, an online travel agency, is using Codex. loveholidays operates across 8 European markets and processes 60 trillion package combinations every day. The story features Dmitri Lerko, Head of Engineering, and Mike Jones, CTO. It shows how adopting Codex changed the old assumption that only engineers write code.

The story so far

Until now, product changes, infrastructure changes, and code deploys were essentially engineering-only work. As Lerko puts it, "Product changes, infrastructure changes, code deploys — these are no longer just an engineering activity." When product managers, designers, or sales came up with an idea, they needed an engineer to actually build it, which created a bottleneck.

Data platform and infrastructure work had the same problem: it required specialized knowledge, so whenever something went wrong, a specialist engineer had to step in.

What changes

With Codex in place, product managers, designers, and even sales teams can now contribute directly to the codebase. A great example is the "Search Playground," where non-engineers can prototype customer experiences themselves. It has produced more than 10 new search experiences, with 3 or more already live in production. The marketing team even built a microsite for the "Crisps from Abroad" campaign in just a few hours.

Data and infrastructure work changed too: Codex provides workflows with best practices and validation logic baked in, so — in Lerko's words — "self-service users of infra or data needs get expert-level knowledge, 24 hours a day."

Dive Deep

The results show up clearly in the numbers.

• AI-assisted success rate on the data platform: 58% → 93%
• Data platform changes per support request: up 4x
• Self-service infrastructure workflow success rate: 63% → 90%
• Share of code changes that are AI-assisted: up from 7% to 79% over one year
• Deployments: up 73%
• Engineering headcount: roughly unchanged

The business impact is real too: cloud storage costs dropped by £36,000 a year, and wasted data processing was cut by roughly £100,000 a year. CTO Mike Jones says, "Technology is just a vehicle, what matters is the impact it has," and loveholidays now measures AI success by business outcomes rather than adoption rates.

Organizationally, Codex has become a "single control plane" that connects engineers, data scientists, and business teams. With routine work handed off to AI, engineers can spend more time on upstream problem-solving. As Jones puts it, "As we give more of the work to AI, the job elevates — you're no longer just the implementer, you have to get closer to the business problem." loveholidays' vision is to build "general intelligence for travel," democratizing technology and human expertise across the whole organization.

Wrap-up

• OpenAI News published a case study on how loveholidays, an online travel agency, rolled out Codex company-wide.
• Product changes and deploys that used to be engineering-only work are now open to PMs, designers, and sales teams too.
• AI-assisted code changes rose from 7% to 79% in a year, and deployments are up 73%.
• The rollout also cut cloud storage costs by £36,000 a year and wasted data processing by about £100,000 a year.
• Worth a read for teams that want to open up development beyond engineers and want to measure AI impact with real numbers.