Case study · AI development
One founder. Many agents.
Building and running a large travel platform with several AI coding agents in defined roles, and translating it into twelve languages with AI pipelines and human review.
Own productTravel discovery platformGlobal · 12 languages
The challenge
A big product.
A small team.
The platform grew to thousands of pages, more than a hundred API routes and twelve languages. One person was responsible for all of it.
That was possible because most of the code was written by AI coding agents working under clear rules. Translation was handled by AI pipelines with human review. Making that work was mostly a management problem. Each agent needed a defined role, a record of what it changed and limits on what it could touch.
The approach
Clear roles.
Human review.
AI agents are productive when each has a defined scope, protected files and a record of what it changed.
Define the roles.
Assign one lead agent and give the others specific areas of responsibility, with files they must not touch.
Keep a record.
Log every handover and deployment so work can be checked and reversed.
Translate in pipelines.
Translate content with AI and DeepL, protecting placeholders and brand terms, and review before release.
Choose languages with data.
Use analytics to decide which pages to translate next, rather than translating everything.
What changed
Scale without a large team.
A single founder ran a platform of thousands of pages in twelve languages.
- Commits
- 1,700+
- Languages
- 12
- Pages made bilingual through one AI pipeline
- 400+
Faster delivery
Agents worked in parallel within agreed limits.
Controlled quality
Human review and automated gates before anything went live.
Targeted translation
Languages rolled out where data showed demand.
Ross’s role
Own product
Ross set the architecture, the agents’ roles and the review rules, and approved what shipped, as founder of the platform.
Published October 2026 · By Ross Brannigan
A potential collaboration
Put AI to work in your team.
For product and marketing teams adopting AI coding agents or AI translation and wanting control as well as speed.
- An AI development operating model.
- Translation pipelines with review steps.
- Guardrails, logging and quality gates.
Related case studies
More from the work.
- Nothing ships unchecked.A pre-deploy audit that reads every rendered page of a large travel site and blocks releases that add broken links, thin pages or unsupported claims.Read the case study
- From duplicate to definitive.Fixing the canonical, hreflang and redirect problems behind a sharp fall in search visibility on a multilingual travel site.Read the case study
- Links where they count.Using Bing position data to find pages stuck just off the first results, then rebuilding internal linking so thousands of orphaned pages joined the site.Read the case study