£1M+ saved on translation costs.
ROI up 1,900%.

35,000 slides, and every one had to keep its shape.

The Academy’s content estate ran to more than 35,000 slides of technical training material — formatted, sequenced and designed for a specific learning experience. Translating it manually, slide by slide and language by language, was the only model in place, and it didn’t scale.

Every translated slide had to keep its original formatting and layout, or it needed manual rework before it could be used. The content itself was a risk multiplier too: technical terminology, special characters and language-specific quirks could produce something misleading if errors slipped through unnoticed.

Three constraints were non-negotiable:

• Formatting and layout had to survive translation intact, slide for slide
• The transformation was a one-time job across a 35,000-slide estate, not an ongoing service
• Future updates needed to run without heavy reliance on external APIs

A pipeline that translates, formats and voices — end to end.

Our Fractional Head of AI designed and deployed an API-led AI pipeline that automated translation, formatting and voice output at scale. Modular microservices handled extraction of slide content, language detection, translation, layout mapping and rendering back into the original presentation format — enabling parallel processing and traceability across 35,000+ slides.

Formatting was preserved throughout: text segmentation, placeholder mapping and adaptive text-wrapping kept every translated slide presentation-ready, without manual rework. Advanced text-to-speech engines were integrated to generate natural-sounding narration aligned to slide timings and speaker cues, so each language got a full multimedia learning experience, not just translated text.

An AI-assisted validation framework used scenario-based datasets to stress-test terminology handling, special characters and layout overflow before they became live problems. To cut ongoing reliance on external APIs, the team also architected a self-hosted storage and content infrastructure, giving the Academy direct control over its data and the ability to adapt content faster.

£1M saved, ROI at 1,900%.

The project turned a one-off, high-cost translation exercise into a repeatable, language-agnostic model the Academy owns outright. It launched a multilingual product offering that future-proofed the business and expanded its reach into international markets — proof that a well-built AI pipeline can cut costs and expand what a business can sell, at the same time.