LEGO Consumer / Toys via OxfordSM

Gamified L&D with AI character generation

Built a gamified open-world learning framework for LEGO using custom AI datasets and living LoRAs of their main characters — operated as a full AI production pipeline by a single person.

AI discovery & readiness audit AI L&D capability mapping Content & production automation strategies

The challenge

What they needed

LEGO's L&D team wanted something genuinely different. Their characters are some of the most recognisable IP in the world, and they wanted to use them as the backbone of an internal learning experience — not as static illustrations in a slide deck, but as living, interactive elements in a gamified learning environment. The problem was that creating content with LEGO characters at the quality and consistency their brand demands traditionally required their full creative production pipeline, which meant every new piece of learning content needed significant production investment. The ambition was an open-world learning framework where employees could explore, interact, and learn through LEGO's own IP. But the production economics of doing that with traditional methods — animation studios, character artists, render farms — made it prohibitively expensive and impossibly slow for an L&D application.

The approach

How I tackled it

Working through OxfordSM, I built custom LEGO datasets that captured the visual identity of their characters with enough fidelity to pass brand scrutiny. From those datasets, I created living LoRAs — fine-tuned AI models of all the main characters that could generate new content on demand while maintaining brand consistency. These aren't rough approximations; they're production-grade character models that can be placed in new scenarios, poses, and environments without going back to a traditional production pipeline. The learning framework itself is open-world and gamified. Rather than linear course modules, learners navigate an environment where LEGO characters guide, challenge, and interact with them. The critical operational detail is that the entire AI production pipeline runs with a single operator. One person generating what would traditionally require a team of character artists, animators, and production coordinators. That's not a cost-saving exercise — it's a fundamentally different production model.

Results

By the numbers

Person operating full pipeline

Production cost reduction

Content output per cycle

What was delivered

Outcomes

The LEGO engagement is the purest example I have of role collapse in action. A production pipeline that would traditionally need character artists, animators, a producer, and a post-production team — all collapsed into a single operator running an AI-first workflow. That’s not about replacing people for the sake of cost reduction; it’s about making a type of content production viable that simply wasn’t viable before at L&D budgets.

The living LoRAs are the technical heart of the project. LEGO’s character IP is sacred — rightly so — and any AI-generated content featuring those characters needs to be indistinguishable from traditionally produced material. Training custom models to that standard takes care and iteration, but once they’re built, you’ve got a production asset that can generate new character content on demand. It’s the difference between hiring an illustration studio every time you need a new image and having that capability in-house, permanently.

The gamified framework changes how people experience corporate learning. Open-world exploration with characters you recognise and care about is a fundamentally different proposition to clicking through slides. The engagement data backs that up. But what excites me most about this project is what it says about the future of enterprise content production — when a single operator can produce what used to require a team, the constraint stops being budget and starts being imagination.

→ Custom LEGO character datasets achieving brand-approved visual fidelity
→ Living LoRAs of all main characters enabling on-demand content generation
→ Open-world gamified learning framework replacing linear course modules
→ Full AI production pipeline operated by a single person
→ Production economics transformed from prohibitive to sustainable for L&D applications