ICH — the International Council for Harmonisation, which sets global technical requirements for pharmaceuticals for human use — needed technical teams around the world to build their own training content with non-technical external support. Before any tool was chosen, my job was to find out which one could actually work inside a scientific and technical environment — then make it usable once they picked one.

A generic vendor tutorial wasn’t going to work.

Off-the-shelf onboarding assumes a generic user at a generic company. ICH’s teams operate under strict internal rules about how AI can be used — restrictions on confidential data, unverified outputs, unsupervised generation. Any training material had to teach the tool and the guardrails at the same time.

The Challenge

ICH needed a way for technical teams around the world to build their own training courses with non-technical external support — but first, they needed to know which AI-assisted authoring tool could actually handle a scientific and technical environment without changing the level of scientific and technical precision of the training content.

The Solution

I evaluated around ten AI-assisted course-authoring tools against ICH’s actual constraints — scientific and technical content, technical teams, strict internal AI rules — and the one I recommended came out as the one they chose. Once it was selected, I evaluated it in depth for that specific audience, then built the entire onboarding system from scratch: a full video series covering course creation, AI-assisted module generation, quiz creation, interactive activities, export workflows, and AI image generation, plus a written guide on how to prompt the tool effectively. I built the onboarding around real examples from their own work instead of the vendor’s generic templates, structured so teams could follow it independently rather than needing an expert at each step. Every part of it had to respect their internal AI usage rules — what could be trusted from an AI output, when human review was required, how to catch a wrong answer before it reached real training material.

The Outcome

Teams across multiple regions can now build their own training courses independently, using a tool and workflow they’d never touched before — without routing back through a central office for repetitive support. Adoption spread on its own, well beyond where I was still directly involved.