AI Workflows
Strategy

ICH — the International Council for Harmonisation, which sets global technical requirements for pharmaceuticals for human use — came to me with six global operational and technical documents — the kind where the wrong term carries real consequences — built up over several years. Each document was internally consistent, but together they used different terms for the same ideas. My job was to harmonize the language without changing what any term actually meant.
A terminology problem AI alone couldn’t solve, not with these documents.
This wasn’t marketing copy — it was global operational and technical material, where a dropped nuance can carry real consequences. The inconsistency wasn’t an error, just years of contributions catching up with each other. I could see manually cross-referencing thousands of pages of technical language would take months and a full team — not realistic for the timeline ICH needed.
The Challenge
I needed to harmonize the language across six global operational and technical documents without changing what any term actually meant, in documents where that distinction carries consequences.
The Solution
I designed the entire system for this from the ground up — there was no off-the-shelf tool for reviewing operational terminology at this scale. AI made reviewing possible, but not on its own — this needed someone who could judge the analysis correctly and manage the process end to end. I structured the review in two layers: one pass on the words themselves, a second on the meaning behind them, broken into 59 clusters by document and stage, with guardrails I built into each step to catch hallucination or misinterpretation. That surfaced over 100 findings. Reviewing them in a spreadsheet became unworkable at that volume, so I designed and built a custom review app instead — each finding shown with its context, reviewed one by one, accepted, rejected, or corrected, before coming back to me for a final check.
The Outcome
I delivered six documents recreated with harmonized terminology throughout, across five phases in five months. The project lead later described the work as “highly systematic and thorough,” pointing to the “considerable flexibility and effective communication” needed to address their concerns and navigate the complexity of the subject matter. As the team lead put it: “No consulting team could have done this in this timeline.”