Medical information only works if people understand it — yet pharmaceutical and marketing texts are notoriously dense. Pfizer needed a tool that reviews commercial and medical copy for understandability, improving communication between pharma experts, marketing and sales teams, and the doctors, nurses, procurement teams and patients who read the results.
Medical and regulatory texts are precise but hard to parse — readers struggled to grasp the point quickly.
Doctors, nurses, procurement, patients and marketing each need a different level of clarity from the same source.
Teams had no consistent, measurable definition of "readable", so quality depended on the individual author.
An AI assistant had to improve real texts against readability science — not just rephrase them superficially.
The checker was built around one idea: start from who will read the text, measure it against proven readability science, and give editors an actionable path to fix it — inside their real workflow. These are the moves that made it work.
Document creation begins with selecting a user group. Based on that choice, the relevant document checks for the target audience are preconfigured automatically — and can then be refined to the user’s specific needs.
Choosing a reader group up front preconfigures the right checks — authors start from the audience, not a blank page.
The tool rates texts against key writing principles — clear, simple, logical, engaging and actionable — using established readability measures such as the Gunning Fog Index, Flesch Reading Ease and SMOG Index, integrated into the AI agent through RAG.
Established indexes — Gunning Fog, Flesch, SMOG — ground the tool in science rather than opinion.
After a check, users review scores across the rating indexes; detected issues are listed by type with the number of occurrences per category — turning “hard to read” into a concrete to-do list.
Issues grouped by type and count turn “hard to read” into a concrete to-do list.
Users create, view, update and archive documents and share them for collaboration on medical and commercial copy — the checker lives inside the real workflow, not beside it.
We embedded the checker inside the real document workflow rather than shipping it standalone — more integration work, but editors actually used it.
A dedicated knowledge hub offers best practices, writing guidelines and before/after examples of patient content that went through health-literacy optimization.
Guidelines and before/after examples spread health-literacy practice beyond the tool itself.
Rough user flows became a testable prototype, moderated tests with representatives of the target groups surfaced usability hurdles, and the refined flows went into production as a lean, effective tool.
Moderated tests with real target-group representatives surfaced usability hurdles before the refined flows shipped to production.
