White paper
Machine-scale. Human-signed.
How commercial pharma organizations will write next, and how to get there first, responsibly.
Greg Johnsen, lead author · Bradford Lee, contributing editor
The argument
Specialty pharma spent two decades perfecting the molecule, then wrapped it in communication written for nobody in particular. The technology to close that gap has arrived: adaptive, personalized language for every patient, prescriber, and payer, at scale, at the moment it matters.
What is missing is not technology. It is a permission structure. Personalization at scale generates thousands of communications no reviewer ever sees, and a review model built to approve artifacts cannot approve instances. Until MLR, privacy, and Legal can say yes to that, the capability stays unapprovable, which is why every “segment of one” pitch of the last decade died at the same gate.
This paper builds that structure by an unusual route: we assembled the strongest case against AI writing from its sharpest critics and converted the harms they identify into operating disciplines: Good Writing Practice, the GxP for the written word (an internal practice standard, not a regulatory designation).
Machine-scale is coming. The only decision is whether it’s human-signed.
Human-signed means under rules you authored, with records you can stand behind, in a voice that is recognizably yours. The alternative is rules nobody wrote, and the voice of everyone at once.
What’s inside
A framework, its instruments, and the climb from where most organizations sit today.
Good Writing Practice: six disciplines at the level of the individual artifact, five at the level of the organization, each one derived from a harm the critics of AI writing actually identified rather than one we invented.
The provenance record behind every asset, the published AI Writing Policy, the data contract behind personalization, and the review architecture that governs generation against review capacity.
Ad hoc, permitted, governed, orchestrated, compounding. Most of commercial pharma sits at stage two, which is the most dangerous place to stand: sanctioned volume, ungoverned trust.
Plus the evidence base: a framework tested against a probability sample of coverage it was not built from, with its failed predictions reported alongside its confirmed ones.
Free to read. No sign-up.
It runs about forty minutes, and it is written for the commercial leader who has to decide whether their organization governs this now or explains it later.
PDF · 37 pages. Written under our AI Writing Policy, and it ships with its own declaration of how it was made.
Good Writing Practice, installed
We help commercial pharma organizations put the practice to work, starting from where they stand today.
The assessment maps your organization against the eleven disciplines and five stages. A pilot proves the economics on your own assets in a quarter.
From there, we install it as a Standards OS your teams run.
We’ll come back with a point of view on your first step.