The Co-Creation Model

Updated 2026-07-13 · markdown version

The co-creation model is the opposing design philosophy to one-button authoring: AI as the medical writer's instrument, not their replacement. The division of labor is drawn along judgment: the machine does the ledger work — finding substantiating passages, maintaining claim–evidence links, generating annotations, detecting drift, pre-running review checks — while the writer keeps every decision about what to say, to whom, and whether it can be defended.

The model rests on a claim about the nature of intelligence work generally: the deliverable is the residue of thousands of judgments, and the judgments — not the typing — are what clients, reviewers, and regulators are actually paying for. Automating the typing while keeping the judgment makes the expert faster. Automating the judgment while keeping a human to click "approve" produces human-as-the-loop — the model's degenerate twin.

In practice, co-creation tooling has a recognizable shape: it lives where the writer lives (Word, PowerPoint — not a separate portal); it treats claims and evidence as first-class linked objects; it is sourced-only (it cannot cite what isn't in the references — nothing is generated into the evidence layer); and its speed shows up as fewer review rounds rather than more drafts. This is the shape we built Claims in, working with medical writers with decades in the craft — which is also our disclosure: we are partisans of this model, not neutral observers of it.

The strategic case for co-creation over automation in regulated industries reduces to one line: when generation becomes free, verification becomes the product — see the verification asymmetry.