Investigator's Brochures, Pre-IND packages and protocols assembled from the program database by a prompt system with a shared rubric and modality overlays — so medical writers edit instead of starting from blank.
Drafts you can argue with, not documents you have to trust. Every section cites its source rows, so a reviewer can disagree in minutes. I gave up fluency for traceability, every time.
— The call I madeAn Investigator's Brochure or a Pre-IND package is weeks of medical-writing time — and most of that time is assembly, not judgement: pulling the pharmacology, the tox summary, the CMC facts out of the program database and arranging them into sections reviewers expect.
Every program re-did it from scratch, and the facts drifted from the database the moment they were copied into a document.
The Investigator's Brochure generator runs 22 canonical prompt rows against the program database, with a shared rubric prepended to every one and overlays for siRNA, monoclonal and bispecific antibodies. Each section cites the rows it drew from. The same architecture powers a 16-section Pre-IND package built in waypointed phases so long runs survive timeouts, plus clinical protocol and RNAi patent drafting.
Six steps from a strategy handoff to an exported document, with a person approving every edit.
First drafts arrive in [hours] instead of [weeks], with every section traceable to its source rows. Writers spend their time on what the document should argue, not on where the numbers came from.
The first IB draft, generated live in a program review, turned a strategy debate into an edit session — which is the whole point.
Prompt rows co-written with medical writing and regulatory; modality overlays reviewed by the scientists who own them. Program database → prompt system → Google Docs, and the same rubric now powers Pre-IND and protocol drafting.
— How it came together