AI manuscript review
Maven Comms
An AI first-pass reviewer that checks promotional manuscripts against client guidelines, references and claims before they reach an editor, reducing repetitive editorial QA.
Background
A healthcare communications agency reviews promotional manuscripts where every claim has to be supported and every house-style and regulatory rule has to hold. The first pass, the slow and careful proofreading, was all done by hand.
The problem
Editors were manually checking grammar and house style, confirming that each claim was backed by an approved reference, and making sure the required elements were present, on every draft, before the real review even began. It is exacting, repetitive work that has to be right every time.
What we built
A tool that runs that first pass automatically. A draft comes in and five AI passes run in parallel, spelling and grammar, medical, legal, regulatory, and claims against the reference library. It returns an annotated document plus a findings dashboard, where each finding carries the exact quote, a plain-language explanation, a suggested fix, a severity and the reference that supports or contradicts the claim. It is multi-tenant, so each brand keeps its own rules, references and data, isolated at the database level.
How it works
- 1
An editor picks a brand, uploads a draft and chooses which checks to run.
- 2
The file is read and split so every finding can point to an exact place in the document.
- 3
Five passes run in parallel, and each claim is matched to the brand's approved references with vector search.
- 4
Findings are consolidated, any quote the model cannot find verbatim in the draft is dropped, and issues the brand already rejected are suppressed.
- 5
An annotated document and dashboard come back, and when the editor marks the pdf up and re-uploads it, their rejections are remembered for next time.
What it does
Got a job like this?
Tell us the task that eats your team’s week and we’ll tell you if we can automate it.
hello@trdspc.com