How to use AI with client manuscripts without hiding the workflow
Professional use begins with authority and disclosure, not with the upload button. Confirm what the client permits, understand the data flow, and keep human verification visible in the deliverable.
Obtain specific authorization
Do not assume a general confidentiality clause permits third-party AI processing. Explain the purpose, provider, content being processed, retention, expected output, and whether the tool may use content for training.
Minimize what you send
Use only the manuscript content required for the agreed task. Remove unrelated personal data, verify that you selected the correct client project, and avoid pasting confidential prose into unapproved general chat tools.
Review every output
Treat AI findings as unverified leads. Check the cited passages, reject hallucinated relationships, distinguish narrator belief from story fact, and write the client-facing interpretation yourself.
Close the data lifecycle
Record the tool and processing date, retain only what the engagement requires, delete projects according to the agreement, and know how to respond if the client withdraws authorization or asks where their text was processed.
Continue the editor workflow
Review Novilot’s processing boundary
Understand the current data flow and limitations before importing a client-authorized manuscript.
Run a free manuscript preflight