One manuscript, not a meeting.

Ideally one your editors have already read and formed a view on, and preferably one where that view was contested. A read you can only agree with tells you nothing about the instrument.

Our production timings come from reads of 47,000 and 64,000 words. A shorter or longer book is read the same way; we will tell you that its timing is an estimate rather than a measurement.

Your organisation, your imprints and your staff’s seats are configured directly with us. The self-service house workspace is in build — while it is, this step is a conversation and a short setup rather than a signup form.

It is also the step at which we would rather be slow. A seat decides what a person at your house can see, and we would rather set that up with you than guess at it.

Once your seats are set up, your staff can bookmark authorslab.ai/publisher. It asks them to sign in, then opens your house’s Books.

Training is included, not an extra. The people who hold seats learn to work the read before they rely on it — including how to disagree with it, which matters more.

About half an hour for a full novel, and it does not depend on anyone keeping a browser open — the job exists server-side, and completion shows on your Books list. What comes back is described on what you get.

Put the report beside your own editor’s view of the same book and find where they disagree. Then ask us why — every editorial call behind a read is recorded, so “why did it say that” is a question we can trace.

A supplier who cannot tell you why their output said what it said is asking you to trust them. We would rather be checked.

• One manuscript file. Ingest is gated: a file that fails validation is refused with a stated reason rather than loaded in a degraded state.
• The names and email addresses of the people at your house who should hold seats.
• Your imprint structure, if the house has more than one — it is what seats are scoped to.