Editorial pre-read
Findings on logic, characters, motivation, language and form are structured before a person deepens the editorial judgement.
AI in Manuscript Work: Publishers, Editors, Agencies
Working with manuscripts professionally is a different problem from writing one. An author knows their own book. You receive other people's texts, often several at once, often under time pressure, and you have to form a defensible picture of each in very little time.
That is where the use lies, and that is also where the limit lies. This page describes both, for all three roles.
Prepare, not decideTextual observation can be structured. The judgement about it stays with you.
What is never automated
We put this section first rather than last, because it is the condition for everything else.
Literary potential is not a checklist. Originality, cultural context, market development and the relationship between an agency and an author cannot be reduced to a score. EPOS does not decide whether a manuscript belongs on a list, whether a sample chapter convinces, or whether a text should be represented.
And it does not improve towards a single ideal. An editor recognises what a text is trying to be, and tells a deliberate quirk from a mistake. That is exactly why the author's voice is not a side issue but the core of the work. A tool that smooths does not take work off your hands, it takes the face off the text.
What remains is the layer underneath: collecting findings, surfacing contradictions, spotting patterns across a whole manuscript. That is the work that costs the most time and where attention drops by page 400.
For publishers
In a publishing workflow, EPOS sits before the editorial assessment, not in its place.
Editing whole manuscripts
Three separate levels keep proofreading, style and substance from coming back mixed into one unreadable list. You choose which level a given project needs.
Handover to Word
Lektorat Pro returns results as a Word file with tracked changes, so existing workflows continue instead of adapting to a new tool.
What EPOS does not do inside a publishing house: list selection, market assessment, positioning. Those decisions stay entirely with the house.
For freelance editors
Four tasks prepare well without making your own work invisible.
Consistency
Contradictory names, attributes, states of knowledge, timelines and motives within the available manuscript context.
Stylistic patterns
Clusters, unclear references and drift in form, without treating every repetition as an error.
Questions of substance
Passages on motivation, causality, tension and character development, collected as material for your own diagnosis.
Documented handover
Word with tracked changes, when the exchange with the author happens there.
The division behind it is simple, and it is the reason this works:
EPOS collects
Findings, patterns and possible contradictions.
You judge
Intent, effect, priority and the right form of feedback.
The author decides
Which change belongs in the work and which quirk stays.
The value of an edit is not in the number of findings. It is in selection, weighting, explanation and relationship. A tool that takes the first layer off you does not make the second redundant, it makes room for it.
For literary agencies
Four questions can be prepared systematically before anyone reads.
Do the synopsis and the manuscript match?
Announced conflict, character development and resolution compared with the text that exists.
Is the sample representative?
Whether voice, perspective and tension hold up across the rest of the manuscript.
Where do contradictions appear?
Findings on what characters know, on timelines, motivation and causality.
Which follow-up questions make sense?
Observations condensed into concrete questions for a human second read or for the conversation with the author.
A useful report names observable passages, not a score. It shows where synopsis and manuscript diverge and which information a character is missing. We explicitly do not recommend automated selection or rejection.
The pilot: one process for all three roles
A meaningful test does not start with the largest manuscript available.
Choose one clear use case
A character check, a stylistic pass or a preparatory analysis. Not everything at once.
Take a work whose history you know
Where existing editorial notes, an assessment or known continuity problems can serve as reference.
Set the criteria first
Which findings are useful, which errors would not be acceptable. Written down beforehand, not after.
Compare against the human reference
Useful findings, false positives, missed problems, rework needed. Plus quality, time spent and the effort of cleaning up afterwards.
Only then scale. A tool fits a workflow only when responsibility, data protection and quality work together.
Frequently asked questions
Does EPOS decide whether a manuscript belongs on our list?
No. EPOS structures textual observations and connections. List selection, market assessment and the editorial decision stay with the house.
Can I narrow the task?
Yes. You do not have to work on language, style and substance at the same time. The three levels are selected individually.
Does this replace my work as a freelance editor?
The first layer is prepared, the second is not. Selection, weighting and explanation are what you are paid for, and they stay.
Can we have manuscripts pre-sorted automatically?
We do not recommend it. A score for literary potential is false precision. What helps is a report that names passages and prepares questions.
How do results reach our workflow?
As a Word file with tracked changes, so the existing exchange continues.
Talk through one concrete workflow
Describe the task, the type of manuscript and the decision that has to stay with your team. Then it can be tested whether EPOS holds at that point. What Lektorat Pro brings and what it costs is on the page for publishers, editors and agencies.