Schematic of document extraction: a page with highlighted clauses beside the fields pulled out of it

Cromwell Press AI Editorial Workflow

A London academic publisher automated copy-editing, fact-checking, and reviewer routing across 180 active titles.

At a glance · Publishing · 2025

What Blackbyrds Digital built, and what changed

Blackbyrds Digital deployed an editorial workflow agent for a London academic publisher producing roughly 280 titles a year across humanities and social sciences. Headline result: 18% to 4% (catalog-deadline slippage).

Published · Updated

Cromwell Press AI Editorial WorkflowOne system
  • Manuscript Submission
  • Editor Assignment
  • Initial Review
Shared data · role-based access3 of the modules shown
Schematic of the shipped system, drawn from the case notes

The challenge

Before

Cromwell Press is a mid-sized London academic publisher producing about 280 titles a year across humanities and social sciences. Editorial workflow was a bottleneck — copy-editing rounds, fact-checking on cited references, peer-reviewer assignment, author correspondence, and production handoff were tracked across a project management tool, several email accounts, and a custom spreadsheet that one production manager had been quietly maintaining for years. Title slippage was endemic, and the firm was missing seasonal catalog deadlines for roughly 18% of its titles.

  • 180 active titles tracked across PM tool, email accounts, and one heroic spreadsheet
  • 18% of titles slipping past seasonal catalog deadlines
  • Copy-editing rounds tracked manually with inconsistent style-guide application
  • Reference fact-checking done by interns sampling rather than checking systematically
  • Peer reviewer assignment by editorial-director recall of who was good at what
  • Author correspondence scattered across editor email accounts
  • Production handoff requiring three meetings and a checklist run by hand
  • Reviewer database growing stale with no recency or response-rate signals
BeforeDisconnected tools
  • 180 active titles tracked across PM tool,…
  • 18% of titles slipping past seasonal…
  • Copy-editing rounds tracked manually with…
  • Reference fact-checking done by interns…
  • Peer reviewer assignment by…
  • Author correspondence scattered across…
Schematic of the starting point, drawn from the case notes

The solution

What we built

We built an editorial workflow agent that owns the production-tracking layer while respecting editorial judgment. Manuscripts move through structured stages — submission, peer review, revisions, copy-editing, fact-checking, typesetting, proof, print. The agent applies the firm's style guide consistently across copy-editing rounds and surfaces ambiguous decisions to the editor rather than guessing. Fact-checking on cited references runs against a structured citation database and flags items that cannot be verified. Reviewer assignment uses a structured fit model (subject expertise, recent response rate, current load, conflict-of-interest history) and surfaces three candidates for editorial choice. Author correspondence is tracked against the manuscript, so any editor picking up the file sees the full history. Production handoff is a generated checklist instead of a meeting. The editorial team retains full creative and strategic authority; the agent removes the production friction.

How the system flows

  1. Manuscript SubmissionEditor AssignmentInitial Review
  2. Peer Reviewer MatchFit ModelThree CandidatesEditorial Choice
  3. Revision RoundAuthor CorrespondenceManuscript Update
  4. Copy-EditingStyle Guide ApplicationAmbiguity Surfaced
  5. Fact-CheckingCitation DatabaseUnverified Items Flagged
  6. Production HandoffGenerated ChecklistSign-off
  7. Catalog Deadline TrackerSlip Risk Alerts
  • Reviewer Database with recency, response rate, and conflict history
  • Editorial judgment preserved on ambiguity and strategic decisions
  • Author correspondence threaded against manuscript record
Schematic of document extraction: a page with highlighted clauses beside the fields pulled out of it
Schematic of document extraction: a page with highlighted clauses beside the fields pulled out of it

Questions about this project.

What problem did Cromwell Press AI Editorial Workflow need to solve?

Cromwell Press is a mid-sized London academic publisher producing about 280 titles a year across humanities and social sciences. Editorial workflow was a bottleneck — copy-editing rounds, fact-checking on cited references, peer-reviewer assignment, author correspondence, and production handoff were tracked across a project management tool, several email accounts, and a custom spreadsheet that one production manager had been quietly maintaining for years. Title slippage was endemic, and the firm was missing seasonal catalog deadlines for roughly 18% of its titles.

What did Blackbyrds Digital build for Cromwell Press AI Editorial Workflow?

We built an editorial workflow agent that owns the production-tracking layer while respecting editorial judgment. Manuscripts move through structured stages — submission, peer review, revisions, copy-editing, fact-checking, typesetting, proof, print. The agent applies the firm's style guide consistently across copy-editing rounds and surfaces ambiguous decisions to the editor rather than guessing. Fact-checking on cited references runs against a structured citation database and flags items that cannot be verified. Reviewer assignment uses a structured fit model (subject expertise, recent response rate, current load, conflict-of-interest history) and surfaces three candidates for editorial choice. Author correspondence is tracked against the manuscript, so any editor picking up the file sees the full history. Production handoff is a generated checklist instead of a meeting. The editorial team retains full creative and strategic authority; the agent removes the production friction.

What changed for Cromwell Press AI Editorial Workflow after launch?

18% to 4%: Catalog-deadline slippage. Catalog-deadline slippage dropped from 18% to 4%. The firm published every Spring catalog title on its committed date for the first time in eight years. -41%: Peer review response time. Peer review response time fell 41% as the fit model surfaced reviewers who were both qualified and likely to respond. Author satisfaction with review quality rose noticeably. +22%: Active titles per editor. Editors now handle 22% more active titles each without quality regression. Time previously lost to production-status chasing went into substantive editorial work.

What kind of project was the Cromwell Press AI Editorial Workflow work?

A Publishing project covering AI & Agents, Custom Software.

Production workflow eating editorial judgment?

We build editorial agents that protect your style and your editors' judgment — they remove production friction, your team owns the creative work.

No retainer lock-in · Month-to-month · Full transparency