An interactive decision desk
Keep the work
under your name.
Choose the part of the process you are working on. See a possible role for a tool, the judgment that stays with you and a concrete check before keeping an output.
Outline / Keep the human decision
Ask for possibilities. Keep the premise.
Make the uncertainty explicit before asking for more words.
Write the question, audience and intended effect yourself. Choose the structure that serves those commitments.
Use a tool to propose alternative orders or identify a missing question. Treat each suggestion as a candidate, not a discovery.
Write one sentence explaining why you kept an option. Remove suggestions that merely repeat the premise.
Try the question, not the promise.
List three possible openings for a fictional essay about a closed railway station. Compare what each opening promises the reader; do not let a tool invent the station’s history.
Original editorial exercise · hypothetical, not a reported test.
Research / Keep the human decision
Use the answer to find a source—not replace it.
A persuasive summary still needs a route back to the original evidence.
Decide which statements need evidence. Open the underlying document and check the exact passage, date and scope.
Use a research tool to locate candidate material or organise notes. Its links are leads until you inspect them.
Keep the claim, source passage and your interpretation separate. Mark an unsupported claim unresolved instead of filling the gap.
Try the question, not the promise.
Suppose a draft says a new editing tool “doubles productivity.” Identify the task, sample, measurement and comparison you would need before using that sentence. No productivity result is asserted here.
Original editorial exercise · hypothetical, not a reported test.
Revise / Keep the human decision
Make one editorial decision at a time.
A smoother sentence is not automatically a more accurate or better one.
Name the problem: rhythm, ambiguity, repetition or unsupported certainty. Decide which meaning and voice must survive.
Request a bounded alternative, then compare it with the saved original. Ask for an explanation you can evaluate.
Read the edit for changed meaning, emphasis and implication. Accept, revise or reject deliberately; retain a recoverable version.
Try the question, not the promise.
Compare “She may have missed the train” with “She missed the train.” The second is shorter, but it removes uncertainty. Decide whether the evidence permits that change.
Original editorial exercise · hypothetical, not a reported test.
Publish / Keep the human decision
Agree on the process before delivering the prose.
A client, editor, journal and retailer may ask different questions.
Read the actual recipient’s current terms and agree on permitted assistance, confidentiality and disclosure before delivery.
Use a tool to organise your own process notes, not to determine whether an agreement permits a use.
Recheck the destination policy and unresolved permissions. Describe the actual assistance accurately; do not use this desk as legal clearance.
Try the question, not the promise.
Draft a private handoff note with three fields: task assisted, human review completed, unresolved questions. The note is a record to discuss, not a universal disclosure formula.
Original editorial exercise · hypothetical, not a reported test.
No manuscript upload, model call or saved personal data. The selection changes an editorial brief, not an AI output.
This is an editorial exercise prepared 19 September 2026, not a legal rule or automated assessment. It sends no manuscript, calls no model and stores no personal data. Use the linked source records to investigate the tool or policy relevant to your work.