When a Word edit should refuse to run

A reviewer can approve a one-word correction and still receive an unacceptable edit. Replacing the entire paragraph may produce the right final text while filling the document with deletions and insertions that hide what actually changed.

In a contract under negotiation, that revision history is part of the work. The next person needs to understand the correction without comparing two versions of an otherwise unchanged clause.

The text on screen is not the whole document

Earlier insertions and deletions complicate the target. The current wording visible in Word can differ from the text encountered by a search through the document's stored structure. Repeated phrases create another problem. Finding the requested words does not establish which occurrence the reviewer approved.

I implemented target resolution against the current revision text and its context. Missing or ambiguous targets remain explicit failures. The tool does not expand an edit into neighboring text just to make the action succeed.

That boundary matters most when the tempting fallback looks helpful. A looser match may turn a failed operation into a plausible edit, but it also transfers the choice of passage from the reviewer to the algorithm.

A successful edit has more than one record

The changed document, the review pane and the saved decision history must agree. A card disappearing from the pane proves very little if the change missed its target or the decision failed to save.

I rebuilt the finding cards around evidence, actions and history, removing repeated outcomes that made the pane harder to read. Failed saves keep the decision unresolved, with retry beside the action. Verification has to reach the document and the persisted state, including what remains after reopening.

Refusal is part of the interaction

Stopping an uncertain edit is only useful if the reviewer can understand what stopped it. The interface needs to keep the intended correction and its context available for inspection.

The product decision was how much authority to give the tool when certainty runs out. I kept that decision visible across the matching logic, the Word operation and the review interface.