Coverage Is Not a Decision: Put the Pick on the Page
A complete comparison table isn't a recommendation. Why AI-made option menus push editing work downstream, and why the pick still matters on the page.
This article is published in English.
Sources: Coverage is not a decision
Someone drops a five-option memo: pros, cons, risks, a polished comparison grid. Ask which option they are backing. They never name one.
The file is not unfinished in the usual sense. Every cell is filled. AI made that kind of assembly easy. The expensive part, the actual pick, stayed expensive.
The same pattern shows up on a weekend. Ask for a park worth visiting and the reply lists walk time from the station, lawn size, kid-friendliness, shops nearby and a warning about Saturday crowds. Nothing essential seems missing. After two scrolls the phone goes down anyway. There was plenty of material and still no sense of having chosen.
At work the rows are clear. There is no obvious gap. There is also no recommendation. Nowhere does a sentence say "this is the one." Options fill the page. A defended choice does not.
Coverage can feel like progress. Choosing is still a different job.
Why people pass along the whole menu
Generating text got cheap. Leaving every option in costs almost nothing. Dropping one means rereading, writing a reason and standing behind a judgment. Keeping the whole set is quicker and safer.
Covering the field looks thorough. It signals that nothing was overlooked. Cutting looks like risk. A reader can call the result incomplete, biased, underthought or taste posing as analysis.
When it is unclear what belongs, two checks help. For the job this memo has to do, can the decision still be made without this piece? Between A and B, which one matters more right now? Ask the same from the reader's chair. Whatever still fails those tests goes to an appendix or a "later" note.
That is not burying work. It is marking the main path. Cutting also protects the person who has to read.
In many review cultures, missing something feels worse than being slow. So the sender plays it safe. Every candidate remains. Every explanation remains. They can say they surveyed the field. The receiver inherits the field.
They thought they were organizing. What arrived was a lineup.
Blank pages used to force the cut
Before AI, drafting was slow enough that weak options fell away on their own. Try writing a case that depends on a partner you have not spoken to. Open a blank page, type the first premise and stall. Without that conversation the argument will not hold, so the option never reaches the draft. Unwritable ideas stayed off the page. The cut happened before anyone else opened the file.
Now the full set shows up first. The machine lines the options up. Writing got cheaper. Discarding did not. It just became easier to postpone.
AI is good at arrangement: putting options in a row, labeling each one, building a table that looks finished. People still have to say which row the team should act on. That is editing.
Under time pressure the two jobs smear together. A neat table starts to stand in for thinking, even when no one has named the row they believe. Someone still has to discard. If the sender does not, the next person must. Reviewers end up doing the cutting that never happened upstream, before they can even talk about the substance. That cutting used to belong to the person who wrote the file.
Meetings drift the same way. The room has to read before it can decide. People scan the page, comparing rows that never got a ranking. Clarifying questions turn out to be editing questions. Twenty minutes vanish. The conversation ends on what to drop, not on what to do next week. A follow-up appears so the same lineup can be "organized" again.
Blaming readers for lacking filters lets the sender off. Readers usually lack the context behind the ask: the deadline, which premises were already locked, which constraints were hard and which were soft. Something still has to be found. Something still has to be decided.
Customer proposals drift the same way. Three priced plans, each explained, none clearly pushed. On a call the customer asks which plan the team would buy with their own budget. The deck never answered, so the customer invents the ranking. The person who shifted that work rarely notices.
Review without a pick
Return to the A-through-E memo. Begin to note that B's premise is thin. That is real feedback. Then stop. If every letter carries the same weight, review becomes commenting on all five. You opened the file to sharpen one path. You leave having annotated the whole set.
Whether the prose came from a model is secondary. A review needs a claim first: a line that says which option is being pushed.
"Which one are you recommending?"
On a shared screen, with the table open and the cursor near the empty recommendation row, wait for a letter or a name. After a pause the answer arrives: all of them could work.
Delete the half-written note about B. There is no sharp place to aim an opinion. The target is the entire document. Equal weight means nothing to narrow toward. The session cannot begin as a review. It has to begin as editing that should already have happened upstream.
At first it looks like a weak filter on the sender's side. After the question is asked, the diagnosis changes. The organizing work was handed downstream.
A file with no pick can look careful and still be unfinished. That is why a complete-looking document does not read as done. Ask for one recommended row in the chat. Without that line the rest cannot be reviewed. The call closes with a follow-up: put the pick at the top, then look at the table.
The skill that still matters
Do not score a file by length, option count or the presence of a table. Look for keep, decide and cut, and whether those moves are visible on the page. One recommended line at the top often says more than five equal columns. When sending, put that line first so the reader meets a choice before a menu.
Call that skill information editing: choosing what reaches the page and what stays off it. When the choice is visible, the other person can answer yes, no or change this. When it is missing, they have to guess the sender's priorities, what was already discarded and what "please review" is supposed to mean. The conversation starts one layer too late.
Assembled is not the same as decided. A useful decision memo does more than list options; it shows which one should lead. Until the page names the option the sender would stand behind, it is not fit for someone else to review. The test is tiny, and it separates passing along a menu from passing along a real decision under way.
Assembly is no longer what sets work apart. Anyone can produce a comparison that looks finished. What still separates good work is what you keep, what you drop and what you push. Before sending, ask which one you are choosing. If the document does not answer yet, the editing is not finished.