The Twenty Minute Board Cleanup Every Friday

Boards do not go wrong in a week. They go wrong over three months of small neglect, and then someone books a two hour meeting to fix it.

The Friday board cleanup is twenty minutes, done by one person, in the same slot every week. Four passes: close what is finished, kill what is dead, re-owner what has drifted, and flag what has not moved. It works because it is short enough to actually happen and because it catches drift while it is still one card rather than forty.

Who does it and when

One named person, not the whole team. A group cleanup turns into a discussion and takes an hour.

Put it late Friday, or first thing Monday if your Fridays are chaotic. What matters is that it is the same slot every week and it belongs to someone by name. Rotating it monthly is fine and keeps any one person from becoming the only one who understands the board.

The person doing it does not need authority to make every call. Anything genuinely uncertain gets a comment and a question, not a decision.

The four passes

Pass one: close what is done

Go to whatever column holds nearly finished work. For each item, ask whether the thing it describes is actually complete by the team's own definition. If there is no honest column to put it back into, the board probably needs a different number of columns.

If yes, close it. If it is complete apart from one small piece, split the small piece into its own card and close the rest. Half finished cards sitting in a done adjacent column are the most common source of a board that overstates progress, and one of the clearest signs a board has become fiction.

Pass two: kill what is dead

Find anything created more than a couple of months ago that has never moved and has no date.

Most of these are ideas somebody had once. They are not tasks. Delete them or move them to a separate list called "someday", which is honest about what they are. A board holding forty aspirations and twelve real tasks trains people to skim.

Deleting feels risky. It is much less risky than the alternative, which is nobody trusting the board. If an idea matters, it will come back.

Pass three: re-owner what drifted

Look for two patterns. Items with no owner, and items owned by someone who has clearly moved on to other work.

For the unowned, add a name or mark it explicitly as unassigned and raise it in the next review. For the drifted, ask the named owner directly rather than guessing. One message: "This is still against your name, is that right?"

Check leave as part of this pass. An item owned by somebody who is away for the next two weeks is not going to move, and it is better to know that on Friday than to discover it on Wednesday as a surprise.

Pass four: flag what has not moved

Anything untouched for a fortnight gets a flag and a one line comment asking what it is waiting for. Items parked in a review column are the usual suspects, and that queue has its own fixes.

Do not chase it now. The flag is the point. It puts the item in front of its owner and in front of the weekly review, where somebody can decide whether it needs help, a smaller scope, or to be killed next Friday.

The checklist

Print this or keep it as a saved note. Twenty minutes, in order.

• Close anything genuinely finished. Split out any leftover piece.
• Move anything "done except" back to a truthful column.
• Delete or archive items older than two months with no movement and no date.
• Add a single owner to anything unowned, or mark it unassigned out loud.
• Confirm ownership on anything whose owner has clearly moved on.
• Check who is on leave next week and note items that will not move because of it.
• Flag everything untouched for a fortnight with a one line question.
• Write three lines for the team: what closed, what was killed, what is flagged.

Those last three lines are what makes the routine visible. Without them it looks like nothing happened, and invisible maintenance gets dropped first when the week is busy.

Where the tooling helps

Most of this is judgement and a person doing it consistently. The parts that benefit from software are the boring lookups.

SicherOne keeps project records and HR records on one set of data, so the leave check in pass three is something the board already knows rather than a second system to open. Its AI agents work against those same records, and a human approves whatever an agent produces before it goes anywhere, which suits a task like this: draft the list of stale items, let a person decide what actually gets killed.

Twenty minutes a week. It is the cheapest thing you can do to keep a board worth reading.