The call solved it. The email proved it.

Why verbal agreements decay

Two days after a conversation, two people can recall it differently and both be sincere. A week later neither is sure. A month later, when it matters, there is nothing to consult.

This is not about distrust. It is that a conversation has no fixed form: each participant remembers the part that concerned them, in their own words, and the parts that didn't concern them fade. Written text doesn't do that. Everybody reads the same sentence.

Which has a second benefit that is easy to miss. A written record de-personalises the disagreement. Two people arguing on a call are two people. Two people looking at the same list are two people looking at a list.

The chain

Meeting, minutes, action, follow-up, closed. Five steps, and projects routinely execute the first three and abandon the last two.

Minutes that record discussion are close to worthless. What earns the time is the action: a thing to be done, a person who is doing it, and a date. Anything in the minutes that is not one of those is background, and can be a sentence.

Follow-up means the item appears again next week whether or not anybody raises it. Closure means somebody says it is done, and what proves it.

The step that gets skipped is the fourth, and skipping it produces the pattern from Week 9: an item resolved verbally in one meeting, raised again in the next, with nobody able to say whether it was ever closed or closed and reopened.

The same list under pressure

When a project is behind, a recovery action list appears. It is usually treated as a different kind of document — more urgent, more senior, more visible.

It is the same document. Same columns, same discipline, same failure mode if follow-up is dropped. The only difference is that the consequences of a lapsed action are now measured in weeks of programme rather than in inconvenience.

Which is an argument for running the ordinary version properly, because a project that can't close routine actions in normal conditions won't suddenly be able to close urgent ones under pressure.

Copy the record holder, not the hierarchy

Who goes on the email decides whether it works, and the instinct to copy upwards is usually wrong.

The people who need to be on it are the ones holding the records that disagree: the engineer who reported the quantity, the department whose figure contradicts it. They can resolve it between them, and they are the ones who will have to stand behind whatever is agreed.

Copying a manager who holds neither record does something different. It converts a technical reconciliation into an escalation, which slows it down and makes the two people defensive rather than useful. Escalation is a tool, and it works because it is rare.

So the default distribution is narrow and complete: everybody who owns a piece of the disagreement, and nobody who doesn't.

What an action needs

Three fields carry almost all the weight, and a fourth is worth the space.

A person, not a department. Departments don't do things; people do, and an action assigned to engineering belongs to nobody.

A date, which is what makes follow-up possible at all.

A statement of what done looks like, because half of all disputed closures are two people with different pictures of completion.

And the consequence of not doing it, which is optional and changes behaviour more than the other three combined. An action with a stated cost is harder to leave alone than one without.

System design

An action log with three columns is a wish list. These are the fields that make follow-up possible at all.

Record	Produced by	Required quality	Verified against	Feeds
Written statement of the disagreement	Project controls	Sent before the call, not after it	The records that differ	The conversation
Minutes	The chair	Decisions and actions, not discussion	The meeting	Action log
Action	The meeting	A person, a date, and what done looks like	Follow-up	Closure
Closure	The owner	Who closed it, when, and on what evidence	The evidence	Whether it reappears next week

Column order fixed, new fields to the right. This log will be read by somebody a year from now who was not in the room.

Practical insight

Take the last four sets of minutes from your weekly meeting and find every action in them. Then check each one: is there a named person, a date, and any record of closure?

Count how many are still open, how many are closed, and how many you can't tell. That third number is the one that matters, because an action nobody can categorise was never really an action.

Then take the oldest open one and close it this week, properly, with a name against the closure. It is a small thing and it demonstrates the mechanism better than any explanation of it.

Key takeaways

• Write it down before you call. The call is for solving, the email is the record of how it was solved.
• Apply the same rule when the mistake is yours. That is what makes it a habit rather than a defence.
• Two days after a conversation, two people can recall it differently and both be sincere.
• A written record de-personalises a disagreement: everybody reads the same sentence.
• Minutes that record discussion are worth little. Actions with owners and dates are the output.
• Follow-up is the step that gets skipped, and skipping it is why items reappear in every meeting.
• A recovery action list is the same document under pressure, with the same failure mode.
• An action needs a person, a date, and a picture of what done looks like. A stated consequence changes behaviour most.

Records born here. Minutes · action log · decision log · recovery action list when the project is behind.

What is coming next

Actions and decisions are two of the registers a project runs. There are half a dozen more, and most of them are taught as separate lists.

Next week: the register system — which record feeds which, and which of them anybody actually reads.