The morning brief that replaced our status meetings

A status meeting is a synchronisation ritual with a bad exchange rate. Six people spend forty minutes so that each of them can learn the two or three things that concern them. The other thirty-odd minutes are the price of the format.

We killed ours and replaced them with a brief. This is a dogfood note about what the format turned out to require, written from running it on ourselves rather than from a template.

The five sections

The morning brief is five sections — what changed, needs attention, likely to happen, should be done, running now — with no more than seven items in total, every number evidence-classed, every line deep-linking to the surface it describes, and a subject line that states the day in one clause.

The order is not stylistic. It follows the loop.

What changed is the state delta since the last brief. It is first because everything after it is only interpretable against it.

Needs attention is the set of things that will get worse without a decision. This is the section people actually open the brief for.

Likely to happen is prediction, and it is labelled as prediction. It sits after attention because a forecast you read before the current problems is a distraction.

Should be done is recommendation with an owner attached. Not a wish list — each line names who and by when.

Running now is the honest tail: work in flight, so that nothing in the four sections above reads as unattended when it is simply in progress.

The cap is the feature

Seven items, across all five sections, total.

We tried it without a cap first. What you get is a digest, and a digest is a newsletter, and a newsletter gets skimmed. The cap forces the composition step to be an act of judgment rather than a query. Something has to be left out, which means something has to be ranked, which means the brief has an opinion.

The cap also creates a useful signal. When the same item survives the cut four days running, that is information about the item. When a section is empty, that is information too, and an empty section renders as empty rather than being padded.

Every number carries its evidence class

This is the rule that took the longest to hold and repaid the most.

A number in the brief is never bare. It renders with what kind of number it is: measured, inferred, or predicted, and where it came from. A predicted figure renders as a range, not a point. A measured figure renders with the moment it was last confirmed.

The immediate effect is that you stop having the "is that real" conversation, which turns out to have been a surprising share of every status meeting we ever ran. The delayed effect is more interesting: when predictions in the brief are ranges with a basis, they can be checked later against what happened, and the brief becomes a record of how well the company reads itself.

For this to work at all, the underlying model has to carry the classification. That is what company state does, and it is why a state model beats a dashboard for anything that composes text about your business.

Deep links, not descriptions

Every line in the brief points at the thing it is about. The account, the mission, the invoice, the candidate. Not a search result — the object.

This sounds like a detail and is the difference between a brief you read and a brief you work from. Without the link, acting on a line means going to find it, which means switching context, which means most lines do not get acted on the morning you read them. With the link, the distance from "this needs a decision" to making the decision is one click, and the decision usually happens.

One composition, three renderings

The same composed brief renders three ways: as email, as an in-product view for the day, and as a short audio brief in the two-to-four minute range.

We expected the audio version to be a novelty. It became the most-used rendering on the days it matters most, because the moment when you have attention for company state is often the moment when you do not have a screen. The constraint audio imposes is also clarifying — a line that does not survive being read aloud was usually a line that was hedging.

The weekly variant carries one extra block, the Win Reel, and it is batched there and only there. Wins that arrive on a Tuesday wait until Friday. Putting them in the daily brief competes for the item cap with things that need decisions, and celebration loses that competition badly. It also loses its effect: a win in a stream of alerts reads as noise.

What it cost and what it did not fix

Two honest notes.

The composition quality is the whole product. A brief assembled by ranking rules produces a technically correct document that nobody trusts, because ranking rules cannot tell a slipping deadline that matters from one that does not. The judgment has to come from a model that has read the state, and it has to be checkable, which is why the evidence classes are non-negotiable.

And the brief did not remove every meeting. It removed the ones whose only content was synchronisation. The meetings that remained are the ones where a decision gets made with several people in the room, and those got better, because everyone arrived having read the same five sections.

The mechanism, including the weekly variant and the audio rendering, is described on run your business.