Design Pattern · Sep 10, 2026

The Briefing as Home Screen

The home screen is a narrative of what changed since the last review: what the system handled, what failed, ranked developments, one intervention worth making, and an evidence link on every claim.

Under the pillar Zero Dashboard Experiences

Thesis

A person opening an operating system for care should be told what happened since they last looked, not shown a set of tiles from which they can work it out.

The argument

The briefing has four parts and they are in order of consequence, not of module. First, what the system did without a person: routine outreach, low-risk exceptions closed, records prepared. Second, what failed, stated as failure rather than absorbed by a silent retry. Third, ranked developments, each one a claim with an evidence link. Fourth, a single recommended intervention, offered as something to simulate rather than as something to do.

The ordering is the design. A briefing that opens with what went well is marketing. A briefing that opens with everything undifferentiated is a log. What makes it a briefing is that something has taken responsibility for saying which item matters most today and can be asked why.

The hardest constraint is the last one: every sentence must be expandable into the record that produced it. That is what separates a briefing from a generated summary, and it is also what makes the surface expensive to build.

A briefing home screen: a paragraph summarising the period, a ranked list of developments each with an evidence link, and sections for autonomous actions and failures.
Prose leads; the ranked developments follow; each carries an evidence link. In the prototype's synthetic data the opening paragraph reports 184 routine actions completed, 27 low-risk exceptions resolved and ten decisions left for a person.

What a legacy vendor would say

That prose is slower to scan than tiles, that different roles need different home screens, and that a narrative written by a model is a liability the moment one of its sentences is wrong.

The last point is fair, and it decides the design. A wrong tile is a data problem. A wrong sentence is an assertion the organization made. The only credible answer is that every sentence is grounded in a source record and any sentence that cannot be grounded is blocked before a person sees it.

What would settle it

Time to first correct action, measured from opening the surface, against a matched dashboard cohort. Plus an audit of the briefing's claims against source records over a month: the tolerable rate of ungrounded assertions is zero, and a system that cannot demonstrate that has not earned the home screen.

Open questions

What the right review interval is when the underlying reasoning is continuous but the reader is not. Whether the briefing should differ by role, or whether one operating picture with different authority to act is the safer design. What the surface should say on a day when genuinely nothing of consequence happened, given the temptation to manufacture a headline.