Put the label where the feature is
Guidance belongs where the feature is, not in a terms page nobody opens. PeopleMuster puts its labels on the screen itself: who can see this, where it came from, and what to check before acting on it. Each one sits next to the thing it describes.
Five labels the running product prints next to the feature.
Why the placement decides everything
A note only works at the moment of use. Somebody asking how many sick days they get is about to act on the answer. That is the only second in which guidance changes anything.
The same words in a footer are decoration. Read at signup, forgotten by Tuesday, and never present when the decision happens.
Placement is also a fair test of whether you believe your own guidance. A label you are willing to put next to the answer is a real one.
Who can see this
Meeting risk signals carry the words admin only on the panel itself. Leadership reading a signal knows it was never shown to the people in the meeting.
A support ticket tells the person raising it that it is private, visible only to the people who handle support. They read that before they type, which is when it matters.
The roles screen lists a baseline headed Everyone can, and says on the page that those items are not permissions and cannot be removed.
Where it came from
Every AI meeting signal is stamped with the prompt version and the model that produced it. When the prompt changes, older signals keep their older stamp, and the screen says so.
A project with no recent meeting shows the gap in days on the project list, driven by a threshold the tenant sets. The number explains itself.
Integration keys say, on the settings screen, that they are stored in your own database and never shown again in full.
- Visibility, printed on the panel it governs.
- Source and version, stamped on generated output.
- Thresholds shown as numbers, not as colours.
- Storage stated where the secret is typed.
What to check before acting
HashBot, the policy assistant, answers questions from your published policies. Next to the answer box it tells people to confirm anything important against the policy document itself.
That keeps the published policy as the source, and makes the assistant the fast way to find it. The authoring screen shows how many chunks each policy produced, so an administrator can see a document was indexed.
What a good label says
Most labels say nothing useful. A badge that only says new tells a reader what they had already guessed.
A label earns its place by answering a question the reader has right now. Who else sees this. Where did this number come from. What should I check before I act.
It also has to stay true. A label that describes last year's behaviour teaches everyone to ignore labels, and the next one is ignored too.
The same rule applied to a report
Once you accept that guidance belongs next to the thing, the argument reaches past product screens.
A report sent to Slack should carry its period and its source in the message, not in a document somewhere else. A monthly attendance export should say when the device last synced.
PeopleMuster's attendance header shows exactly that: when the terminal sync last ran. A day missing from the end of a month is visible before anyone quotes the total.
Questions people ask
Who can see meeting risk signals?
Administrators only, and the signal panel says so on the screen, so nobody mistakes a private assessment for a shared one.
How does the policy assistant handle important questions?
It answers from your published policies and tells people, next to the answer, to confirm anything important against the policy document itself.
How does a policy become searchable?
Publishing it chunks and indexes the text. The authoring screen shows how many chunks each policy produced, so an administrator can see it worked.