Suggestions and interruptions
The assistant mostly waits to be asked. A small, deliberately bounded part of it doesn’t: it can put a suggestion in front of you, and — at most once a day — interrupt you with one.
What this does
Section titled “What this does”Four channels, from quietest to loudest. Each is a different level of claim on your attention, and they are not interchangeable.
| Channel | What it looks like | Cost to you |
|---|---|---|
| Ambient | A quiet chip on a page. | None — it doesn’t move anything. |
| Inline | A suggestion chip in the flow of what you’re doing. | None. Doesn’t compete for the interrupt slot and doesn’t spend the daily budget. |
| Inbox | Collected for you to look at when you choose. | You go to it. |
| Interrupt | A banner. | It comes to you. Capped at one a day. |
Inline suggestion chips
Section titled “Inline suggestion chips”Chips are page-aware, and they read your route, not your data: where you are, and where you just came from. The transition is the useful part — someone who was in Enrollment and is now on a program is doing something specific, so “Who did I just add?” is a question they might actually have, where “How many learners are enrolled?” would be wallpaper by Thursday.
Examples you’ll see: Who did I just add? · Who is on this? · How is it going? · What’s missing? · How many questions? · What needs me?
They’re produced from a fixed table, not a model. That’s deliberate — a model would cost a call on every page load and would phrase them slightly differently each time, so the chips would move under your cursor.
Pressing one asks the assistant that question. Nothing happens until you press.
When Lurno will not interrupt you
Section titled “When Lurno will not interrupt you”An interruption is refused outright — not queued, not deferred — while any of these is true:
- You have a graded attempt open. Same boundary as the learner tutor, for the same reason.
- You’re in focus mode.
- You have an unsaved edit.
- It’s between 20:00 and 08:00 in your local time — quiet hours.
- You’ve already been interrupted today. The cap is one per person per day, and it’s counted per day, not per session, because reloading the page is the thing an interrupted person does most.
If the cap can’t be read for any reason, it’s treated as already spent. Failing towards silence is the only safe direction.
Required legal notices are not suggestions
Section titled “Required legal notices are not suggestions”A consent request and a pending-deletion notice never go through any of this. They are not ranked, not capped, and not suppressible by quiet hours — because a required legal notice that an interruption budget could swallow isn’t a notice. They’re structurally excluded rather than given a high priority, so no configuration change can accidentally silence them.
Make it stop — two different controls
Section titled “Make it stop — two different controls”These look similar and mean different things, and Lurno keeps them separate on purpose.
✕ — “Not now.” A snooze. That category comes back after 14 days. Use it when you meant not right now, not never.
Turn these off — “Never this kind again.” A durable, per-category opt-out. Reversible any time in Account → Notifications.
And one rule you don’t have to find: snoozing the same kind of suggestion three times turns it off by itself. Someone who has said no three times has told you.
Both follow you, not your browser. Dismissing on your laptop holds on your phone, survives clearing your cookies, and works in a private window — because “I told it no and it asked again” is the specific complaint that gets a feature switched off.
Where suggestions come from
Section titled “Where suggestions come from”Nothing here is a model deciding something about a person. Every trigger is arithmetic on data Lurno already holds, so a suggestion can be read aloud to the person it’s about and reproduced tomorrow with the same answer.
- Staff get the counted Needs you band on their home page, and the at-risk learner queue. See Your home page and Integrity & analytics.
- Learners get one thing, in the lesson player: an offer of a hint when they look stuck. It’s never a label about them, never during a graded attempt, and never more than once on the same question. See Help & wellbeing.
Reference
Section titled “Reference”| Setting | Value | Set by |
|---|---|---|
| Proactivity level | off · ambient · inbox · interrupt | Per organization, by Lurno |
| Interruption cap | 1 per person per day | Fixed |
| Quiet hours | 20:00–08:00, local | Fixed |
| Snooze (✕) | 14 days | You |
| Auto-off | After 3 snoozes of the same category | Automatic |
| Durable opt-out | Per category, reversible | You, in Account → Notifications |
| Never suppressed | Consent and pending-deletion notices | Structural |
| Never interrupts during | A graded attempt · focus mode · an unsaved edit | Structural |
Troubleshooting
Section titled “Troubleshooting”I never see any suggestions. Your organization’s proactivity level may be off, or you’ve turned the categories off. Check Account → Notifications.
I dismissed something and it came back. ✕ is a snooze, not an off switch — it returns after 14 days. Use Turn these off for permanent.
I turned suggestions off but the assistant still works. Correct. Turning off suggestions stops Lurno offering; it doesn’t stop you asking.
I turned suggestions off and still see a consent banner. Required legal notices are outside this system entirely and can’t be suppressed.
A colleague gets banners and I don’t. The daily cap is per person, and a banner is refused while you have a graded attempt open, are in focus mode, have an unsaved edit, or it’s outside working hours where you are.
I want to stop my data being sent to AI, not just hide the cards. That’s the AI assistance consent in Account → Consents, not this.