Which of these actually needs you?
AI triage grades every customer conversation — a priority from 1 to 5, whether it needs human attention, a set of labels and a one-line summary of what happened. The sweep runs on its own every half hour with nothing to configure, so the inbox is already sorted when you open it.
Triage runs on conversations you already have, drawing on your plan’s included credits. There is nothing to set up and no per-conversation fee.
- billing@northvale.co Third time asking about the March statement For you 9:04a
- +1 206 ··· ···· Caller upset about a missed appointment For you 8:51a
- Marcus Reid do you do same-day crowns? Answered 8:32a
- Website visitor what time do you close on Saturdays? Answered 8:20a
- Priya Raman Thanks, that works perfectly Answered 7:58a
Sixty-one conversations overnight; these are the two that need a person. Priority and needs-a-human are separate judgements, because an urgent thing can already be handled.
Or just ask for it
Ask what needs you. Get a short answer.
The grading has already happened — you’re just reading it.
Inbox3 need you, 38 handled
41 open. Three flagged needs-attention: the Kepler refund, an invoice dispute, and a caller who asked for you by name. The other 38 are graded but handled — priority and needs-you are separate questions, so an urgent thing that is under control does not shout at you.
Assign the invoice one to Marcus and label the Kepler thread “refund”.
@Octo what’s priority 1 right now?
On it 🔧(runningtriage_conversations…)
Two. Both unassigned. Five more are priority 2 and already answered. (edited)
✅ 22 replies · last reply 16m ago
give me both
Email octo@agent.omniocto.com · WhatsApp 1-MAN-ASK-OCTO · Slack #omniocto
What it works out
Four things about every conversation.
Not a sentiment score. Four specific, separately useful facts, so the inbox can be sorted by the one you care about this morning.
A priority, 1 to 5
How much this one matters against everything else waiting, so the list has a top rather than just a date order.
Whether it needs a human
A separate flag from priority, because they are different questions. Plenty of urgent things are already handled, and some quiet ones are not.
Labels
What it is about, applied consistently, so you can filter to billing or complaints without having tagged anything by hand.
One line of what happened
Enough to decide whether to open it, which is the actual decision you make forty times a morning.
The sweep runs every thirty minutes on its own. There is nothing to configure and no rules to maintain.
How it fits your morning
It has already run before you look.
Do nothing
That is genuinely the setup step.
No rules, no keyword lists, no training. The sweep grades what is in the inbox, including channels you connected today.
Filter to what needs you
A short list instead of a long one.
Sort by priority, or filter to the ones flagged as needing a human. The rest are still there when you want them.
Handle those, ignore the rest
And let the rest keep being answered.
The conversations that did not need you are still being handled on whatever mode you set. Triage decides what to show you, not what to answer.
What is actually happening
Why it is useful rather than another dashboard.
Every thirty minutes, on its own
An always-on sweep re-grades conversations rather than waiting for you to open something. A thread that turned bad at lunchtime is flagged before you are back at your desk.
always-on · 30-minute sweepZero configuration, deliberately
No rules to write and no keyword lists to keep up to date. The most common way a triage system stops working is that nobody maintained the rules, so there are none.
no rules · no training stepOne scale across every channel
Email, WhatsApp, SMS, web chat and phone calls are all conversations in the same inbox, graded the same way — so a bad call and an angry email sort against each other.
one inbox · 17 channels · one priority scaleSearch that handles your customers’ languages
Full-text search works natively across Chinese, Japanese and pinyin rather than degrading to substring matching, so finding a conversation does not depend on writing in English.
CJK + pinyin native searchOr have it come to you
Breakage and budget alerts, and a digest you can reply to, both respect the quiet hours you set — so being told about something at 3am is a choice rather than a default.
alerts · digest · quiet hoursQuestions
The things people actually ask.
- What do I have to configure?
- Nothing. There are no rules to write, no keyword lists to maintain and no training step. The sweep grades whatever is in the inbox, including channels you connected this morning.
- What does the priority actually mean?
- A number from 1 to 5 for how much this one matters relative to the others, alongside a separate flag for whether a human specifically needs to do something. They are different questions: a routine order update can be urgent, and a rambling complaint can be low priority but still need a person.
- How current is it?
- The sweep runs every thirty minutes, so a conversation that turned bad over lunch is flagged before you get back rather than at the end of the day.
- Does it work across channels?
- Yes. Email, WhatsApp, SMS, web chat and phone calls are all conversations in the same inbox, so a phone call that went badly is graded on the same scale as an angry email.
- Can I still find things myself?
- Yes, and search handles Chinese, Japanese and pinyin natively rather than falling back to substring matching — which matters if your customers do not write in English.
- Will it tell me, or do I have to look?
- Both are available. The grading is there when you open the inbox, and separately you can get a digest and alerts that come to you — those respect quiet hours so a 3am threshold does not wake you.
Keep reading
Related jobs you can hand over.
Triage runs on conversations you already have, drawing on your plan’s included credits. There is nothing to set up and no per-conversation fee.