Never Miss an Escalation Again: Notification Rules

The expensive failure isn't the assistant giving a wrong answer. It happens when the assistant correctly concludes it can't help, passes the conversation to your team, and six hours go by before anyone notices. The notification settings exist to close that gap, and getting them right takes about five minutes.

The failure that actually costs you money

A customer asks something your knowledge base doesn't cover, and the assistant behaves as it should: it admits it can't answer and offers a person. That escalation is now sitting in your inbox. Leave it alone until the afternoon and the customer has had a worse experience than if you ran no automation at all, because they were told help was coming and then left waiting.

Every other category below earns its place, yet this one is why the feature exists at all. What shape the handover itself should take is a separate question, and it gets an answer in AI chatbot against human support.

The eight things we can tell you about

There are eight categories, and each one is switched on or off on its own:

• Human escalation: the assistant has passed a conversation along to your team.
• Appointment activity: a booking was created, rescheduled or cancelled.
• Billing: payments, invoices and changes to your plan.
• Usage limits: your message allowance is nearly gone, or already is.
• Onboarding: setup steps that remain unfinished on a new account.
• Weekly digest: a roundup of the week instead of a single event.
• Integration health: one of your connected channels or stores has stopped working.
• Product news: the things we have released.

Two of those carry weight while the rest are informational. Human escalation is the first. Integration health is the second, and it is the one people forget: a WhatsApp connection that drops quietly never says so in your inbox, it shows up as a week that was quieter than usual.

Right now, or once a day?

A category either fires immediately or gets folded into a daily digest. The defaults have everything on and immediate, which is noisy on purpose: we would rather you dial things down after watching them for a week than never learn that a category existed at all.

One test is worth applying: could someone act on this notification within the hour? A customer waiting on a human passes. A billing receipt does not, and product news clearly does not either. Whatever fails that test belongs in the digest, and moving it there is precisely what keeps the immediate alerts meaningful, since a team fielding forty alerts a day stops reading any of them.

Where the notification actually lands

Notifications can land in four places: email, WhatsApp, web push and the bell in the dashboard. These are not interchangeable, and picking a destination per category matters far more than picking one and applying it across the board.

WhatsApp reaches a person who isn't at a desk, so it fits escalations and doesn't fit product news. Web push only works while the dashboard is open, which suits a team already working the inbox. Email acts as the durable record: good for billing and the weekly digest, poor for anything urgent. The bell is for items you would rather find when you go looking than be interrupted by.

Your company clock

Your timezone is captured at signup, and it deserves a check, because several behaviours hang off it. Non-urgent notifications get held overnight and released in the morning on your company's clock instead of going out at 3am. Broadcasts run on that same clock, so a campaign set for 10am leaves at 10am where your business is, not where the server is.

You will find the setting under company settings, given as a city plus an offset. When a team is spread over several timezones, the company setting governs scheduling, and individual users inherit it unless they configure one of their own.

Why the assistant needs to know your opening hours

The assistant draws on that same clock, and it changes two of its answers. Asked whether you are open right now, it replies correctly rather than reciting your hours and leaving the customer to work out the answer. Beyond your hours it hands the conversation over instead of promising a reply shortly, and that is the promise which breeds complaints once shortly turns out to mean eleven hours.

An assistant that knows the local time where you operate behaves differently from one that does not, and customers pick that up straight away even though the configuration behind it is small. Shaping out-of-hours behaviour is covered further in handling after-hours customer messages.

The set-up we'd recommend

For a small team, this is the starting point we would suggest, and you can tune it after a week of living with it:

• Human escalation: immediate, to WhatsApp. This is the alert that has to interrupt somebody.
• Integration health: immediate, to WhatsApp and email. A dead channel is an emergency you would otherwise catch late.
• Appointment activity: immediate to the dashboard bell if you take bookings, daily digest if you take a great many.
• Usage limits: immediate, to email. A mid-week message shortage is worth hearing about in advance.
• Billing, onboarding, product news: daily digest, to email.
• Weekly digest: on, to email, and genuinely read on a Sunday.

Before you close the page, confirm the timezone is correct. Of every setting on this screen, it is the one that quietly shapes all the others.