Watch Your Assists in Real Time: Usage Alerts Are Here

Minimum Version: Australia Patch 4 Feature: Usage Alerts in AI Admin Hub Role: sn_nowassist_admin

💡 Note: As of Australia Patch 5 or Zurich Patch 12, Now Assist Admin Console (NAAC) has been renamed to AI Admin Hub (AIAH).

Overview

Here's a scenario that may be familiar. An admin notices assist consumption looks higher than expected, digs into it, and discovers one skill has been firing way more than anyone exspected. Maybe it's a misconfigured automation. Maybe it's a user who found a creative way to loop a skill. Either way, assists are consumed, and nobody got a heads-up.

If you've been waiting for an answer to that concern, it's here. Usage Alerts in AI Admin Hub gives you a configurable rule engine that watches your skills and fires a notification when consumption crosses a threshold you define. It's the kind of thing admins have been asking for since AI-powered skills started showing up.

💡 What you're setting up: Assists are a purchased, metered resource. Burning through them without visibility isn't just annoying, it's an operational risk. Usage Alerts puts you back in control by letting you define thresholds per skill, choose how you want to be notified, and get the signal.

Two Metrics, One Rule

When you create a rule, you pick the skill you want to watch and choose which metric to count: assists or executions.

Assist consumption

Tracks the number of assists the skill has consumed within a given time window. This is your direct spend signal. If a skill is eating assists faster than expected, this metric surfaces it.

Execution count

Tracks how many times the skill actually ran, regardless of consumption. This one is valuable for catching runaway automations or unexpected call patterns before they compound into a consumption problem. A high execution count with low consumption can mean a skill is running more often than intended, but at a low cost. Having this available means you can monitor and alert for usage for a given skill.

How the Timing Works

You select the alert timeline window: The options are 1 day or 1 week, so a daily rule asks, "Did this skill exceed the threshold within the last 24 hours?" and a weekly rule looks at the last 7 days. What actually triggers the check is a scheduled job called "NAAUsageRuleProcessor: Calculate Now Assist usage and generate alerts," which runs hourly out of the box. So while your rule is defined in day or week increments, the system is evaluating it against that window every hour. Practically, that means you'll get notified pretty quickly after a threshold is crossed, not just once at the end of the day.

Severity is configurable too: Critical, Warning, or Info. Use this to give your team the right signal without training them to ignore noise. 

The Alerts Feed Is More Than a Log

This is the part worth calling out, because it changes how you actually respond to an issue.

From any alert, you can navigate directly to the detail page for the affected skill. You can edit the rule that triggered the alert without leaving the feed. And you can dismiss an alert from the active view without deleting the underlying rule, which matters because you probably don't want to throw away a rule that's working just because you've already handled what it surfaced.

Those three actions, in context, mean the workflow from "I got an alert" to "I've investigated and adjusted" stays within the tool. No context switching to find the rule, no hunting for the skill record.

 

 

Getting Started

Once you're on Australia Patch 4, navigate to Admin > AI Admin Hub > Settings > Usage Alerts and select Rules. Walk through the modal: name the rule, pick the skill, choose your metric, set the threshold, pick a timeline window, set the severity, and decide whether to enable email notifications. 

 

 

💡 Quick win: Start with your highest-volume skills and the ones tied to automations that run on a schedule. Those are the most likely candidates for unexpected consumption spikes, and getting a rule on them now means the next time something goes sideways at 3 am, you'll know about it before anyone has to go looking.

Why This Matters

Assist consumption is real spend. Being reactive about it creates friction downstream. Usage Alerts flip the script by putting visibility exactly where you need it, in the moment when you can actually do something. No more finding out at the end of the month that something went sideways. No more hunting through logs to understand what happened. You get notified, you investigate, you adjust, all without leaving the workspace.

If you're on Australia Patch 4, this one's worth setting up right away.

Learn more in the Usage Alerts documentation