Manage your agentic assists consumption with these system properties

[Victor Chen]

Note: The following properties are available as of the Australia Patch 3 release (June 2026)

Now Assist AI Agents does not charge for tokens, it charges for value in the form of assists. Here are properties you can use to manage and protect your assists from runaway consumption found in the sn_aia_property table. 

Trigger throttling

ServiceNow provides five configurable system properties that controls AI Agent triggers from over-executing on the same record within a short period of time. Any over-use in a given time period is considered a "breach"; the time periods are configurable per below. The system properties control email notifications that warn about runaway triggers and potentially de-activate them:

• kill_switch.mode: Options are off, warn_only, or enforce. Default value: warn_only. 
  • Warn_only = Sends a daily warning. 
  • Enforce = Sends a warning on Days 1 and 2, before deactivating trigger on Day 3. 

kill_switch.max_fires_per_window	Fires per record that mark it as breaching.	Default value: 5
kill_switch.min_distinct_records	Breaching records needed for the window to count as runaway.	25
kill_switch.window_size	Length of one observation window.	1440 min, i.e., 24 hours
kill_switch.consecutive_windows_duration	The total look back span.	4320 min, i.e., 3 days

Recursive check properties

ServiceNow provides properties that cap how many executions a trigger can create in a certain time frame.

recursive_check.create_max_executions	Maximum number of matching executions creating records	Default value: 50
recursive_check.create_time_window	Time window, in minutes, for checking for matching executions creating records	15
recursive_check.update_max_executions	Maximum number of matching executions updating a record	5
recursive_check.update_time_window	Time window, in minutes, for checking for matching execution updates	15

Assist alert properties

ServiceNow provides properties that alerts users to sudden assists spikes. 

alert.assist_spike_hours_to_check	Time interval, in hours, between running the scheduled job that checks for spikes in assist usage	Default value: 3
alert.assist_spike_usage_percentage_threshold	Percentage increase required from previous job results to trigger the spike in assist usage notification	0.5, i.e., 50%
alert.assist_spike_usage_threshold	Minimum number of assists to trigger the spike in assist usage notification	5000

More Best Practices

• Always monitor your assists usage in either the AI Agents Analytics dashboard (navigate to AI Agent Studio > Analytics > Assist Consumption) or Now Assist subscription management (Subscription management > Subscriptions > Now Assist usage). You can view our Platform Academy video on Now Assist usage in Subscription Management.
• Always review and test your AI Agent trigger configurations. Avoid potential loops when combining "On Record Create" and "On Record Update". Avoid over-scheduling triggers. Review this article on AI Agents Triggers Best Practices
• Do not turn off assists consumption-related notifications, even for sub-prod environments. Sub-prod environments also consume assists!
• AI Control Tower also provides consumption dashboards by agent for proactive monitoring.
• Review your AI Agents and agentic workflows for "over-tooling". Is it possible to combine or reduce the number of tools in use? Could a Virtual Agent topic or conversational catalog also fulfill your use case? As a reminder, using 0-4 tools consumes 25 assists; 5-8 tools consumes 50 assists; and 9-20 tools consumes 150 assists. 
• Update your Now Assist AI Agents store app where appropriate. ServiceNow recommends at least version 6.0+.