Entity monitor

The entity monitor is a way to point an agent at one specific thing in your operation — a delivery, order, route, return, or job — and have it keep an eye on that thing over time, checking back on a regular cadence and flagging trouble as soon as it’s warranted. It is not a one-off lookup and it is not a generic rule builder. It’s a persistent “watch this until it’s done” capability: the agent re-evaluates the entity on a fixed interval and produces a fresh health read each time, until the entity reaches a final state (completed, failed, or canceled) or can no longer be found.

​

When to use this

Reach for the entity monitor when a single delivery or route matters enough that you want eyes on it without sitting and refreshing a screen. For example:

• A high-value or VIP delivery you can’t afford to have slip.
• A route that’s already shown early signs of running behind.
• An order tied to a promise you’ve made a customer, where you want to know the moment risk appears.

For broad, recurring work across your whole operation — “check every morning for stuck deliveries” — a scheduled custom agent is the better fit. The entity monitor is for focusing on one named thing and staying on it.

​

How it works

Nash ships an Entity Monitor agent in every organization’s agent list. It’s a Nash-curated agent purpose-built to watch a single entity and assess its health. The mental model has three parts: You point it at one entity. A watch is always scoped to a specific delivery, order, route, return, or job — identified by its Nash ID. The agent doesn’t roam your whole operation; it concentrates on that one thing. It checks back on a cadence. Once a watch is open, the agent re-evaluates the entity on a regular interval (every 15 minutes by default). Each pass is a fresh, focused look — it carries forward a short summary of what it found last time rather than re-reading the entire history, so it stays sharp on what changed. It produces a health read each time. Every check yields a structured assessment: an overall risk level (green, yellow, orange, or red), what the agent observed, whether the situation is trending better or worse, and a recommended action. These land in a dedicated session you can open from your sidebar, so you can watch the picture evolve over time. The watch closes itself automatically once the entity reaches a terminal state — completed, failed, or canceled — or if it can no longer be found. You don’t have to remember to turn it off.

​

What it does on a match

What happens when the agent spots a problem depends on the agent’s mode:

• Report-only (the default): the agent recommends. It tells you what it found and what it would do — but it doesn’t take the action itself.
• Auto-execute: the agent can take the recommended action on its own. Destructive actions (canceling, refunding, reassigning, and the like) still surface as confirmations you approve before anything happens, based on your organization’s guardrails.

​

Set it up

The most direct way to start a watch is to ask the agent in plain conversation. You don’t configure a rule or build a trigger — you tell the agent what to keep an eye on.

​

Managing watches

• List active watches. Ask the agent what it’s currently watching to see the open watches in your org.
• Stop a watch. Tell the agent to stop watching a given entity. It closes the watch and won’t reopen it.

​

Customizing the Entity Monitor agent

The Entity Monitor that ships with your org is Nash-curated and read-only — you can run it as-is for recommendations, but you can’t edit it in place. To tailor it (or to let it take actions), duplicate it into your organization from the agent list. The copy is yours to edit. Duplicating, editing, and running agents requires the schedule-management permission, which is typically held by ops managers and admins rather than every user.

​

What to expect

• Recommendations, not silent action, by default. Unless you’ve deliberately switched a duplicated agent to auto-execute, the monitor surfaces what it would do and waits for you.
• A running health log. The watch session is a timeline of assessments, so you can see how risk built (or eased) rather than just a single snapshot.
• It stops on its own. When the entity finishes, fails, or is canceled, the watch closes. There’s also a hard safety deadline so a watch can never run indefinitely (12 hours by default — see Scheduling & execution), and the agent stops itself if it hits repeated errors.

​

Related