Dispatch Assistants (AI Agents) | FieldCamp AI Dispatcher

Create AI dispatch assistants that automatically assign jobs on a schedule or after updates. Configure playbooks, guardrails, and monitoring to keep humans in the loop.

What Are Dispatch Assistants?

Dispatch Assistants are saved AI agents that automatically dispatch jobs without requiring manual approval for every assignment. Instead of reviewing each suggestion one by one, you define the rules once and let the assistant handle routine dispatching on a schedule or in response to job updates.

You can find Dispatch Assistants on the Agents page in AI Dispatcher at dispatcher.fieldcamp.ai.

AI Dispatcher Dispatch Assistants page with stats and guardrails configuration

The Agents page shows:

Active assistants count — How many assistants are currently running
Recent activity — A feed of recent dispatch actions taken by your assistants
Needs review — Items flagged for human attention based on your guardrails

Use the filter tabs at the top to view All, Active, Paused, or Draft assistants.

Creating an Assistant

To create a new dispatch assistant:

Open the Agents page in AI Dispatcher
Click Create Assistant
Give it a name that describes its purpose (e.g., "Weekday HVAC Dispatcher" or "Emergency After-Hours")
Choose a Playbook that defines how the assistant behaves
Set your Guardrails to control when the assistant should pause and ask for help
Save the assistant as Draft to test, or set it to Active to start dispatching immediately

Start with a Draft assistant and monitor its decisions for a few days before switching to Active. This lets you fine-tune the playbook and guardrails without affecting live operations.

Playbooks

AI Dispatcher Create Assistant form with playbooks, review rules, and guardrail configuration

A Playbook is a predefined set of rules that tells the assistant how to make dispatch decisions. Playbooks control:

Which jobs the assistant handles — By job type, priority, or time window
How aggressively it dispatches — Confidence threshold for auto-assignment (e.g., only dispatch if confidence is above 80%)
Scheduling behavior — Run on a fixed schedule (every 15 minutes, hourly) or react to new job creation and status changes
Technician preferences — Whether to prioritize proximity, workload balance, or customer continuity

You can assign one playbook per assistant. Multiple assistants can share the same playbook or each can have a unique one.

Guardrails

Guardrails are safety controls that determine when the assistant should stop and escalate to a human. Three guardrails are available:

Escalate exceptions only — The assistant dispatches autonomously for routine jobs but flags anything unusual (low confidence, skill gaps, schedule conflicts) for your review
Explain every recommendation — The assistant logs detailed reasoning for every dispatch decision, giving you a full audit trail even when it acts automatically
Keep external actions gated — The assistant can suggest and prepare assignments, but any action that touches external systems (customer notifications, calendar syncs) requires your approval first

You can enable multiple guardrails on a single assistant. For most teams, starting with Escalate exceptions only provides the best balance of speed and control.

Assistant Statuses

Each assistant has one of three statuses:

Active — Currently running and dispatching jobs according to its playbook and guardrails
Paused — Temporarily stopped. All configuration is preserved. Resume anytime without reconfiguring.
Draft — Not yet live. Use this to build and refine an assistant before activating it.

You can change status at any time from the Agents page. Pausing an assistant does not undo any dispatches it already made.

Monitoring Your Assistants

The Agents page gives you two monitoring tools:

Recent Activity

A chronological feed showing every action your assistants have taken — which jobs were dispatched, which technicians were assigned, and the confidence score for each decision. Use this to spot patterns and verify the assistant is making good choices.

Needs Review

Items that triggered a guardrail escalation appear here. These are jobs the assistant could not dispatch confidently and is waiting for your input on. Review each item, then accept the suggestion, reassign manually, or dismiss it.

Keeping Humans in the Loop

Dispatch Assistants are designed around a "human-in-the-loop" philosophy. Even at their most autonomous, they:

Never override a manual assignment you already made
Always flag low-confidence matches for review
Log every decision with full reasoning so you can audit at any time
Respect your guardrails as hard boundaries, not suggestions

The goal is to handle the 80% of routine dispatches automatically so you can focus your attention on the 20% that actually need human judgment. This always-on approach is part of a broader concept called continuous planning.