Skip to main content

​
What are Autonomous Agents?

While the FusionX Assistant requires you to prompt it for every action, Autonomous Agents are AI workers that run independently in the background. You assign an Agent a goal, give it access to specific tools (like MailX or FileX), and it determines the necessary steps to accomplish that goal, executing them without further human intervention.

​
Types of Agents

AIBAMS provides several pre-configured agent types, optimized for different business functions:

The Organizer

Monitors a FileX folder. Automatically renames, tags, and sorts incoming files based on their content.

The Triage Agent

Monitors a MailX Shared Inbox (e.g., support@). Categorizes emails, drafts suggested replies, and assigns them to the right team member.

The Researcher

Given a topic, searches the web, analyzes documents in your workspace, and compiles a comprehensive report.

The Monitor

Watches specific metrics or data sources and sends an alert if anomalies are detected (e.g., a spike in website contact form submissions).

​
Deploying an Agent

To create and deploy an agent:
  1. Go to FusionX → Agents
  2. Click Deploy New Agent
  3. Select a Template or choose “Custom Agent”
  4. Configure the Goal: Describe exactly what the agent should achieve.
    • Example: “Monitor the ‘Invoices’ folder in FileX. Whenever a new PDF is uploaded, extract the total amount, vendor name, and due date, and log it into the ‘Finance Tracking.xlsx’ spreadsheet.”
  5. Assign Tools: Grant the agent access to the specific apps it needs (e.g., FileX Read/Write).
  6. Set the Trigger:
    • Continuous: Runs 24/7, reacting to events instantly.
    • Scheduled: Runs at a specific time (e.g., every Friday at 5 PM).
    • Manual: Only runs when you click “Start”.
  7. Click Deploy.

​
Agent Supervision & Safety

We built Autonomous Agents with strict guardrails to ensure they operate safely within your business.

​
1. Human-in-the-Loop (HITL) Approval

By default, destructive actions (like deleting files or sending emails to external clients) require human approval. When an agent reaches an approval step, it pauses and sends a notification. You can review its proposed action in the Agent Dashboard and click Approve or Reject. You can turn off HITL for trusted workflows.

​
2. Strict Tool Scoping

Agents only have access to the specific tools and data you assign them. An agent tasked with organizing files cannot read your emails unless explicitly granted MailX access.

​
3. Execution Limits

To prevent runaway processes, every agent is subject to execution limits (e.g., max 50 actions per run, max 10 minutes runtime). If an agent hits a limit, it pauses and notifies the owner.

​
Monitoring Agent Activity

You can track exactly what your agents are doing at any time.
  1. Go to FusionX → Agents
  2. Click on an active Agent to view its Run Log.
The Run Log shows the agent’s “thought process”—how it broke down the goal, which tools it decided to use, the results of those tool calls, and its final action.
StatusMeaning
RunningCurrently executing a task.
WaitingPaused, waiting for Human-in-the-Loop approval.
SleepingContinuous agent waiting for a trigger event.
FailedEncountered an error or hit an execution limit.
StoppedManually halted by a user.

​
Advanced: Multi-Agent Workflows

(Enterprise Feature) For highly complex processes, you can chain multiple agents together. For example, an onboarding workflow:
  1. Agent 1 (Triage) receives a signed contract via email and saves it to FileX.
  2. Agent 2 (Data Extractor) reads the contract, extracts the client’s details, and passes them to Agent 3.
  3. Agent 3 (Provisioner) uses those details to generate a welcome packet and send the onboarding email.
To set up multi-agent workflows, use the Workflow Builder tab in the Agents interface.