Agent Architecture

FreightConnect runs autonomous agents that handle complex tasks without human intervention. This is our core differentiation — we're not a copilot (you do the work), we're an autopilot (the AI does the work).

What Are Agents?

Agents are AI-powered services that:

1. Monitor events (load posted, shipper inquiry, delivery exception)
2. Make decisions using rules and historical data
3. Take actions autonomously
4. Log their work for audit and learning

Think of each agent as a tireless team member who specializes in one job.

The Four Core Agents

1. Quoting Agent

Job: Rate shop your loads with 50+ carriers and rank offers

What it does:

• Takes a load you create
• Queries 50+ carriers based on your network, equipment type, and geography
• Ranks quotes by price, FSC score, and carrier fit
• Displays ranked results to you in 60-90 seconds
• Can auto-accept quotes if you enable that feature (Enterprise)

Key decisions:

• Which carriers to query (based on recent activity, equipment availability)
• How to rank results (price vs. reliability vs. capacity)
• Whether to send follow-up requests if no quotes come in

Customizable: You set minimum FSC score, insurance requirements, and territory preferences.

2. Carrier Vetting Agent

Job: Verify carrier credentials and flag fraud

What it does:

• Runs every carrier through FMCSA SAFER database checks
• Verifies insurance policies (real-time via third-party confirmations)
• Screens ownership against sanctions lists (OFAC, SDN)
• Analyzes claim history and chargeback frequency
• Assigns fraud risk score (0-100)

Key decisions:

• Which carriers to approve for your network
• When to flag a carrier for manual review
• When to auto-reject a carrier (failed compliance check)
• When to downgrade a carrier's FSC score (compliance violations detected)

Customizable: You set fraud risk threshold, minimum insurance amounts, and FMCSA rating requirements.

3. Tracking Agent

Job: Monitor shipments and flag exceptions in real-time

What it does:

• Queries ELD/telematics data every 1-2 minutes
• Monitors for exceptions: delays, route deviations, harsh events
• Updates shipper with revised ETA automatically
• Notifies you of critical exceptions (late delivery likely)
• Logs all updates for audit trail

Key decisions:

• Is the truck on schedule? If not, by how much?
• Should we escalate to manual intervention? (e.g., contact carrier directly)
• Should we trigger a recovery carrier? (for critical delays)
• What communication should the shipper receive?

Customizable: You set alert thresholds (delay tolerance, route deviation radius, etc.).

4. Customer Service Agent

Job: Handle shipper inquiries 24/7

What it does:

• Monitors your support email (or a dedicated FreightConnect inbox)
• Answers common questions: "Where's my shipment?" "What's my rate?" "When's delivery?"
• Escalates complex issues to your team (if needed)
• Sends proactive updates (shipper didn't ask, but there's important info)
• Maintains conversation history and context

Key decisions:

• Can I answer this question automatically? (e.g., tracking status)
• Does this need human review? (e.g., rate dispute)
• Should I escalate to a manager? (angry customer, complex issue)
• What tone should I use? (friendly, professional, apologetic)

Customizable: You set escalation rules and define what questions get auto-answered vs. escalated.

Agent Autonomy Levels

Each agent runs at one of three autonomy levels:

Full Autonomy (Agent Decides & Acts)

• Agent makes decision and executes
• You see the result after the fact
• Best for routine tasks (quoting, customer service)

Example: Quoting agent posts a load, queries carriers, customers see results automatically.

Human-in-Loop (Agent Recommends, You Approve)

• Agent analyzes and proposes action
• You review in your dashboard
• You click "Approve" or "Reject"
• Agent executes your decision

Example: Carrier vetting agent flags a carrier with a safety violation. You review the flag and decide whether to remove them from your network.

Audit-Only (Agent Logs, You Decide)

• Agent monitors and documents everything
• You have full visibility
• You make all decisions manually
• Agent provides recommendations

Example: Tracking agent monitors shipment but doesn't auto-escalate. You review tracking dashboard and decide when to contact the carrier.

How to Configure Agents

Go to Settings → Agents to control each agent:

1. Enable/Disable — turn agents on or off
2. Autonomy Level — choose from Full, Human-in-Loop, or Audit-Only
3. Rules & Thresholds — set parameters for decision-making
4. Escalation Rules — define when humans should step in
5. Communication Templates — customize how agents talk to carriers/shippers

Example Configuration

Quoting Agent:

• Autonomy: Full (auto-post quotes to dashboard)
• Min FSC score: 70
• Insurance required: Yes
• Fraud risk threshold: 50
• Territory: All

Tracking Agent:

• Autonomy: Human-in-Loop (escalate exceptions for approval)
• Delay threshold: 2 hours
• ETA update frequency: every 2 minutes
• Shipper notification: automatic on major changes
• Route deviation radius: 5 miles

Customer Service Agent:

• Autonomy: Human-in-Loop (requires approval for non-standard responses)
• Auto-answer questions: tracking status, rate confirmation, pickup time
• Escalate if: angry tone detected, refund requested, dispute filed
• Response time target: < 5 minutes

Agent Performance & Monitoring

Go to Analytics → Agent Health to see:

• Quoting Agent: avg quote volume per load, time to first quote, acceptance rate
• Carrier Vetting: carriers approved/rejected, compliance checks passed/failed
• Tracking Agent: exception accuracy (how often we correctly predict late delivery), shipper satisfaction
• Customer Service: resolution rate, escalation rate, response time

Use these metrics to fine-tune agent behavior.

Learning & Improvement

Agents improve over time by learning from outcomes:

• If the quoting agent quotes a carrier and they fail to deliver, the agent learns to deprioritize that carrier next time
• If the tracking agent predicts a delay and it comes true, it learns the pattern
• If the customer service agent's response is marked "helpful," it learns that communication style

You can also provide feedback:

1. Go to Analytics → Agent Logs
2. Find an agent action
3. Click Feedback → "This was correct/incorrect"
4. Agent incorporates your feedback

Agent Limitations

Agents can't:

• Create new policies (only follow existing rules)
• Authorize refunds (only recommend escalation)
• Access carrier bank accounts (only verify via official channels)
• Bypass your approval rules (if you set human-in-loop, they must wait)

Agents are powerful but bounded by your preferences and rules.

Cost

Agent usage is included in all plans:

• Solo: Basic agents (quoting + tracking; no AI customer service)
• Brokerage: All four agents with customization
• Enterprise: All agents + custom agents (e.g., "pricing optimization agent" for your specific use case)

Custom agents are available on Enterprise — contact matt@freightconnect.ai.

What's Next

• Quoting Agent — detailed guide to how rate shopping works
• Carrier Vetting Agent — compliance and fraud screening
• Tracking Agent — exception handling and proactive alerts
• Customer Service Agent — 24/7 support automation