AI Agents

The Agent Org Chart

Commerce Index deploys one AI CEO (goal_manager), 13 department heads (VP Commerce, Head Supply Chain, Head Pricing, Head E-commerce, Head Retail, Head Wholesale, Head Finance, Head Customer Service, Head Tax, Head Brand & IP, Head Corp Compliance, CMO, Head Creator Success), and 60+ specialist worker agents. Each agent observes signals, makes decisions within guardrails, and learns from outcomes. Together, they form an autonomous commerce team.

The Agent Lifecycle

Every agent follows the same four-phase lifecycle. This creates a consistent, predictable, and auditable pattern across all agent operations:

Phase 1

Observe

Agents continuously monitor incoming signals — price changes, order events, inventory updates, customer reviews, search rankings, and more. They build a real-time picture of your commerce operations.

Phase 2

Decide

Using the Decision Graph, agents evaluate signals against your guardrails, historical patterns, and business context. They calculate a confidence score for each potential action.

Phase 3

Act

Based on their trust level, agents either take action autonomously (adjust price, create deal, reorder stock) or submit a recommendation for your approval.

Phase 4

Learn

Every action outcome is tracked. Agents learn from your approvals, rejections, and overrides to improve future decisions. The more you interact, the smarter they get.

ObserveDecideActLearnRepeat

Meet the Agents

Pricing Agent

Competitive pricing, automated

Monitors competitor prices across marketplaces, analyzes price elasticity, and recommends or executes optimal pricing strategies. Supports MAP compliance, dynamic pricing rules, and margin protection guardrails.

Observes
  • Competitor price changes
  • Sales velocity shifts
  • Margin threshold alerts
  • Seasonal demand patterns
Decides
  • Optimal price points per SKU
  • When to match vs. undercut competitors
  • Dynamic pricing windows
Actions
  • Adjust product prices
  • Send price change notifications
  • Generate pricing reports
Default Guardrails
  • Minimum margin threshold (default 20%)
  • Maximum daily price change (10%)
  • MAP compliance enforcement
  • Blackout date restrictions

Deal Agent

Promotions that perform

Creates, manages, and optimizes promotional campaigns. Analyzes deal performance in real-time, suggests discount depths based on inventory age and margin, and auto-expires underperforming deals.

Observes
  • Inventory age thresholds
  • Conversion rate drops
  • Competitor promotions
  • Seasonal opportunities
Decides
  • Which products to promote
  • Discount depth and duration
  • When to extend or expire deals
Actions
  • Create deals on thedeals.ai
  • Publish to distribution channels
  • Auto-expire underperformers
Default Guardrails
  • Maximum discount percentage (default 40%)
  • Minimum units in stock before promoting
  • Maximum concurrent deals per category
  • Budget caps per campaign

Inventory Agent

Never stock out again

Predicts stockouts using sales velocity and lead time analysis. Calculates optimal reorder points, suggests safety stock levels, and alerts you before problems happen — not after.

Observes
  • Stock level changes
  • Sales velocity trends
  • Supplier lead times
  • Seasonal demand forecasts
Decides
  • Reorder timing and quantity
  • Safety stock levels
  • Which supplier to use
Actions
  • Generate purchase orders
  • Send low-stock alerts
  • Update reorder points
Default Guardrails
  • Maximum order value per PO
  • Minimum days-of-supply threshold
  • Approved supplier list only
  • Budget limits per reorder cycle

Fulfillment Agent

Ship it right, every time

Tracks shipments across carriers, detects delivery exceptions in real-time, proactively notifies customers of delays, and monitors fulfillment SLAs. Flags patterns like repeated carrier issues.

Observes
  • Shipping status updates
  • Delivery exception alerts
  • Carrier performance data
  • SLA breach warnings
Decides
  • When to escalate delays
  • Carrier selection for new shipments
  • Customer notification timing
Actions
  • Send proactive delay notifications
  • Flag carrier issues
  • Generate fulfillment reports
Default Guardrails
  • SLA breach threshold (default 24h)
  • Maximum auto-refund amount
  • Escalation rules for repeat issues
  • Carrier performance minimums

CX Agent

Customer sentiment, quantified

Monitors customer sentiment across reviews, support tickets, and social mentions. Calculates NPS trends, identifies emerging issues before they escalate, and routes critical feedback for immediate attention.

Observes
  • New reviews and ratings
  • Support ticket sentiment
  • Social media mentions
  • NPS score changes
Decides
  • Escalation priority level
  • Response templates to suggest
  • Trend alerts to surface
Actions
  • Route critical tickets
  • Generate sentiment reports
  • Trigger escalation workflows
Default Guardrails
  • Never auto-respond to customers
  • Escalate all 1-star reviews
  • Minimum review volume before trends
  • PII handling restrictions

SEO Agent

Be found, be bought

Optimizes product titles, descriptions, and metadata for search engine visibility. Monitors keyword rankings, suggests content improvements, and tracks organic traffic impact per product.

Observes
  • Keyword ranking changes
  • Search impression data
  • Click-through rate shifts
  • Competitor content updates
Decides
  • Title and description rewrites
  • Meta tag optimization
  • Keyword targeting strategy
Actions
  • Update product titles/descriptions
  • Generate SEO reports
  • Suggest image alt text
Default Guardrails
  • Brand voice guidelines
  • Maximum title length (70 chars)
  • No keyword stuffing (density < 3%)
  • Require approval for title changes

Distribution Agent

Everywhere your customers are

Manages product distribution across multiple channels and marketplaces. Syncs listings, pricing, and inventory across Amazon, eBay, Walmart, and other platforms. Detects channel conflicts.

Observes
  • Channel performance metrics
  • Listing sync failures
  • Cross-channel price conflicts
  • New marketplace opportunities
Decides
  • Channel prioritization
  • Price parity enforcement
  • Listing optimization per channel
Actions
  • Sync listings across channels
  • Resolve price conflicts
  • Generate channel reports
Default Guardrails
  • Price parity rules across channels
  • Approved channel list only
  • Maximum listing changes per day
  • Channel-specific compliance rules

How Agents Make Decisions

Every agent decision follows a structured flow: signal received, guardrails checked, confidence scored, and action taken or approval requested. Here is a simplified example:

// Simplified agent decision flow

signal_received = await agent.observe(event)
  // e.g., competitor lowered price by 15%

guardrails_check = await agent.checkGuardrails({
  min_margin: 0.20,         // Never go below 20% margin
  max_change_pct: 0.10,     // Max 10% price change per day
  map_compliance: true,      // Respect MAP pricing
  blackout_dates: ["2024-11-29"]  // No changes on Black Friday
})

if (!guardrails_check.passed) {
  return agent.log("Blocked by guardrail: " + guardrails_check.reason)
}

confidence = await agent.scoreConfidence({
  signal_strength: 0.85,    // How strong is the signal?
  historical_success: 0.72, // Past success rate for similar actions
  data_quality: 0.90        // How complete is our data?
})
// → confidence = 0.82

if (confidence >= agent.trust_level.auto_threshold) {
  // Trust level allows autonomous action
  await agent.execute(action)
  await agent.notify("Price adjusted: $49.99 → $44.99")
} else {
  // Needs human approval
  await agent.requestApproval({
    action: "Reduce price to $44.99",
    reason: "Competitor dropped to $42.99, maintaining margin at 22%",
    confidence: 0.82,
    expires_in: "24h"
  })
}
1

Signal Received

An event triggers the agent

2

Guardrails Checked

Business rules validated

3

Confidence Scored

Action certainty calculated

4

Action or Approval

Execute or request review

Related Documentation

What's Next?