Use Case: AI Procurement Agent

How an AI procurement agent uses SpendGuard to control vendor payments and spending.

The Scenario

You have an AI procurement agent that approves and processes vendor payments. When a department requests a purchase, the agent evaluates it and processes the payment.

The problem: Without guardrails, the agent could pay unapproved vendors, exceed budget limits, send payments to sanctioned countries, or process the same invoice twice.

The solution: SpendGuard enforces your procurement policy on every payment. The agent checks before paying, and SpendGuard returns allow, block, or escalate.

---

Step 1: Create Your Vendor Spend Policy

Start with the Vendor Spend Policy template and customize it:

curl -X POST https://spendguardapi.com/v1/policies \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $SPENDGUARD_API_KEY" \
  -d '{
    "policy_id": "vendor_spend_policy",
    "name": "Vendor Spend Policy",
    "description": "Controls payments made by our procurement agent",
    "rules": [
      {
        "rule_id": "r1",
        "rule_type": "max_amount",
        "description": "Block payments over $10,000",
        "parameters": { "limit": 10000, "currency": "USD" }
      },
      {
        "rule_id": "r2",
        "rule_type": "vendor_allowlist",
        "description": "Only approved vendors",
        "parameters": {
          "vendors": ["vendor_acme", "vendor_globex", "vendor_initech", "vendor_aws", "vendor_gcp"]
        }
      },
      {
        "rule_id": "r3",
        "rule_type": "escalate_if",
        "description": "Payments over $2,500 need finance approval",
        "parameters": { "amount_above": 2500, "action_types": ["spend"] }
      },
      {
        "rule_id": "r4",
        "rule_type": "geography_block",
        "description": "Block sanctioned countries",
        "parameters": { "blocked_countries": ["RU", "KP", "IR", "CU"] }
      },
      {
        "rule_id": "r5",
        "rule_type": "duplicate_guard",
        "description": "Block duplicate payments within 60 minutes",
        "parameters": { "window_minutes": 60 }
      }
    ]
  }'

---

Step 2: Wire the Agent to Check Before Paying

def process_vendor_payment(agent_id, vendor_id, amount, country=None):
    """Check with SpendGuard before processing a vendor payment."""

    check = requests.post(
        f"{SPENDGUARD_URL}/v1/checks",
        headers={
            "X-API-Key": SPENDGUARD_KEY,
            "Content-Type": "application/json",
        },
        json={
            "agent_id": agent_id,
            "policy_id": "vendor_spend_policy",
            "action_type": "spend",
            "amount": amount,
            "currency": "USD",
            "counterparty": vendor_id,
            "metadata": {
                "country": country,
            },
        },
    ).json()

    decision = check["decision"]

    if decision == "allow":
        execute_payment(vendor_id, amount)
        return f"Payment of ${amount} to {vendor_id} processed."

    elif decision == "escalate":
        create_finance_ticket(vendor_id, amount, check["message"])
        return f"This payment needs finance team approval: {check['message']}"

    elif decision == "block":
        return f"Payment blocked: {check['message']}"

---

Step 3: See It In Action

$1,500 payment to approved vendor → Allow

{ "decision": "allow", "message": "Action is within policy. Proceed." }

$3,000 payment to approved vendor → Escalate

Above the $2,500 threshold:

{
  "decision": "escalate",
  "reason_code": "escalation_threshold_exceeded",
  "message": "Spend of $3000.00 exceeds the escalation threshold of $2500.00."
}

$5,000 payment to unapproved vendor → Block

Vendor is not on the allowlist:

{
  "decision": "block",
  "reason_code": "vendor_not_on_allowlist",
  "message": "Vendor 'vendor_unknown' is not on the approved vendor list."
}

Payment to sanctioned country → Block

{
  "decision": "block",
  "reason_code": "blocked_geography",
  "message": "Actions from country 'RU' are blocked by policy."
}

---

Step 4: Update the Vendor Allowlist

When you onboard a new vendor, update the policy. SpendGuard creates a new version — the old version is preserved for audit:

curl -X POST https://spendguardapi.com/v1/policies \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $SPENDGUARD_API_KEY" \
  -d '{
    "policy_id": "vendor_spend_policy",
    "name": "Vendor Spend Policy",
    "description": "Controls payments made by our procurement agent",
    "rules": [
      ...same rules, but with "vendor_newco" added to the vendors array...
    ]
  }'

This creates version 2. All future checks use version 2. Past audit records still reference version 1.