Building a Donor Engagement Agent: Donation Processing, Receipts, and Thank-You Messages

Why Donor Engagement Needs an Agentic Approach

Nonprofits live and die by their donor relationships. A first-time donor who receives a generic confirmation email is far less likely to give again than one who receives a timely, personalized acknowledgment that connects their gift to a specific impact. Yet most small nonprofits lack the staff to deliver that level of engagement consistently.

An AI donor engagement agent can handle the full lifecycle: accepting donations through payment APIs, generating compliant tax receipts, crafting personalized thank-you messages based on donor history, and managing recurring giving schedules. The agent operates autonomously for routine transactions while escalating edge cases to human staff.

Designing the Donor Data Model

Before building the agent, define the data structures that represent donors and their contributions.

sequenceDiagram
    autonumber
    participant Caller as Caller
    participant Agent as CallSphere Agent
    participant API as CRM API
    participant DB as CRM Database
    participant Webhook as Webhook Listener
    Caller->>Agent: Inbound call begins
    Agent->>Agent: STT plus intent detection
    Agent->>API: Lookup contact by phone
    API->>DB: Read contact record
    DB-->>API: Contact and history
    API-->>Agent: Personalized context
    Agent->>API: Create call activity
    Agent->>API: Update deal stage
    API->>Webhook: Outbound webhook fires
    Webhook-->>Agent: Confirmed
    Agent->>Caller: Spoken confirmation

from dataclasses import dataclass, field
from datetime import datetime, date
from enum import Enum
from typing import Optional
from uuid import uuid4

class DonationType(Enum):
    ONE_TIME = "one_time"
    RECURRING_MONTHLY = "recurring_monthly"
    RECURRING_ANNUAL = "recurring_annual"
    IN_KIND = "in_kind"

class DonationStatus(Enum):
    PENDING = "pending"
    COMPLETED = "completed"
    FAILED = "failed"
    REFUNDED = "refunded"

@dataclass
class Donor:
    donor_id: str = field(default_factory=lambda: str(uuid4()))
    first_name: str = ""
    last_name: str = ""
    email: str = ""
    phone: Optional[str] = None
    total_lifetime_giving: float = 0.0
    first_gift_date: Optional[date] = None
    last_gift_date: Optional[date] = None
    gift_count: int = 0
    is_recurring: bool = False
    preferred_fund: Optional[str] = None
    communication_preference: str = "email"

@dataclass
class Donation:
    donation_id: str = field(default_factory=lambda: str(uuid4()))
    donor_id: str = ""
    amount: float = 0.0
    donation_type: DonationType = DonationType.ONE_TIME
    status: DonationStatus = DonationStatus.PENDING
    fund_designation: str = "General Fund"
    payment_method: str = "card"
    transaction_ref: Optional[str] = None
    receipt_number: Optional[str] = None
    created_at: datetime = field(default_factory=datetime.utcnow)
    is_tax_deductible: bool = True

Building the Payment Processing Tool

The agent needs a tool that integrates with a payment processor. Here we use Stripe as an example, wrapping the API call so the agent can invoke it as a tool.

import stripe
from agents import Agent, Runner, function_tool

stripe.api_key = os.environ["STRIPE_SECRET_KEY"]

@function_tool
async def process_donation(
    donor_email: str,
    amount_cents: int,
    fund: str,
    payment_method_id: str,
    is_recurring: bool = False,
) -> dict:
    """Process a donation payment through Stripe."""
    try:
        if is_recurring:
            customer = stripe.Customer.create(
                email=donor_email,
                payment_method=payment_method_id,
                invoice_settings={
                    "default_payment_method": payment_method_id
                },
            )
            subscription = stripe.Subscription.create(
                customer=customer.id,
                items=[{"price": os.environ["STRIPE_RECURRING_PRICE_ID"]}],
                metadata={"fund": fund},
            )
            return {
                "status": "active",
                "subscription_id": subscription.id,
                "donor_email": donor_email,
            }

        intent = stripe.PaymentIntent.create(
            amount=amount_cents,
            currency="usd",
            payment_method=payment_method_id,
            confirm=True,
            receipt_email=donor_email,
            metadata={"fund": fund},
        )
        return {
            "status": intent.status,
            "transaction_id": intent.id,
            "amount": amount_cents / 100,
        }
    except stripe.error.StripeError as e:
        return {"status": "failed", "error": str(e)}

Generating Tax-Deductible Receipts

Nonprofits must issue receipts that meet IRS requirements. The agent generates these automatically after a successful payment.

@function_tool
async def generate_receipt(
    donor_name: str,
    donor_email: str,
    amount: float,
    donation_date: str,
    fund: str,
    transaction_id: str,
    org_ein: str = "12-3456789",
) -> dict:
    """Generate an IRS-compliant donation receipt."""
    receipt_number = f"RCP-{datetime.utcnow().strftime('%Y%m%d')}-{transaction_id[:8]}"

    receipt_text = (
        f"OFFICIAL DONATION RECEIPT\n"
        f"Receipt #: {receipt_number}\n"
        f"Organization: Hope Community Foundation\n"
        f"EIN: {org_ein}\n\n"
        f"Donor: {donor_name}\n"
        f"Date: {donation_date}\n"
        f"Amount: ${amount:.2f}\n"
        f"Designation: {fund}\n"
        f"Transaction ID: {transaction_id}\n\n"
        f"No goods or services were provided in exchange "
        f"for this contribution. This receipt may be used "
        f"for tax deduction purposes."
    )

    return {
        "receipt_number": receipt_number,
        "receipt_text": receipt_text,
        "donor_email": donor_email,
    }

Personalizing Thank-You Messages

The agent crafts thank-you messages that reference the donor's history and the specific impact of their gift.

@function_tool
async def lookup_donor_history(donor_email: str) -> dict:
    """Retrieve donor history for personalized messaging."""
    # In production, query your donor database
    return {
        "first_name": "Sarah",
        "total_lifetime_giving": 2500.00,
        "gift_count": 8,
        "first_gift_date": "2024-06-15",
        "preferred_fund": "Youth Programs",
        "is_recurring": True,
    }

@function_tool
async def send_thank_you(
    donor_email: str,
    subject: str,
    message_body: str,
) -> dict:
    """Send a personalized thank-you email to the donor."""
    # In production, use SendGrid, SES, or similar
    print(f"Sending to {donor_email}: {subject}")
    return {"status": "sent", "recipient": donor_email}

Assembling the Donor Engagement Agent

Wire the tools together into a single agent with instructions that guide its behavior across the full donation lifecycle.

donor_agent = Agent(
    name="Donor Engagement Agent",
    instructions="""You are a donor engagement agent for Hope
Community Foundation. Your responsibilities:

1. Process donations via the payment tool
2. Generate IRS-compliant receipts for every gift
3. Look up donor history to personalize thank-you messages
4. For recurring donors, acknowledge their ongoing commitment
5. For first-time donors, welcome them warmly
6. Always confirm the fund designation before processing
7. If a payment fails, offer to retry or suggest alternatives
8. Never store or repeat full payment card details""",
    tools=[
        process_donation,
        generate_receipt,
        lookup_donor_history,
        send_thank_you,
    ],
)

result = Runner.run_sync(
    donor_agent,
    "Sarah at [email protected] wants to donate $100 to Youth Programs "
    "using payment method pm_card_visa. Please process the donation, "
    "generate a receipt, and send a personalized thank-you.",
)
print(result.final_output)

The agent will call each tool in sequence: process the payment, generate the receipt, look up Sarah's history, then compose and send a thank-you that references her eight previous gifts and ongoing support.

FAQ

How does the agent handle failed payments?

The agent checks the payment status returned by Stripe. If the status is failed, it informs the donor with a clear message and suggests retrying with a different payment method. The agent never stores card details itself — it works only with tokenized payment method IDs from Stripe.

Can this agent handle in-kind donations?

Yes. You can add a separate tool for logging in-kind donations that skips the payment processing step but still generates a receipt. The receipt for in-kind gifts should describe the donated item rather than listing a dollar amount, since the IRS requires the donor to determine fair market value.

How do you ensure receipt compliance across states?

The receipt template includes the organization's EIN, a statement that no goods or services were exchanged, and the exact donation amount and date. For gifts over $250, the IRS requires a written acknowledgment, which this agent provides automatically. State-specific requirements can be added as conditional logic in the receipt generation tool.

---

#NonprofitAI #DonorManagement #PaymentIntegration #AgenticAI #Python #LearnAI #AIEngineering