Quick Start - Agentic AI Process Automation

Get started with autonomous process agents in 5 minutes!

---

🚀 Installation

1. Install Dependencies

pip install -r requirements.txt

2. Set API Keys (Optional)

export OPENAI_API_KEY="your-key-here"

---

🧾 Invoice Processing Agent

Run Demo

python examples/process_automation/invoice_processing_demo.py

Use in Code

from packages.agents.process_agents import InvoiceProcessingAgent

# Initialize
agent = InvoiceProcessingAgent()

# Process invoice
result = await agent.process_invoice("path/to/invoice.pdf")

# Check result
if result.approval_decision.approved:
    # Auto-approved!
    print(f"✓ Approved: ${result.invoice_data.total_amount}")
else:
    # Route for approval
    print(f"⚠ Route to: {result.approval_decision.escalation_level}")

Configuration

validator_config = {
    "auto_approve_threshold": 1000.00,
    "approved_vendors": ["Acme Corp", "Global Supplies"]
}

router_config = {
    "auto_approve_limit": 1000.00,
    "supervisor_limit": 5000.00,
    "manager_limit": 25000.00
}

agent = InvoiceProcessingAgent(
    validator_config=validator_config,
    router_config=router_config
)

---

📧 Email Response Agent

Run Demo

python examples/process_automation/email_response_demo.py

Use in Code

from packages.agents.process_agents import EmailResponseAgent, Email
from datetime import datetime

# Initialize
agent = EmailResponseAgent()

# Create email
email = Email(
    email_id="email-001",
    from_address="customer@example.com",
    to_address="support@company.com",
    subject="How do I reset my password?",
    body="I can't login to my account...",
    received_at=datetime.utcnow()
)

# Process email
result = await agent.process_email(email)

# Check result
if result.auto_send:
    print(f"✓ Auto-sent: {result.draft.subject}")
else:
    print(f"⚠ Requires approval: {result.classification.category}")

With RAG Context

# Add retriever for context
from packages.rag.retrievers import HybridRetriever

agent = EmailResponseAgent(retriever=your_retriever)

# Now drafts will include KB context
result = await agent.process_email(email)

---

📊 Research & Report Agent

Run Demo

python examples/process_automation/research_report_demo.py

Use in Code

from packages.agents.process_agents import ResearchAgent, ReportFormat

# Initialize
agent = ResearchAgent()

# Conduct research
result = await agent.conduct_research(
    research_question="What are AI trends in healthcare?",
    output_format=ReportFormat.MARKDOWN
)

# Access report
print(f"Title: {result.report.title}")
print(f"Summary: {result.report.executive_summary[:200]}...")

# Save report
agent.report_generator.save_report(
    result.report,
    "outputs/research_report.md"
)

With Custom Context

# Add context to guide research
context = {
    "focus": "enterprise_adoption",
    "industry": "healthcare",
    "timeframe": "2024-2025"
}

result = await agent.conduct_research(
    research_question="What are AI trends in healthcare?",
    context=context
)

---

🔄 Multi-Agent Workflow

Combine Multiple Agents

from packages.agents.process_agents import (
    InvoiceProcessingAgent,
    EmailResponseAgent,
    ResearchAgent
)

# Initialize all agents
invoice_agent = InvoiceProcessingAgent()
email_agent = EmailResponseAgent()
research_agent = ResearchAgent()

# Workflow: Process invoice → Research vendor → Send email
async def invoice_verification_workflow(invoice_pdf, vendor_query):
    # Step 1: Process invoice
    invoice_result = await invoice_agent.process_invoice(invoice_pdf)
    
    # Step 2: Research vendor if unknown
    if not invoice_result.validation_result.vendor_approved:
        research_result = await research_agent.conduct_research(
            f"Research vendor: {invoice_result.invoice_data.vendor_name}"
        )
    
    # Step 3: Send notification email
    email = create_notification_email(invoice_result)
    email_result = await email_agent.process_email(email)
    
    return {
        "invoice": invoice_result,
        "research": research_result,
        "notification": email_result
    }

---

⚙️ Configuration Options

Invoice Agent

config = {
    # Validator
    "validator_config": {
        "auto_approve_threshold": 1000.00,
        "max_invoice_age_days": 90,
        "require_po_above": 5000.00,
        "approved_vendors": ["Vendor1", "Vendor2"]
    },
    
    # Router
    "router_config": {
        "auto_approve_limit": 1000.00,
        "supervisor_limit": 5000.00,
        "manager_limit": 25000.00,
        "auto_approve_enabled": True
    }
}

Email Agent

config = {
    # Classifier
    "classifier_config": {
        "model": "gpt-4o-mini",
        "temperature": 0,
        "enable_pii_detection": True
    },
    
    # Drafter
    "drafter_config": {
        "model": "gpt-4o",
        "temperature": 0.3,
        "templates_path": "path/to/templates.json"
    }
}

Research Agent

config = {
    # Planner
    "planner_config": {
        "model": "gpt-4o",
        "max_sub_questions": 7
    },
    
    # Gatherer
    "gatherer_config": {
        "enable_web_search": True,
        "enable_internal_search": True
    },
    
    # Generator
    "generator_config": {
        "default_format": "markdown"
    }
}

---

📊 Performance Expectations

Processing Times

Agent	Task	Time	Cost
Invoice	Process 1 invoice	4-6s	$0.002
Email	Draft 1 response	4-6s	$0.004
Research	Generate 1 report	40-55s	$0.03-0.05

Automation Rates

Agent	Auto-Handle	Manual Review	Escalation
Invoice	70%	25%	5%
Email	75%	20%	5%
Research	100%	0%	0%

---

🧪 Testing

Run Unit Tests

# All tests
pytest tests/process_agents/ -v

# Specific agent
pytest tests/process_agents/test_invoice_agent.py -v

# With coverage
pytest tests/process_agents/ --cov=packages.agents.process_agents

---

📚 Next Steps

1. Try the Demos

Run all three demos to see the agents in action.

2. Review Documentation

• Week 1 Complete - Invoice
• Week 2 Complete - Email
• Week 3 Complete - Research

3. Customize for Your Needs

• Modify business rules in config files
• Add custom templates
• Integrate with your data sources

4. Deploy to Production

• Add your own PDF invoices
• Connect to your email system
• Integrate with your knowledge base

---

💡 Common Use Cases

Finance Operations

# Automate invoice processing
agent = InvoiceProcessingAgent()
for invoice_file in pending_invoices:
    result = await agent.process_invoice(invoice_file)
    if result.approval_decision.approved:
        process_payment(result)
    else:
        send_for_approval(result)

Customer Support

# Automate email responses
agent = EmailResponseAgent(retriever=kb_retriever)
for email in inbox:
    result = await agent.process_email(email)
    if result.auto_send:
        send_email(result.draft)
    else:
        queue_for_review(result)

Business Intelligence

# Automate research reports
agent = ResearchAgent(retriever=docs_retriever)
result = await agent.conduct_research(
    "Market analysis for Q4 2025"
)
agent.report_generator.save_report(result.report, "q4_analysis.md")

---

🔧 Troubleshooting

Import Errors

# If you get import errors, ensure project root is in path
import sys
from pathlib import Path
project_root = Path(__file__).parent.parent.parent
sys.path.insert(0, str(project_root))

API Key Not Set

# Set OpenAI API key
export OPENAI_API_KEY="your-key"

# Or use .env file
echo "OPENAI_API_KEY=your-key" > .env

PDF Parsing Issues

# Install pdfplumber if missing
pip install pdfplumber>=0.10.3

---

📞 Support

• Documentation: /docs/docs/services/agentic-process-automation/
• Examples: /examples/process_automation/
• Tests: /tests/process_agents/
• Code: /packages/agents/process_agents/

---

Ready to build AI employees! 🤖🚀