Build a Voice AI Agent for Property Management with Python SDK

🏢 Enterprise AI Consulting

Get dedicated help specific to your use case and for your hardware and software choices.

TLDR: Build a voice AI agent that handles tenant requests on-device — without cloud dependencies, transcription errors, or privacy risks. This tutorial shows how to create an AI-powered property manager using Python that listens for commands and returns structured JSON you can integrate directly into your property management software.

What you'll build: AI-powered voice agent that outputs structured JSON to integrate with property management apps, backend workflows, or phone systems, when tenants say "Hey Assistant! Schedule emergency plumbing in unit 20."

{
"intent": "maintenanceRequest",
  "slots": {
    "priority": "emergency",
    "issue": "plumbing",
    "unitNumber": 20
  }
}

Modern property management platforms increasingly benefit from AI agents to streamline tenant communication and building operations. On-device Voice interfaces offer one of the most practical entry points for this automation, especially for teams looking to build voice AI for property management without cloud dependencies. A property management voice AI agent can listen for tenant requests, interpret intent, and trigger automated actions, such as submitting maintenance tickets or booking amenities.

This tutorial shows how to build a voice AI agent for property management with Python. The AI-powered property manager uses wake word detection, Porcupine Wake Word, for hands-free activation, and intent recognition, Rhino Speech-to-Intent, for understanding voice commands — both running entirely on-device.

Conventional voice assistants use a two-stage pipeline for intent recognition: cloud-based Automatic Speech Recognition (ASR) converts speech to text, then Natural Language Understanding (NLU) extracts intent. This process means ASR transcription errors can compound into failures in intent understanding, and each service requires separate API management and cost considerations. Rhino Speech-to-Intent combines these steps into a single end-to-end model, eliminating the transcription step that can cause errors, and simplifying the architecture of AI-powered property managers.

Moreover, tenant interactions contain sensitive personal information. Cloud-based APIs send voice data to external servers, raising compliance concerns under regulations like CCPA and GDPR. On-device processing keeps voice data within the property management software. This reduces liability for property managers handling confidential information and ensures that sensitive data remains private.

Train a Custom Wake Word for PropTech AI App

Sign up for a Picovoice Console account and navigate to the Porcupine page.
Enter your wake phrase such as "Hey Summit Towers" or "Hey Assistant" and test it using the microphone button.
Click "Train," select the target platform, and download the .ppn model file.

For tips on designing an effective wake word, review the choosing a wake word guide.

Define Voice Commands for Property Management Automation

Create an empty Rhino Speech-to-Intent Context.
Click the "Import YAML" button in the top-right corner of the console and paste the YAML provided below to define intents for the Property Management AI Agent.

Test model with the microphone button and download .rhn context file for your target platform.

You can refer to the Rhino Syntax Cheat Sheet for more details on how to build your custom context.

YAML Context to Automate Tenant Communication

context:
  expressions:
    maintenanceRequest:
      - "(@politeness) [schedule, create, submit] (a) (new) maintenance [request, ticket] for $issue:issue (in, at) [unit, apartment, suite] $pv.TwoDigitInteger:unitNumber"
      - "(@politeness) [schedule, create, submit] (a) (new) maintenance [request, ticket] for $issue:issue (in, at) $location:location"
      - "(@politeness) $urgency:urgency $issue:issue [issue, problem, repair] (in, at) [unit, apartment, suite] $pv.TwoDigitInteger:unitNumber"
      - "(@politeness) $urgency:urgency $issue:issue [issue, problem, repair] (in, at) $location:location"
      - "(@politeness) report $issue:issue (in, at) [unit, apartment, suite] $pv.TwoDigitInteger:unitNumber"
      - "(@politeness) report $issue:issue (in, at) $location:location"
      - "(I need) $urgency:urgency priority maintenance for $issue:issue"
    
    bookAmenity:
      - "(@politeness) [book, reserve, schedule] (the) $amenity:amenity [for, on] $date:date"
      - "(@politeness) [use, access] (the) $amenity:amenity [on, for] $date:date"
      - "(@politeness) [check, is] (the) $amenity:amenity available (on) $date:date"
    
    generalInquiry:
      - "[when is, what is] [trash pickup, rent due, office hours, parking rules, guest policy]"
      - "(@politeness) tell me about [trash pickup, rent due, office hours, parking rules, guest policy]"
    
    updateContact:
      - "(@politeness) [update, change] (my) [contact information, phone number, email]"
      - "I need to update [my, the] [contact, phone, email]"

  slots:
    urgency:
      - "emergency"
      - "urgent"
      - "routine"
      - "standard"
    
    issue:
      - "plumbing"
      - "heating"
      - "cooling"
      - "garbage disposal"
      - "dishwasher"
      - "refrigerator"
      - "oven"
      - "lock"
      - "door"
    
    location:
      - "lobby"
      - "parking garage"
      - "hallway"
    
    amenity:
      - "gym"
      - "swimming pool"
      - "lounge"
      - "rooftop"
    
    date:
      - "today"
      - "tomorrow"
      - "monday"
      - "tuesday"
      - "wednesday"
      - "thursday"
      - "friday"
      - "saturday"
      - "sunday"
      - "weekend"

  macros:
    politeness:
      - "please"
      - "can you"
      - "could you"

Set Up Python SDK for the Voice AI Property Management Agent

Install the Porcupine Wake Word Python SDK pvporcupine, Rhino Speech-to-Intent Python SDK pvrhino and Picovoice Python Recorder library pvrecorder.

pip install pvporcupine pvrhino pvrecorder

Enable Hands-Free Activation for the AI Agent

Set up wake word detection to run as a continuous background process:

import pvporcupine
from pvrecorder import PvRecorder

ACCESS_KEY = "${ACCESS_KEY}"  # Obtain from Picovoice Console
KEYWORD_PATH = "${KEYWORD_PATH}"  # Path to .ppn model

porcupine = pvporcupine.create(
    access_key=ACCESS_KEY,
    keyword_paths=[KEYWORD_PATH]
)

recorder = PvRecorder(frame_length=porcupine.frame_length)
recorder.start()

print(f"Listening for wake word...")
while True:
    pcm = recorder.read()
    keyword_index = porcupine.process(pcm)
    if keyword_index >= 0:
        print("Wake word detected. Ready for command...")
        break

recorder.stop()
porcupine.delete()

Process Tenant Voice Commands in Real-time

Extract structured intent data from the tenant requests:

import pvrhino
from pvrecorder import PvRecorder

ACCESS_KEY = "${ACCESS_KEY}"      # Obtain from Picovoice Console
CONTEXT_PATH = "${CONTEXT_PATH}"  # Path to .rhn context

rhino = pvrhino.create(
    access_key=ACCESS_KEY,
    context_path=CONTEXT_PATH
)

recorder = PvRecorder(frame_length=rhino.frame_length)
recorder.start()

print("Speak your command...")

while True:
    pcm = recorder.read()
    is_finalized = rhino.process(pcm)
    
    if is_finalized:
        inference = rhino.get_inference()
        
        if inference.is_understood:
            print(f"Intent: {inference.intent}")
            print(f"Slots: {inference.slots}")
            
            # Route to property management system
            handle_property_command(inference.intent, inference.slots)
        else:
            print("Command not understood. Please try again.")
        
        break

recorder.stop()
rhino.delete()

Route Voice Commands to Property Management APIs

Integrate voice command output with the property management software:

def handle_property_command(intent: str, slots: dict):
    """Route voice intents to property management system"""
    
    if intent == "maintenanceRequest":
        category = slots.get('issue')
        unit = slots.get('unitNumber') or slots.get('location', 'unknown')
        priority = slots.get('priority', 'routine')
        
        # Create maintenance ticket via property management API
        create_maintenance_ticket(
            category=category,
            unit=unit,
            priority=priority
        )
        
    elif intent == "bookAmenity":
        space = slots.get('amenity')
        when = slots.get('date')
        
        # Reserve amenity via booking system
        schedule_amenity_booking(
            amenity=space,
            date=when
        )
        
    elif intent == "generalInquiry":
        # Route to FAQ system
        handle_inquiry(slots)
    
    elif intent == "updateContact":
        # Trigger profile update workflow
        initiate_contact_update()

Complete Python Script for Property Management Voice Agent

Here's the complete Python script integrating all the components above:

# Property Management Voice AI Agent


import argparse
import os
from datetime import datetime

import pvporcupine
import pvrhino
from pvrecorder import PvRecorder


def handle_property_command(intent: str, slots: dict):
    """Route voice intents to property management system"""
    
    if intent == "maintenanceRequest":
        category = slots.get('issue')
        unit = slots.get('unitNumber') or slots.get('location', 'unknown')
        priority = slots.get('priority', 'routine')
        
        print(f"[{datetime.now()}] Creating maintenance ticket:")
        print(f"  Priority: {priority}")
        print(f"  Category: {category}")
        print(f"  Location: {unit}")
        # API call: create_maintenance_ticket(category=category, unit=unit, priority=priority)
        
    elif intent == "bookAmenity":
        space = slots.get('amenity')
        when = slots.get('date')
        
        print(f"[{datetime.now()}] Booking amenity:")
        print(f"  Space: {space}")
        print(f"  Date: {when}")
        # API call: schedule_amenity_booking(amenity=space, date=when)
        
    elif intent == "generalInquiry":
        print(f"[{datetime.now()}] Processing inquiry: {slots}")
        # API call: handle_inquiry(slots)
    
    elif intent == "updateContact":
        print(f"[{datetime.now()}] Initiating contact update workflow")
        # API call: initiate_contact_update()


def main():
    parser = argparse.ArgumentParser()

    parser.add_argument(
        '--access_key',
        help='AccessKey obtained from Picovoice Console (https://console.picovoice.ai/)',
        required=True)

    parser.add_argument(
        '--keyword_path',
        help='Absolute path to wake word model file (.ppn)',
        required=True)

    parser.add_argument(
        '--context_path',
        help='Absolute path to context file (.rhn)',
        required=True)

    args = parser.parse_args()

    print("Property Management Voice AI Agent")
    print("=" * 50)

    # Main loop for continuous operation
    while True:
        # Stage 1: Wake Word Detection (Porcupine)
        try:
            porcupine = pvporcupine.create(
                access_key=args.access_key,
                keyword_paths=[args.keyword_path])
        except pvporcupine.PorcupineError as e:
            print("Failed to initialize Porcupine")
            raise e

        keywords = list()
        keyword_phrase_part = os.path.basename(args.keyword_path).replace('.ppn', '').split('_')
        if len(keyword_phrase_part) > 6:
            keywords.append(' '.join(keyword_phrase_part[0:-6]))
        else:
            keywords.append(keyword_phrase_part[0])

        print('Porcupine version: %s' % porcupine.version)

        recorder = PvRecorder(frame_length=porcupine.frame_length)
        recorder.start()

        print('Listening for wake word... (press Ctrl+C to exit)')

        try:
            while True:
                pcm = recorder.read()
                result = porcupine.process(pcm)

                if result >= 0:
                    print('[%s] Detected %s' % (str(datetime.now()), keywords[result]))
                    break

        except KeyboardInterrupt:
            print('\nStopping ...')
            recorder.delete()
            porcupine.delete()
            break

        finally:
            recorder.delete()
            porcupine.delete()

        # Stage 2: Speech-to-Intent (Rhino)
        try:
            rhino = pvrhino.create(
                access_key=args.access_key,
                context_path=args.context_path)
        except pvrhino.RhinoError as e:
            print("Failed to initialize Rhino")
            raise e

        print('Rhino version: %s' % rhino.version)
        print('Context info: %s' % rhino.context_info)

        recorder = PvRecorder(frame_length=rhino.frame_length)
        recorder.start()

        print('Listening for command...')

        try:
            while True:
                pcm = recorder.read()
                is_finalized = rhino.process(pcm)

                if is_finalized:
                    inference = rhino.get_inference()
                    if inference.is_understood:
                        print('{')
                        print("  intent : '%s'" % inference.intent)
                        print('  slots : {')
                        for slot, value in inference.slots.items():
                            print("    '%s' : '%s'" % (slot, value))
                        print('  }')
                        print('}\n')
                        
                        # Route to property management system
                        handle_property_command(inference.intent, inference.slots)
                    else:
                        print("Didn't understand the command.\n")
                    break

        except KeyboardInterrupt:
            print('\nStopping ...')
            break

        finally:
            recorder.delete()
            rhino.delete()


if __name__ == '__main__':
    main()

Run the Voice AI Agent for Property Management

Run the script from your terminal with the following commands. Copy your AccessKey from the Picovoice Console and replace the placeholder values with your actual ACCESS_KEY, KEYWORD_FILE_PATH and CONTEXT_FILE_PATH.

python3 voice_service.py \
  --access_key "$ACCESS_KEY" \
  --keyword_path "$KEYWORD_FILE_PATH" \
  --context_path "$CONTEXT_FILE_PATH"

Start Free