AI Agent for Property Inspections: Checklist Management and Report Generation

Why Automate Property Inspections?

Property inspections happen at move-in, move-out, annually, and whenever maintenance concerns arise. Inspectors walk through units with a clipboard, photograph issues, and then spend an hour back at the office formatting a report. An AI inspection agent structures this workflow — generating checklists, categorizing photographed issues, and producing formatted reports instantly.

The Inspection Data Model

We start with a structured representation of inspections and their findings.

flowchart LR
    CALLER(["Buyer or Seller Lead"])
    subgraph TEL["Telephony"]
        SIP["Twilio SIP and PSTN"]
    end
    subgraph BRAIN["Real Estate AI Agent"]
        STT["Streaming STT<br/>Deepgram or Whisper"]
        NLU{"Intent and<br/>Entity Extraction"}
        TOOLS["Tool Calls"]
        TTS["Streaming TTS<br/>ElevenLabs or Rime"]
    end
    subgraph DATA["Live Data Plane"]
        CRM[("CRM and Notes")]
        CAL[("Calendar and<br/>Schedule")]
        KB[("Knowledge Base<br/>and Policies")]
    end
    subgraph OUT["Outcomes"]
        O1(["Showing scheduled"])
        O2(["Lead routed to agent"])
        O3(["Pre-qual handed to broker"])
    end
    CALLER --> SIP --> STT --> NLU
    NLU -->|Lookup| TOOLS
    TOOLS <--> CRM
    TOOLS <--> CAL
    TOOLS <--> KB
    NLU --> TTS --> SIP --> CALLER
    NLU -->|Resolved| O1
    NLU -->|Schedule| O2
    NLU -->|Escalate| O3
    style CALLER fill:#f1f5f9,stroke:#64748b,color:#0f172a
    style NLU fill:#4f46e5,stroke:#4338ca,color:#fff
    style O1 fill:#059669,stroke:#047857,color:#fff
    style O2 fill:#0ea5e9,stroke:#0369a1,color:#fff
    style O3 fill:#f59e0b,stroke:#d97706,color:#1f2937

from dataclasses import dataclass, field
from datetime import datetime
from enum import Enum
from typing import Optional

class InspectionType(Enum):
    MOVE_IN = "move_in"
    MOVE_OUT = "move_out"
    ANNUAL = "annual"
    MAINTENANCE = "maintenance"

class IssueSeverity(Enum):
    COSMETIC = "cosmetic"      # scuff marks, minor wear
    MINOR = "minor"            # small holes, loose fixtures
    MODERATE = "moderate"      # appliance issues, damaged flooring
    MAJOR = "major"            # structural, plumbing, electrical
    SAFETY = "safety"          # mold, fire hazard, code violation

@dataclass
class InspectionItem:
    room: str
    area: str  # walls, floor, ceiling, fixtures, appliances
    condition: str  # good, fair, poor, damaged
    notes: str
    severity: Optional[IssueSeverity] = None
    photo_url: Optional[str] = None

@dataclass
class Inspection:
    inspection_id: str
    unit: str
    inspection_type: InspectionType
    inspector: str
    date: datetime
    items: list[InspectionItem] = field(default_factory=list)
    overall_condition: str = "pending"
    tenant_present: bool = False

Dynamic Checklist Generation

Different inspection types need different checklists. The agent generates them based on the unit configuration.

ROOM_CHECKLISTS = {
    "kitchen": [
        "Countertops", "Cabinets (open/close all)", "Sink and faucet",
        "Dishwasher", "Stove/oven", "Refrigerator", "Microwave",
        "Floor condition", "Walls and ceiling", "Light fixtures",
        "Outlets (test GFCI)", "Under-sink (check for leaks)",
    ],
    "bathroom": [
        "Toilet (flush test)", "Sink and faucet", "Shower/tub",
        "Tile and grout", "Mirror and medicine cabinet",
        "Exhaust fan", "Floor condition", "Outlets (test GFCI)",
        "Under-sink (check for leaks)", "Caulking condition",
    ],
    "bedroom": [
        "Walls and ceiling", "Floor/carpet condition",
        "Closet (doors, shelves, rod)", "Windows (open/close, locks)",
        "Window coverings", "Light fixtures", "Outlets",
        "Smoke detector (test)", "Door and hardware",
    ],
    "living_room": [
        "Walls and ceiling", "Floor condition", "Windows",
        "Window coverings", "Light fixtures", "Outlets",
        "Thermostat", "Front door (locks, deadbolt)",
    ],
}

def generate_checklist(
    unit_rooms: list[str],
    inspection_type: InspectionType,
) -> dict[str, list[str]]:
    """Generate an inspection checklist based on unit layout."""
    checklist = {}
    for room in unit_rooms:
        room_key = room.lower().split()[0]  # 'Master Bedroom' -> 'bedroom'
        base_items = ROOM_CHECKLISTS.get(room_key, ROOM_CHECKLISTS["bedroom"])
        checklist[room] = list(base_items)

    # Add move-specific items
    if inspection_type == InspectionType.MOVE_OUT:
        checklist["General"] = [
            "All personal belongings removed",
            "Unit cleaned to move-in standard",
            "All keys returned",
            "Forwarding address collected",
            "Garage/storage cleared",
        ]
    return checklist

Photo-Based Issue Categorization

When an inspector photographs damage, the AI categorizes it automatically.

from openai import AsyncOpenAI
import base64
import json

async def categorize_issue_from_photo(
    image_path: str,
    room: str,
) -> dict:
    """Analyze a property inspection photo to categorize the issue."""
    client = AsyncOpenAI()

    with open(image_path, "rb") as f:
        img_b64 = base64.b64encode(f.read()).decode()

    response = await client.chat.completions.create(
        model="gpt-4o",
        messages=[{
            "role": "user",
            "content": [
                {
                    "type": "text",
                    "text": f"""Analyze this property inspection photo from the {room}.
                    Return JSON with:
                    - area: what part of the room (wall, floor, ceiling, fixture, appliance)
                    - issue: brief description of the problem
                    - severity: one of cosmetic, minor, moderate, major, safety
                    - recommended_action: what should be done to fix it
                    - estimated_cost_range: low and high estimate in USD""",
                },
                {
                    "type": "image_url",
                    "url": {"url": f"data:image/jpeg;base64,{img_b64}"},
                },
            ],
        }],
        response_format={"type": "json_object"},
    )
    return json.loads(response.choices[0].message.content)

Building the Inspection Agent

from agents import Agent, function_tool

@function_tool
async def start_inspection(
    unit: str,
    inspection_type: str,
    rooms: str,
) -> str:
    """Start a new property inspection and generate the checklist."""
    room_list = [r.strip() for r in rooms.split(",")]
    insp_type = InspectionType(inspection_type)
    checklist = generate_checklist(room_list, insp_type)

    output = f"Inspection started for unit {unit} ({inspection_type})\n\n"
    for room, items in checklist.items():
        output += f"**{room}:**\n"
        for item in items:
            output += f"  [ ] {item}\n"
    return output

@function_tool
async def record_finding(
    room: str,
    area: str,
    condition: str,
    notes: str,
    severity: str = "cosmetic",
) -> str:
    """Record a finding during an inspection."""
    return (
        f"Recorded: {room} > {area} - Condition: {condition} "
        f"(Severity: {severity})\nNotes: {notes}"
    )

@function_tool
async def generate_inspection_report(unit: str) -> str:
    """Generate the final inspection report for a completed inspection."""
    # In production, pulls all recorded findings from the database
    return (
        f"## Inspection Report - Unit {unit}\n\n"
        f"Date: 2026-03-17 | Inspector: AI-Assisted\n\n"
        f"### Summary\n"
        f"- Total items inspected: 47\n"
        f"- Issues found: 4\n"
        f"- Safety concerns: 0\n\n"
        f"### Issues Requiring Action\n"
        f"1. Kitchen - Faucet drip (minor) - Est. $75-150\n"
        f"2. Bathroom - Grout cracking (moderate) - Est. $200-400\n"
    )

inspection_agent = Agent(
    name="PropertyInspectionAgent",
    instructions="""You are a property inspection assistant.
    Guide inspectors through their checklist, record findings,
    categorize issues by severity, and generate reports.
    Flag any safety concerns immediately.""",
    tools=[start_inspection, record_finding, generate_inspection_report],
)

FAQ

Can the photo analysis detect issues that are hard to spot visually?

Vision models can identify obvious damage like cracks, water stains, mold, and broken fixtures reliably. Subtle issues like hidden water damage behind walls or electrical problems are beyond visual analysis — those still require professional inspection techniques.

How do you handle discrepancies between move-in and move-out inspections?

The system stores both inspection records linked to the same unit and tenancy period. A comparison tool diffs the two reports item by item, highlighting new damage that appeared during the tenancy. This comparison forms the basis for security deposit deduction decisions.

Is the AI-generated report legally sufficient?

AI-generated reports should be reviewed and signed by a licensed inspector or property manager. The AI handles data collection and formatting, but the human provides the professional judgment and legal accountability. Most jurisdictions accept digitally signed inspection reports.

#PropertyInspections #ReportGeneration #RealEstateAI #Python #ComputerVision #AgenticAI #LearnAI #AIEngineering