Concept

LotSentry is an autonomous parking-enforcement agent for private lots — apartment complexes, HOAs, and small retail plazas — that plugs into existing security cameras, watches the lot 24/7, and runs the entire violation workflow end-to-end. Property managers only see a daily audit log and a small queue of disputes or tow-approval decisions; everything else — detection, matching, evidence capture, and first-notice enforcement — happens with no one watching.

How it works

• Property manager connects an existing IP/RTSP camera (or orders a $99 plug-and-play PoE camera kit) and draws zones on a snapshot: resident spots, visitor spots, fire lane, handicap, EV charging.
• A data-refinery onboarding step turns their messy permit list (spreadsheet, PDF, or property-management software export) into a structured, queryable permit database.
• An always-on vision agent (built on a cheap open multimodal model) reads plates, classifies zones, and cross-references the permit DB in real time to flag violations: no permit, expired permit, blocked fire lane, EV-spot misuse, suspected handicap placard fraud, visitor overstay.
• For clear-cut first offenses, the agent auto-generates an evidence packet (photo, timestamp, rule cited) and auto-sends a warning notice via text/email.
• For repeat offenders or towable violations, it queues a one-click approval for the property manager, who can dispatch a partnered local tow company directly from the dashboard.
• Faces and irrelevant PII are redacted at the edge before storage, keeping only plate/vehicle data relevant to enforcement — addressing privacy law and liability from day one.

Why now

Cheap, MIT-licensed multimodal models with long context (e.g. GLM-5.3-class) make accurate plate-reading and zone classification affordable without training custom models. Real-time video infrastructure (per-second metered streaming, built-in PII redaction) removes what used to be months of infra work, letting a small team ship the core loop fast.

MVP

A software-only product: connect one existing camera stream, upload a permit list, draw zones on a snapshot, and get a working violation dashboard with automated first-notice texts within a day of signup. No custom hardware, no on-site install crew — just a browser onboarding flow and a background agent.

Business model

Per-lot SaaS ($79–$149/month based on spaces monitored) plus an optional referral fee from partnered tow/enforcement companies for dispatched jobs.

Unique selling points

• Serves the massive underserved long tail (small private lots) that enterprise LPR vendors ignore.
• Zero enforcement staff needed — the agent works unsupervised and only escalates ambiguous or high-stakes cases.
• Privacy-by-design via edge redaction, reducing legal exposure versus raw camera footage.
• Buildable and launchable by a very small team using existing camera hardware and off-the-shelf multimodal AI, with a clear path to a hardware SKU later.

Concept

RoomCall AI turns the room-readiness inspection into a live, AI-guided video call instead of a manual walkthrough or after-the-fact photo upload. A housekeeper or inspector opens the app after cleaning a room, starts a real-time video session, and an AI agent — watching the live camera feed and listening to short verbal confirmations ("yes," "done," "clean") — walks them through the resort's custom checklist (bed, linens, minibar, bathroom, amenities, damage, safety equipment). The agent flags misses immediately ("I don't see a bath mat — please check") and, at call's end, auto-generates a structured, timestamped inspection record with photos attached.

Why now

Recent advances in real-time voice-agent infrastructure (low-latency handling of short utterances, per-second usage metering, PII redaction, and multi-participant session support) make it newly practical to run always-on, privacy-safe, live video+voice AI sessions on the floor of a hotel — not just async chat or batch photo review.

How it addresses resort pain points

• Replaces inconsistent, supervisor-dependent manual inspections with a standardized AI check on every single room turnover
• Structures previously fragmented inspection "evidence" (texts, WhatsApp photos, radio calls) into clean, queryable records
• Auto-routes flagged issues (stains, damage, missing amenities) as structured tickets directly into existing PMS / maintenance systems (e.g., Opera, Mews, ALICE, Quore)
• Builds a historical, room-by-room dataset of recurring issues to guide capex and refurbishment planning
• Reduces guest complaints/refunds caused by pre-arrival issues slipping through

MVP

• Mobile web app: start live video/audio session, AI agent narrates a resort-configurable checklist and listens for short confirmations
• Real-time vision model flags obvious misses (unmade bed, visible stain, missing item) during the call
• On session end: structured JSON record (pass/fail per item + timestamped photos) stored and exportable via CSV/API/Zapier into the property's existing PMS or maintenance ticketing tool
• Simple ops dashboard for supervisors to review flagged rooms and trends across the property

Unique selling points

• Camera + voice, not chat — an inspection tool built on atoms (rooms, linens, amenities), not documents
• Live-guided workflow catches issues in the moment, no second pass needed
• Privacy-safe by design (redaction of guest belongings/personal items in any stored footage)
• Usage-based, per-inspection-minute pricing model that scales naturally with resort occupancy
• Small, buildable MVP: one mobile flow, one AI checklist engine, one integration layer — deployable by a lean team using existing multimodal/voice-agent APIs, no custom hardware or model training required