Patrol Sight¶

Sales & Business Kit — Master Index · HumanityAI'd — July 2026

Confidential — A–Z package to take the product to market

Download

What this is¶

Everything needed to sell Patrol Sight to Middle East security buyers — from technical due-diligence to pitch decks, pricing, market and competitor intelligence, a target list, and an outreach playbook. Every document is provided as a finished PDF (decks also as editable PPTX; the target list also as CSV). Editable Markdown sources live in _src/, and the underlying research dossiers in _src/research/.

The 30-second story¶

Your officers patrol on foot. Your facial recognition lives on fixed cameras and reports to a control room. So when an officer stands three metres from someone on your watchlist, the officer is the last to know. Patrol Sight closes that gap: AR glasses stream to a server inside your building, which matches faces against your own watchlist and returns a colour-coded identification into the officer's lens in about 68 milliseconds. Nothing leaves the premises. Every match goes to a human to verify. It runs in production today against a 93,728-person index on a single GPU.

Folder map¶

01-technical/¶

Technical-Due-Diligence-Report.pdf — buyer-grade review of the deployed system: architecture, measured performance, severity-ranked findings, licensing, and a 10–16 engineer-week hardening plan. Send at technical vetting — including the parts that don't flatter us.
Deployment-and-Ops-Guide.pdf — hardware sizing, the wireless requirement, install, runbook, security and privacy posture. For customer IT and security operations.

02-product/¶

PRD-Patrol-Sight.pdf — full product requirements, with measured figures against every non-functional target.
BRD-Patrol-Sight.pdf — business case, revenue model, stakeholders, risks.
POC-Playbook.pdf — the 6-week pilot blueprint, with the information-security gate and a shadow-running week. The core conversion tool.
Feature-Matrix-and-Roadmap.pdf — what ships today, honest limits, v1.2→v2 roadmap.

03-market/¶

Market-Analysis-MENA.pdf — the GCC security-capex opportunity, sourced.
Competitor-Analysis.pdf — NEC, IDEMIA, Corsight, Oosto, the Chinese wearable vendors, Clearview, and the Dubai precedent that proves the category.
RFP-Response-Template.pdf — reusable tender response, including the administrative envelope per market.
RFP-Response-Sample-Dubai-Profile.pdf — a worked example on a flagship policing profile.

04-sales/¶

One-Pager.pdf — first-touch leave-behind.
Master-Pitch-Deck.pptx / .pdf — the core deck.
Dubai-Police-Profile-Pitch.pptx / .pdf — tailored flagship variant.
Case-Study-Internal-Deployment.pdf — the proof, with real metrics and the production problems we hit.
Pricing-and-Packaging.pdf — indicative licence + hardware + services.
Demo-Script.pdf — the live-demo talk track, the limits section, and objection handling.

05-contacts/¶

ME-Security-Targets.csv — 35 rows: buyers, integrator partners and channels, sortable by priority, country and procurement channel.
ME-Security-Targets-and-Contacts.pdf — priority buyers, procurement channels per market, and the partner map.

06-package/¶

00-START-HERE-Master-Index.pdf — this document.
Outreach-Sequence-Guide.pdf — buyer sequencing, what to send when, email skeletons, objection handling.

Honesty notes (read before sending anything)¶

The Due-Diligence Report deliberately discloses deployment-grade security debt (default credentials, a committed API key, plaintext wearable transport, published datastore ports) and the model-licensing item, each with a costed fix. Keep it as-is for technical buyers. In a security sale, undisclosed risk is the one thing that ends the relationship permanently.
We do not hold a NIST FRTE ranking. Never imply otherwise. The substitute we offer is a fully measured pipeline plus accuracy measurement in the customer's own environment during the pilot.
State the recognition envelope: reliable to ~5 m, 6–7 m in good light, 8–10 m in long-range mode. Anyone promising more from a head-worn camera is describing a fixed-camera problem.
Concurrent devices per GPU is unmeasured. Do not estimate it. Measure it in the pilot and contract on that figure.
All pricing is indicative for planning, not a quotation.
Contacts contains no individual names, by design — security-sector person-level discovery runs through supplier registration, integrator introductions and events, not scraped lists. Nothing is fabricated.
Market and competitor facts carry public sources in _src/research/market_dossier.md, which also lists the items marked "to confirm" — including the reported scale of Dubai Police's smart-glasses work, sourced only to trade press. Verify before citing to a customer.
Never pitch covert use, automated enforcement action, or emotion/demographic inference. None exist in the product and all three are deliberately out of scope.

Product status at time of writing¶

Patrol Sight runs today as a live 34-service Docker Swarm stack on a single RTX 4060 Ti sharing the GPU with other production workloads: 93,728 enrolled persons, glasses-to-label recognition measured at ~68 ms typical (49–103 ms) with 0% packet loss, 14–20 effective recognition fps, watchlist search P95 8.5 ms over 1,931 real searches, and a two-layer tracker eliding 97.6% of vector searches. Measured headroom to 13 million faces at 80–130 ms. The same stack, tooling and console a customer receives.

Two gates before revenue¶

Model licensing — the recognition weights derive from a research-licensed lineage; resolve before any paid production deployment.
Security hardening Phases 0–1 — credential rotation, per-device tokens, network-surface reduction, encrypted wearable transport (~4–5 engineer-weeks) before any government operational deployment.

Discovery, demos and evaluation pilots can proceed now. Both gates are detailed in the Technical Due-Diligence Report.