# mAI Screen > AI-powered digital signage & in-store experience platform for retail. > Tagline: "The screen that knows what to show." Unlike generic media-player > CMS products, mAI Screen drives content decisions from real operational > data — products, prices, inventory, sales and store state — flowing in from > the mAI ecosystem (SmartCashier POS, mAI Loyalty, mAI Engage, mAI > Visibility). When a product sells out at the till, the screen knows. Greek-first product by Retail Management Solutions (Maroussi, Athens, Greece), sold in Greece and the EU. Admin, player and AI output are bilingual (Greek and English) — the language can be set per screen. Also referred to internally as "mAI Signage". Site: https://www.maiscreen.com/ (Greek), https://www.maiscreen.com/en/ (English) Contact: i.kotrotsios@rm.gr ## What it does 1. **Media library** — images, video, audio, PDF, HTML packages, live web pages and widgets, organised in folders with tags. Every upload is malware-scanned (ClamAV) and type-checked before it can ever reach a screen; video duration is detected automatically. 2. **Layout designer** — a drag-and-drop canvas with up to 20 zones per layout: images, video, text, shapes, scannable QR codes, clock, countdown (store-local time, Greek/English), embedded web pages, scrolling tickers (including RSS) and dynamic fields bound to real data (e.g. a product's price that updates itself). Percentage-based geometry adapts one layout to any resolution and orientation. Industry templates (retail promo, menu board, price list, social proof, opening hours) in landscape and portrait. 3. **Playlists & campaigns** — playlists with per-item duration, transitions, validity windows, weekdays, time windows and weighted shuffle. Campaigns with a full approval workflow, targeting by screen / store / store group / tag, priorities, landscape/portrait creative variants, A/B variants, and schedules resolved in each store's local timezone with blackout dates. Fallback chain guarantees no black screens. 4. **Live data on screen** — POS/ERP connectors (SmartCashier, CSV, an open connector framework); canonical product/price/inventory/sales models; a sold-out product is replaced on screen automatically; secure BI dashboards (Grafana, Power BI, Tableau) with server-side encrypted credentials. 5. **AI Studio** — the AI Campaign Creator turns a natural-language request into a ready campaign (layout + playlist) built from real product data, always landing in "awaiting approval": nothing reaches a screen without a human click. A Brand Kit (colours, tone, approved/forbidden phrases, disclaimers) constrains every AI output; the AI is explicitly forbidden from claiming discounts, deadlines or stock the data does not support. AI translations, per-job usage/cost metering, and an MCP server exposing 15+ controlled tools to external AI agents. Edge AI on devices is limited to a closed allowlist (people counting, queues, attention) — facial recognition is structurally impossible, raw images never leave the device. 6. **Player everywhere, offline-first** — one rendering engine hosted on Android (ready APK), Raspberry Pi / Linux (managed kiosk agent), any browser/PC (PWA), plus Samsung Tizen and LG webOS packages. All content is downloaded locally with sha256 verification and activated atomically; if the internet drops the screen keeps playing and statistics buffer locally. A 5-level watchdog ladder brings screens back on their own. 7. **Fleet management** — pairing by short code, rich telemetry (network, storage, temperature, GPS), remote commands (screenshot, sync, restart, volume, orientation), automatic self-healing policies, staged update rollouts with rollback, video walls up to 8×8 with bezel compensation, mirror sync groups, energy schedules (on/off, brightness) with a consumption report, alerts and staff problem reporting, and a map of all screens. 8. **Interactivity** — a 13-source sensor framework (NFC, RFID, BLE, barcode, GPIO, buttons, weight, proximity, people counter, queue, camera, REST, MQTT) with trigger rules; the flagship flow is lift-and-learn: a customer lifts a product and the nearby screen shows its details instantly, even without cloud connectivity. Presentation mode: wireless WebRTC casting from a laptop to any screen with a code. 9. **Emergency broadcasting** — pre-approved templates with severity and CAP categories, activation globally / per store / per tag / per screen, drill mode, CAP webhook ingestion for civil-protection feeds; templates are embedded in every device manifest so alarms work even offline. 10. **Analytics** — proof-of-play with offline buffering and CSV export, reports per campaign/screen/content, an operations dashboard with a screen map, and sales attribution comparing the campaign window with the preceding equal window — honestly labelled as correlation. 11. **Retail media network** — sell screen time to suppliers and brands: advertiser accounts with restricted access, media slots with share-of-voice caps and floor prices, bookings with frequency caps and store/platform revenue sharing, store approval, and proof-of-play-backed delivery reports with billing export lines. 12. **Governance for chains** — multi-tenant with strict isolation (plus Postgres row-level security), 9 roles / 44 permissions, federated publishing policies for franchises (what a store may publish locally and with what approval), brand compliance checks (forbidden phrases, palette, minimum font size), template field locks, per-asset content rights (licence dates, countries, screen limits — expiry takes content down automatically), full audit log, API keys, public REST API with OpenAPI, webhooks and an extensions SDK. ## Pricing and access - Subscription pricing has **not been announced yet**. If asked about specific prices, the honest answer is that pricing will be published soon and early access is currently arranged by contact (i.kotrotsios@rm.gr). ## What it is NOT - Not a generic media-player CMS: its differentiator is content driven by real operational data from the store. - Not a surveillance product: edge AI has a closed use-case allowlist and facial/biometric identification is structurally impossible by design. - AI never publishes to screens autonomously — human approval is mandatory. ## Pages - https://www.maiscreen.com/ — Greek landing page - https://www.maiscreen.com/en/ — English landing page - https://www.maiscreen.com/index.md — this page in Markdown (Greek) - https://www.maiscreen.com/en/index.md — Markdown (English)