Dubai’s Polynome launches agentic AI for real-world operations
The platform, which is now generally available, combines vision-language models (VLMs), a bilingual voice assistant and an agentic reasoning layer designed to transform data captured by cameras and sensors into operational actions
05 June, 2026
TT
16
Dubai-based artificial intelligence (AI) company Polynome has launched an agentic AI platform designed to provide organisations with real-time intelligence and autonomous responses across physical environments, targeting sectors including government services, hospitality, retail, banking and telecommunications.
The platform, which is now generally available, combines vision-language models (VLMs), a bilingual voice assistant and an agentic reasoning layer designed to transform data captured by cameras and sensors into operational actions.
Polynome said the platform is being deployed across government services, hospitality and retail, and is designed to help organisations make use of existing camera infrastructure without requiring model retraining or data labelling.
The company said enterprises have historically relied on conventional computer vision systems that require separate models for different use cases, creating fragmented systems that generate alerts but limited operational outcomes.
Polynome’s platform is delivered through two flagship applications.
Its Digital Floor Manager serves as a live operational layer built around existing camera infrastructure, monitoring passengers, customers and vehicles in real time. The system measures touchpoints against operational service-level agreements (SLAs) and can identify service-quality gaps, queue build ups, safety incidents, compliance breaches and dispatch requirements. The deployment also generates a real-time digital twin of the monitored environment.
AI Concierge
The company’s AI Concierge is a bilingual voice agent designed to replace legacy interactive voice response (IVR) and call-centre workflows. The Arabic-first model supports regional dialects and code-switching between Arabic and English and is integrated with customer backend systems to complete tasks such as balance enquiries, billing, top-ups, licence renewals, bookings, status updates, complaints and fraud alerts.
Polynome said its platform is built on an architecture that combines in-house vision-language models with object tracking technology. Operators can describe requirements in plain English or Arabic, enabling the system to configure itself without retraining.
The company said its voice assistant processes audio as a continuous stream rather than discrete user turns, allowing it to manage overlapping speech, pauses and language switching. End-to-end response time is 2.2 seconds.
The platform is designed to run on edge hardware located within customer facilities, processing live video without cloud connectivity.
Polynome said the system is fully on-premise, with no cloud dependency, third-party model API or telemetry leaving the facility, making it suitable for regulated industries where data residency and security requirements are critical.
Using cameras
“Every airport, mall, and warehouse already has thousands of cameras. The problem was never the hardware. It was that those cameras couldn’t think. With VLMs, you describe what matters in plain language, and the system sees it. With an agentic layer, the system acts on what it sees. That is the platform we are putting into production today,” said Alexander Khanin, CEO and co-founder of Polynome.
The company said the platform is being used in government and transport applications to replace legacy IVR systems handling millions of citizen interactions, while in hospitality it monitors premium lounges and traveller touchpoints against service-level targets.
In retail, the platform is used for footfall measurement, dwell time analysis, conversion tracking and queue management, while also supporting in-store voice assistance.
Banking and telecommunications customers are using the system for branch queue management, anti-money laundering (AML) acceleration, compliance monitoring and customer service applications.
Polynome describes itself as an agentic AI company focused on the physical world, with a platform designed to turn cameras, microphones and sensors into systems that can perceive, understand and act in real time.
The company’s research team includes co-authors of the DETR (end-to-end object detection with transformers) paper and contributors to research presented at conferences including CVPR, ICML, NeurIPS, ECCV, ICRA and IROS.
Polynome also operates the Polynome AI Academy, an enterprise education programme focused on physical-operations AI, and machines can see (MCS), an annual conference focused on computer vision, robotics and agentic AI.























