On-Device AI or the Cloud? Pick the Job Before You Pick the Model

2026 catalogues are full of “AI-ready” boxes, tablets and cameras. The useful split is quieter. Some jobs must finish on the device because the network, the delay or the footage should never leave the room. Other jobs are happier in the cloud, where a person can change the rule without touching every screen.

Click the job you actually have

The same chip can be the right answer in a hotel room and the wrong one above a school desk. Choose a situation. The panel below says what should stay local, what can live in the cloud, and the mistake teams make when they buy the feature before the workflow.

Why “send everything to the cloud” aged badly

A store or campus link is not a studio uplink. Video of a queue, a shelf or a classroom is also personal data. India’s data protection rules, and the plain cost of bandwidth, both push the same design: count or classify on the device, send a number or an alert, and keep the raw picture unless a person has a reason to see it. That is a product decision, not a model-size contest.

Cloud still earns its place. Playlists, policies, app updates and the words a manager changes at 9 a.m. belong in a console such as digital signage or DCM. The device should keep working when that console cannot be reached. Read why edge devices matter if you are still comparing a camera that only streams with a player that can decide locally.

Buy the failure mode, not the slogan

Ask one question in the pilot: what still happens when the WAN is down for an hour? If the answer is a black screen, a frozen queue board or a tablet that cannot open the class app, the AI feature was never the product. The product is a device with a local job, a policy, and a cloud that updates it when the link returns.

Maram plans that split with the hardware — Android players, commercial tablets and the software that manages them — so “AI-ready” means a job you can point at.