Platform

How IndoAI Works

IndoAI turns a plain-language description of your site into a working AI camera system. The AI Advisor designs the layout and bill of materials, you order it in one transaction, partner OEMs ship to site, installation is arranged and managed by IndoAI, and app.indo.ai becomes the single screen your team uses daily.

Dr Vivek Gujar, Chief Strategy Officer · 27 August 2026 · 12 min read

The problem is not the technology. It is that nobody owns the outcome

A facility manager does not wake up wanting cameras. They wake up wanting to stop shrinkage at the back door, or to know that the loading bay crew is wearing helmets, or to have an attendance record that survives an audit. That is the request. It is a business outcome, described in plain words.

What the market sends back is a parts list. A camera vendor quotes cameras. A recorder vendor quotes an NVR and storage. A cabling contractor quotes cabling and mounts. A software reseller quotes an annual licence. A system integrator quotes to make the five of them talk to each other, and an install crew quotes to physically mount it. Somewhere in there a site survey happens, the quote gets revised, and six weeks have gone.

At no point in that chain does anyone take responsibility for whether the original outcome is achieved. The camera vendor's obligation ends at the camera. The integrator's obligation ends at integration. If the door camera is mounted too high and every face that walks through is unusable, that is nobody's contractual failure. The buyer discovers it months later, and by then there are five parties to argue with.

The fragmentation is not a failure of any one vendor. It is the shape of the industry. Each participant sells a component because a component is what they make. The buyer is left holding the integration risk, which is precisely the risk they are least equipped to carry.

IndoAI exists to take that risk off the buyer. One conversation, one order, one screen, one number to call. The rest of this page explains mechanically how that happens.

How IndoAI works, end to end

1. Scope it

   You describe the site in plain language on the AI Advisor. Type of premises, rough area, what worries you, and optionally photos, a video walkthrough or a floor plan. The Advisor returns a design: where cameras go, what each one can actually resolve at that distance, blind spots measured in square metres, an itemised bill of materials, storage sized on real bitrates, and an installation plan.

   One conversation

2. Order it

   Every line in that design is priced and orderable. Cameras, the Edge Box, accessories, the AI models you need, and installation go into one cart and one purchase order. There is no separate negotiation with a hardware seller and a software seller.

   One order

3. Deliver and install it

   The order routes to the OEM partner who supplies that line, and the goods ship directly to your site. IndoAI tracks the delivery and keeps you updated. Once material is on site, installation is arranged and managed by IndoAI: survey and marking, cabling and mounts, configuration and aiming, and a documented acceptance test before handover.

   Delivered and installed

4. Go live

   app.indo.ai becomes the screen your team opens every morning. Live feeds, incidents, evidence packs, routing to whoever owns the response, and a weekly report. It runs on the cameras we supplied and on the cameras you already had.

   One screen

5. Stay live, then grow

   Annual maintenance, health monitoring, spares and swaps sit under one contract with one number to call. When a new need appears, it is usually a new AI model on the same hardware rather than a new tender: a new zone, a new camera, a new site, same platform.

   One number

What the AI Advisor actually does

The AI Advisor is a solution engineer expressed as software. It has been built around the cameras IndoAI sells, the mounting rules that govern them, and the implementation patterns that have survived contact with real Indian sites.

Conventionally, getting to a design means an engineer visiting your premises two to four times, then days spent preparing a quotation, and weeks before you see a final number. The Advisor compresses that into a single session. It is not magic, it is codification: the geometry of camera coverage is arithmetic, and the rules for what a lens resolves at a given distance do not change from site to site.

Why coverage geometry is the part that matters

Most surveillance disappointment traces back to a decision made before anything was installed. A pixel-per-metre (PPM) figure tells you what a camera can genuinely resolve at a given distance. Detecting that a person is present needs relatively little. Recognising someone you already know needs more. Identifying a stranger from footage, to a standard that is useful in an incident review, needs roughly 80 PPM at the subject.

Under-specify the placement and the system fails exactly when you need it, because low light and motion blur eat whatever margin you had. The camera still records. The footage is simply useless for the question you are asking of it. This is the single most common and most expensive error in Indian CCTV deployments, and it is invisible until an incident.

The Advisor addresses it by designing coverage before hardware is chosen: every camera is traced to a named risk, so no camera exists without a job, and every blind spot is measured with a stated fix and its cost.

What the design report contains

The output is a working document, not a brochure. It contains three drawings and a straight answer on cost:

Alongside the drawings: a bill of materials that includes the recorder, drives and labour rather than quietly omitting them; storage sized on real bitrates with retention options costed; an installation guide with an acceptance test; and guidance on signage, retention periods and the rooms a camera should never see.

What happens after you place the order

This is the part of the process the industry never publishes, so it is worth being precise about.

The Advisor specifies on fit rather than on brand. Once you add the design to your cart, each line routes to the OEM partner who supplies it. IndoAI holds supply agreements with a small set of camera and accessory manufacturers, and orders go to the priority supplier for that line, who ships directly to your site. You are not chasing shipments; IndoAI tracks them and keeps you updated through delivery.

Installation follows delivery. It is part of the order rather than a separate problem you solve afterwards. IndoAI is appointing regional installation partners so that the crew arriving at your site is local to it, which matters for the follow-up visits that every deployment eventually needs. Whoever holds the spanner, the accountability stays with IndoAI: one contract, one number, no finger-pointing between parties.

It works on the cameras you already own

Most sites that need AI vision already have cameras. Ripping them out is expensive, disruptive, and usually unnecessary. The IndoAI Edge Box ingests RTSP and ONVIF streams, which means it can front-end an existing fleet regardless of who manufactured it. Analog estates connect through an encoder.

This is the practical reason the platform is described as an intelligence layer. The capture layer, meaning your cameras and recorders, stays yours and stays whoever you chose. IndoAI adds the layer above it that decides what is worth a human's attention. If you would rather buy the capture layer from IndoAI too, you can, and many customers do because a single vendor is simpler. Both routes end at the same dashboard.

Camera count often falls rather than rises after a proper design, because coverage designed on geometry replaces coverage accumulated over years of ad hoc additions.

The loop a recorder cannot close

A recorder records. That is the whole of its job, and it does it well. What it cannot do is act on what it recorded, which is where the operational value actually sits.

The loop matters more than any individual model. A detection that nobody acknowledges is not security, it is a notification. Closing the loop is what converts a camera estate into an operational system, and it is the reason the dashboard rather than the hardware is what customers open every morning.

Where your video actually lives

Inference runs on site, on the camera or on the Edge Box. Video does not leave your premises by default. This is an architectural choice with three consequences worth stating plainly.

It removes the bandwidth problem, because you are not pushing continuous streams to a data centre. It removes the latency problem, because a detection that matters in three seconds is not helped by a round trip. And it changes your compliance position, because personal data that never leaves the building is a considerably narrower obligation than personal data replicated to a cloud region.

India's Digital Personal Data Protection framework makes this concrete. Rule 4 commences on 13 November 2026, and organisations processing personal data through video systems will need to be able to describe what they collect, why, for how long, and on what basis. Facial recognition raises the bar further and generally warrants a documented impact assessment before deployment.

An edge-first architecture does not make you compliant on its own. Consent notices, retention policy, signage and the lawful basis for processing remain your organisation's obligations. What it does is remove the hardest category of question, which is what happens to Indian residents' video once it has been copied somewhere you do not control.

The honest limits

Every claim above has an edge. These are the ones worth knowing before you commit.

If a vendor tells you none of these constraints apply to them, that is worth more scrutiny than it usually receives.

Frequently asked questions

Related reading: Appization and installable AI models, the IndoAI Edge Box, and the AI vision glossary for the terms used on this page.