What Dragonfruit actually is

Dragonfruit is an AI video intelligence layer that sits on top of the security cameras a site already owns. Instead of selling you cameras, it turns the IP cameras and video management system you already run into a set of purpose-built AI agents. Those agents cover security, loss prevention, store operations, and compliance, reading live video to flag what matters instead of leaving it to a person watching a wall of monitors.

The platform is camera-agnostic and built to work with existing VMS and camera systems, including sites with limited bandwidth or constrained infrastructure. That's the headline difference from a closed camera platform: you keep the cameras you have and add the intelligence on top instead of ripping out hardware to get modern AI.

Under the hood it uses a split architecture. Real-time inference runs on a small on-prem device so heavy video processing happens inside the building, paired with cloud services for aggregation, search, and reporting. The agents are tunable. Sensitivity, alert thresholds, and even recognition of specific products can be adapted to a particular store's risk profile.

What the agents do

The most developed part of the platform is retail loss prevention. The shoplifting agent flags suspicious behavior as it happens using posture, object, and movement analysis. The self-checkout and sales-analysis agents combine live video with point-of-sale data to catch scan-avoidance and loss across staffed lanes and self-checkout. Because it ties video to transactions, it lines up what the camera saw against what the register recorded.

Beyond loss prevention, the same overlay runs store-operations agents: traffic counting, dwell-time and queue analysis, and shopper-behavior insight from the cameras you already have. There are safety and compliance agents too, so a single layer can serve both a security goal and a merchandising one. For grocery, convenience, and multi-store retail, that consolidation is the appeal.

Where Dragonfruit fits, where it doesn't

We install both AI overlays and full camera platforms, so this is the unbiased read on when an overlay is the right call.

Where Dragonfruit is the right call:

  • You already own working cameras: Dragonfruit runs on existing IP cameras and existing VMS. If you have a fleet of Axis, Hanwha, or other working cameras, an overlay adds modern AI without a rip-and-replace camera project. Usually the cheaper path to the same outcome.
  • Retail loss prevention and self-checkout: The loss prevention agents are the strongest part of the platform: shoplifting detection on real-time behavior, plus self-checkout and scan-avoidance detection that ties video to point-of-sale data. Grocery, convenience, and multi-store retail are the clearest fit.
  • Low-bandwidth and many-site fleets: Real-time inference runs on a small on-prem device, so heavy video doesn't have to push to the cloud constantly. That makes it workable in bandwidth-constrained sites and lets a chain scale the same setup from a handful of stores to thousands, with central management.
  • One layer, several jobs: The same overlay covers loss prevention, store operations analytics, traffic and dwell, and safety or compliance. For an operator who wants security and merchandising insight from the same cameras, that consolidation is real.

Where it's the wrong pick:

  • Your cameras are too old or poorly placed: An AI overlay is only as good as the footage under it. Low-resolution cameras, bad angles, or heavy backlighting cap accuracy no matter how good the agents are. If half the fleet needs re-aiming or replacing first, fix the cameras before layering AI on top. We'll say so in the consultation.
  • You want a single closed end-to-end stack: If your IT team prefers one vendor for cameras, recording, and analytics with a single support line, a closed cloud platform like Verkada is a cleaner operational story than an overlay across mixed hardware. The overlay path trades that simplicity to keep the cameras you own.
  • Pure access control or alarms: Dragonfruit is a video intelligence layer, not an access control or intrusion platform. If the project is doors, readers, and credentials, that's a separate stack. We design those alongside video, but the overlay isn't the answer to an access problem.

How Tec-Tel installs and configures Dragonfruit

An overlay deployment is a different shape of project than a camera install. There's usually little or no cable to pull. The work is making sure the footage is good enough to drive accurate agents, sizing and connecting the on-prem inference appliance, wiring in the data the agents need, and tuning everything against the real site.

  1. Week 1: camera + network review: We inventory the existing cameras (resolution, codec, frame rate, mounting, sight lines), confirm the VMS or NVR in place, and read the network. Dragonfruit runs on what you own, so the work here is confirming the footage is good enough to drive accurate agents, not pulling new cable. Output: a written assessment of which cameras are agent-ready and which need re-aiming or a lens change first.
  2. Weeks 1 to 2: edge appliance + connection: Dragonfruit uses a split approach: real-time inference runs on a small on-prem device so video doesn't have to leave the building, paired with cloud for aggregation, search, and reporting. We size and place that appliance, connect it to the camera streams and existing VMS, and confirm bandwidth headroom. Low-bandwidth and multi-site fleets are a designed-for case here.
  3. Weeks 2 to 3: agents + POS data: We turn on the agents that match the site: shoplifting and self-checkout loss prevention, queue and dwell-time analysis, traffic counts, safety and compliance. Where loss prevention is the goal, we connect point-of-sale data so the system lines up video against transactions and flags scan-avoidance and self-checkout loss.
  4. Tuning against your baseline: Out-of-the-box sensitivity is rarely right for a specific store. We tune alert thresholds and sensitivity against the traffic and risk profile of the site so operators get real alerts, not a flood of false positives that trains them to ignore the system.
  5. Week 4: training + runbook: Loss prevention, store operations, and security staff trained on the dashboards, search, and alert workflows. We deliver a written runbook so the deployment doesn't depend on one person's tribal knowledge, plus a service-level agreement in writing.

The biggest predictor of whether an overlay succeeds is camera quality. A low-resolution or badly aimed camera caps accuracy no matter how good the agent is. That's why the first step is an honest camera review, not a software demo. If a chunk of the fleet needs re-aiming, a lens change, or replacement before the AI can perform, we'll tell you that up front rather than turn on agents that under-deliver.

How it fits the rest of your stack

Dragonfruit plugs into the cameras and VMS you already run, and into point-of-sale data where loss prevention is the goal. It doesn't replace your access control, intercom, or alarm systems. It's a video intelligence layer, so it lives alongside those rather than absorbing them. We design the overlay in the context of the whole site, so the cameras feeding Dragonfruit and the doors on your access platform are planned together, not as disconnected projects.

If you're weighing an overlay against a full camera refresh, that's a real decision worth modeling. A camera-agnostic AI layer keeps your existing investment and adds intelligence on top. A closed cloud platform replaces the cameras but gives you one vendor and one support line. We walk through both paths in the consultation. See AI video analytics for how we run overlays end to end, and Coram vs Verkada for the overlay-friendly vs closed-platform tradeoff. Comparing AI video vendors directly? Intenseye vs Spot AI covers two more overlays in the same category.

What Tec-Tel adds vs going direct

Buying Dragonfruit as software leaves the hard part, the install, as your problem: confirming the cameras are good enough, sizing and placing the on-prem appliance, connecting it to streams, VMS, and POS, tuning the agents so operators get real alerts, and training the team. Tec-Tel owns all of that. We're a 15-year nationwide integrator, and one accountable project manager runs the deployment from the first call through every site, using our own team plus vetted, supervised field technicians held to one spec.

We also bring multi-vendor honesty. We install AI overlays and full camera platforms, so we have no reason to push an overlay when a refresh is the better answer, or the reverse. If a Dragonfruit overlay on your existing fleet is the right move, we'll deploy it. If half your cameras need replacing first, we'll say that before you sign anything. The platforms and services we deliver are at AI software layer, and the full set of vendors we install is at all vendors.

A note on partner-status language. Tec-Tel installs, integrates, and configures Dragonfruit. We don't claim a specific Dragonfruit partner certification on this page. If you need a vendor-certified install for a contractual reason, ask in the consultation and we'll confirm the credentials we hold or pair the deployment with a certified partner where required.