The trigger moments that start this conversation, what actually changes on day one, and a four-phase plan that keeps every site recording throughout. Written for teams already running edge AI on the cameras they own.
Both platforms are camera-agnostic and both keep full-resolution video on site, so the physical layer does not change. The switch is about the layer above it.
Four moments when teams running an edge AI video platform start this conversation. If none of them are true, staying put is a reasonable answer.
Lumana publishes a long detection list and self-learning alerts from over 100 possible triggers, which covers most of what a security team asks for. What is not publicly specified is a self-serve path to a detection that is not on the list, or how long a new one takes. If your operations or EHS team is queueing requests that are specific to your process, ask both vendors who builds a new detection, how, and by when.
Lumana prices an annual recurring license by the number of camera feeds, days of storage and license term, with all features standard and no tiers. That is a clean model and it makes the renewal date the natural comparison point. Ask both vendors to quote the identical site list at the identical retention, and make sure recorder hardware is itemized on each side rather than folded in.
Lumana publishes automated responses that include broadcasting audio through a loud speaker, lockdowns, staff notification and actions pushed over API or GPIO, with emergency dispatch after human verification. The trigger is when a broadcast stops working on your sites, usually because the people it is aimed at have learned to ignore it, and what you need instead is something that addresses the person in the frame.
Lumana Core replaces the DVR or NVR and stores the video on the device, so a capacity or refresh conversation is really a platform conversation. Since retention is priced by days of storage on both sides, that is the moment to model three years of the actual obligation rather than the number somebody agreed to at signing, and to check what happens to the hardware if you do not renew.
The structural differences that decide whether a switch is worth starting.
Swipe the table sideways to see every column.
Lumana details come from Lumana's own site, pricing page and Response agent page, read on August 14, 2026. Where Lumana publishes nothing, this table says so rather than guessing.
A rollout that keeps coverage live throughout. Most sites go live inside six weeks.
Inventory the fleet per site: manufacturer, model, resolution, codec and network path, plus the Core device serving each site and the days of storage its license currently carries. Both platforms are camera-agnostic, so this phase is a coverage map rather than a replacement list. The useful work here is the second list: every detection your team has asked for and not got, because that list is the actual subject of the comparison.
Run one site live on both platforms at once. That overlap is the point: with both reading the same cameras, the comparison stops being architectural and becomes concrete, which is the only honest way to judge two edge AI platforms. Pick the site with the most open detection requests and try to close two of them during the pilot, one from the shipped agent set and one built in Iris, then compare how long each took.
Move site by site rather than estate-wide, keeping Lumana recording at each site until its replacement has run a full week. Cameras do not move and cabling does not change, so a site cutover is a configuration change plus an IVR. Leave the Core device racked through the overlap rather than pulling it on the day, and retire it per site only once that site has passed its sign-off.
Once coverage is stable, the work moves from parity to what the old platform could not do: adding operations and safety agents on the same cameras, building the detections that were stuck in the request queue, and setting what fires on site when nobody is watching. Teams that get value out of a switch do it here, not at cutover.
The six that come up in every one of these conversations.
The license is annual and recurring, so the renewal date is rarely far away and the comparison costs little before it arrives. Run assess and pilot during the overlap, on one site, and time the cutover waves to the renewal rather than paying twice across an estate. A pilot that runs while both platforms are live is also the only version of this comparison that proves anything.
No, and the plan is built to make that structural rather than a promise. Lumana keeps recording at each site until its replacement has run a full week, the cameras never move, and cutover happens site by site. If a site fails its sign-off test, that site stays where it is while the others proceed.
It stays on the Core device under the days of storage you bought, and it does not migrate. Export anything under a legal or insurance hold before the license lapses, because that window closes with the subscription rather than with the cutover. Treat the export list as a deliverable of the assess phase, not an afterthought.
Camera reuse and on-site recording are common ground, and neither is a reason to move. Three things change. Who can build a new detection and how quickly, since Iris takes about eight minutes in natural conversation. Whether the deterrence addresses the person in the frame or broadcasts to the area. And how far the shipped agent set reaches into operations and safety on the same fleet.
That cost is real and worth naming. The offset is that natural language search removes most of the scrubbing a video workflow used to need, so the skill being replaced is largely the one nobody enjoyed. Budget a fortnight of overlap in the pilot so the team compares on their own footage rather than on a demo reel, and let the people who use it daily call it.
Anyone answering without your site list is guessing. Both price by camera and by retention, which makes them genuinely comparable: quote the identical site list at the identical days of storage over three years, with recorder hardware itemized rather than folded into a bundle on either side.
Bring a camera list and we will tell you what works as it is.
Customer-reported outcomes from named Spot AI customers.
Staccato went from first conversation to full deployment across an 800-acre campus in seven weeks, on the camera infrastructure it already had.
Bridge33 Capital standardized video across 25 plus properties, each acquisition arriving with a different camera system, and cut footage search from hours to minutes.
The YMCA of Greater Richmond deployed 17 locations in two weeks and standardized retention across the estate.
"Within four minutes of the alarm going off, Spot AI gave us video footage of the incident on the responding officers' phones. We had five recoveries within the hour."
No. Lumana is camera-agnostic and so is Spot AI, so any IP camera that works with Lumana works with Spot AI as it is. Analog cameras connect through the Intelligent Video Recorder rather than being replaced. Nothing on the wall moves and no coax is pulled.
The Intelligent Video Recorder does the same job in the same place: it sits on site, writes full-resolution video in the building and lets only event metadata leave it. That means the network design, the bandwidth profile and the data-residency answer stay broadly as they are, which is what makes this migration a configuration change per site rather than a re-architecture.
Most estates are live inside six weeks: a week to assess, a fortnight of pilot on one site, then site-by-site cutover. Sites go live individually rather than in a big-bang switch, so the timeline scales with how many sites you have rather than how many cameras.
Yes, and the plan depends on it. The pilot runs one site on both at once so search, detection and response can be compared on identical footage. Lumana keeps recording at every site until that site's replacement has passed a full week of sign-off.
Both run AI at the edge on cameras you already own, so the difference is not architecture. It is the agent layer on top. Spot AI ships 15+ pre-trained agents across security, safety and operations, adds Iris so your own team can build a detection in about eight minutes, and answers an incident with AI Talk Down through standard speakers already on site.