What Lumana's own product, pricing and integrations pages say its edge AI platform does not do, written for security, loss prevention and multi-site operations teams running due diligence. Every constraint below carries the documentation behind it, the operational impact, and the workaround.
Every constraint below comes from Lumana's own current product, pricing, integrations and Insight pages at lumana.ai, checked on August 17, 2026. Where Lumana documents nothing, this page says so rather than guessing. Spot AI sells a competing platform, so nothing here rests on an anonymous source or an aggregated user rating, and four claims that circulate about Lumana are corrected on this page: it does have point-of-sale capability, it is camera agnostic in its own words, it is not cloud-only because the Core records on site, and it is no longer thin on operations, because Insight now publishes named agents for occupancy, dwell time, line counting, heatmaps, inventory, cook time, wait time and cleaning verification. Anyone still running those four as arguments is working from old material.
Camera reuse, on-site recording, the detection catalog and the response layer are all published. The route to a detection nobody ships is the part that is not.
A constraint list is only useful next to an honest account of the product.
Lumana names notifying personnel, broadcasting audio through a loud speaker, triggering lockdowns and pushing actions to third-party systems over API or GPIO, plus emergency dispatch combined with human verification. A large part of this category still stops at an alert and leaves the rest to a partner, and Lumana can disprove that reading of itself from its own product pages.
On the security side, weapon, fire, fall, violence, face, people, vehicle and license plate recognition are published as standard. On the operations side Insight names occupancy, dwell time, point-of-sale, line counting, heatmap, inventory, cook time, wait time and cleaning verification as agents. The catalog reaches well past security, which is more than most platforms in this category publish.
Lumana states pricing is an annual recurring license determined by the number of camera feeds, days of storage and license terms, with the option to pay yearly or upfront, and that all features and capabilities come standard to avoid different licensing tiers. It publishes unlimited cameras, locations and users, third-party camera support, and a lifetime warranty on Lumana hardware for as long as you are a customer.
Each one is a consequence of how the platform is designed rather than a defect. What matters is whether it collides with your starting point.
Lumana publishes a continuously learning model and over 100 possible alert triggers. No self-serve builder, no request process and no time to a new detection appears on the pages this page read. The documentation covers what the platform already recognizes, not how something new gets added or by whom.
Every question specific to your own process becomes a vendor conversation with an unknown lead time. An operations team that keeps a queue of requests, such as dwell at one dock door or a step in a store opening routine, has no published way to work out whether that queue moves in weeks or in a release cycle.
Take your list of open requests into the evaluation and ask three questions in writing: who builds the next detection, through what process, and by when. Then ask for two of them to be closed during a trial, because the answer that matters is the one you can time rather than the one in the reply.
Lumana states the Core device sits on site, replaces the DVR or NVR and stores the video on the device, with Lumana Cloud and VMS+ above it. Pricing is published as an annual recurring license determined by the number of camera feeds, days of storage and license terms. So retention is a commercial input rather than a setting, and the recorder is a box per location.
A 40-site estate is 40 Core devices to specify, rack and eventually refresh, and a retention obligation that moves from 30 to 90 days moves the license at every one of those sites at once. Both are honest consequences of a clean architecture, and both belong in a three-year model rather than a first-year quote.
Establish the real retention requirement per site type before the quote is built, since a claims process at one site rarely justifies the same figure everywhere. Model the Core refresh cycle inside the same three years as the license, and ask what happens to retention pricing at renewal.
Point-of-sale is published as one of Lumana's Insight agents, sitting alongside occupancy, dwell time, line counting, heatmap, inventory, cook time, wait time and cleaning verification. The point-of-sale platforms themselves are not named on the pages this page read. The capability is documented; which registers it speaks to is not.
For a convenience, fuel or specialty retail operator that is the question the retail case turns on. A chain running an older till version, or two systems across acquired stores, can approve a platform on a capability that is real and then spend the first quarter establishing whether it reaches their own registers.
Name your point-of-sale vendor and version in the first message and ask for written confirmation, plus one reference running the same combination at a similar store count. Where two systems exist across the estate, ask which one is supported today and what the second one needs.
Lumana publishes emergency dispatch combined with human verification. That pairing is deliberate and it is the right design, because verification is what keeps a false dispatch from reaching a responder. What the pages do not state is who performs the verification, at what hours, or under what response time.
The automated responses on the platform, audio, lockdowns and API or GPIO actions, run in software. The dispatch path runs through people, so its coverage and its cost live in a service arrangement rather than in the license, and an estate assuming 24 hour dispatch should confirm that rather than infer it.
Ask four questions and put the answers in the contract review: who verifies, during which hours, under what target response time, and whether it is included in the annual license or bought alongside it. Confirm the same for weekends, which is when the difference usually shows.
The same four constraints in one view, sized to paste into an evaluation document.
Swipe the table sideways to see every column.
Lumana data comes from Lumana's own product, pricing, integrations and retail pages at lumana.ai, checked on August 17, 2026. Gaps are marked as not publicly specified.
If the questions on your sites are the ones Lumana already ships, and the register is one of them, most of the constraints above stop applying, because you are buying a catalog that already covers you. Camera reuse is the design rather than a bridge product, the Core keeps video in the building, the response layer is published down to lockdowns and API actions, and a single license tier with a warranty attached keeps the commercial side simple to compare.
These two platforms agree on more than they disagree on, and a comparison that hides that is not worth reading. Both are camera agnostic, both keep full-resolution video in the building, both publish a response at the moment of an incident rather than only a notification. Camera reuse and on-site recording are not wedges against Lumana, and anyone using them as such is arguing with a page Lumana can produce.
Where the two genuinely part company is who adds the next detection and how fast. Spot AI ships 15+ pre-trained Video AI Agents across vehicle break-in, fire, cash register theft, after-hours intrusion, personal protective equipment, forklift near-miss, falls and crowding in hazard zones, and Iris builds anything not on that list in natural conversation in about eight minutes, which turns a request queue into an afternoon rather than a roadmap conversation. Full-resolution video stays on the Intelligent Video Recorder in the building and only event metadata crosses the network.
Take your own list of open requests into a fortnight of overlap on one site, close two of them, and time each one. That answers what a feature table cannot.
None of that makes Spot AI the right answer for every buyer. Lumana publishes point-of-sale as a shipped Insight agent where Spot AI wires transaction data through open APIs, webhooks and an MCP endpoint instead, so a retail operator whose whole case sits at the register will find more of it already built on Lumana's side. Lumana also publishes a commercial model in detail, including days of storage as an explicit input, all features standard with no tiers, unlimited cameras, locations and users, and a lifetime warranty on its hardware, and it publishes emergency dispatch, which Spot AI does not. A constraint list earns its keep by holding each platform's shape against your starting point, and on this pair the shapes are closer than either marketing page suggests.
A live pilot on your cameras answers in a week what a spec sheet cannot.
Customer-reported outcomes from named Spot AI customers.
Blackmon Oil runs a vehicle loitering filter to keep parking lots clean and safe at its around-the-clock stores, with a smaller overnight crew.
Unique Industries covers more than a million square feet with a three-person safety team, catching near misses and falls on the cameras already installed.
Cambridge City cut footage search from two hours to 30 seconds after consolidating seven municipal locations on one platform.
"The Agentic AI is incredible technology and I love the unified dashboard."
Four, all documented: there is no published route to a detection outside the catalog, and no stated time to build one; the Core is a device per site and days of storage is one of the three inputs to the annual license; point-of-sale is published as an Insight agent while the till platforms themselves are not named; and emergency dispatch runs through a human verification step whose owner, hours and response time are not stated.
Yes, and this is worth stating plainly because the opposite still circulates. Lumana states it is completely camera-agnostic and compatible with any IP camera that supports standard streaming protocols, and its retail pages state it works with every camera you already have with no rip-and-replace. There is no gateway appliance to buy per site before the AI can see anything. The Core replaces the recorder rather than adding a bridge in front of the cameras.
Yes. Point-of-sale is published as one of Lumana's Insight agents, alongside occupancy, dwell time, line counting, heatmap, inventory, cook time, wait time and cleaning verification. Any comparison claiming Lumana has no point-of-sale capability is out of date. The open question is narrower: the point-of-sale platforms themselves are not named on the pages this page read, so confirm your own till system and version during the evaluation.
Lumana publishes the automated responses rather than leaving them to an integrator: notifying personnel, broadcasting audio through a loud speaker, triggering lockdowns and pushing actions to third-party systems over API or GPIO, plus emergency dispatch combined with human verification. That is a genuine response layer and one of the stronger parts of the platform. The part to scope is the dispatch path, since verification runs through people rather than software.
It depends how often your team asks for something new. If the shipped catalog covers your questions, Lumana is hard to fault on architecture or response. If the next detection is specific to your own process and nobody wants to wait for a roadmap, a platform such as Spot AI fits, because Iris builds a custom detection in natural conversation in about eight minutes and 15+ pre-trained Video AI Agents already span security, safety and operations on any ONVIF or RTSP camera.