A warehouse record can be internally consistent and still point a picker at an empty slot. Someone may have moved the pallet without recording the move. Gather AI's starting point is an independent observation of the floor. Its drones and vehicle-mounted cameras collect evidence that staff can compare with the warehouse management system. The choice for a buyer is which differences merit a recount, a stock correction or a conversation with the customer who owns the goods.
- 01What gets observed Drone Vision reads racked storage. Cameras on material handling equipment capture other parts of the journey as goods move.
- 02What gets compared Gather matches a physical capture with the expected location, identity or quantity and gives the team evidence to review.
- 03What gets deployed A site study determines capture coverage and the fleet. Case counting and battery swapping have hardware and layout implications.
01 / ProductA second view of stock that is already on the books
Gather AI describes its Prana platform in three parts. Vision collects observations using drones and equipment such as forklifts. Sage helps teams interpret those observations alongside the warehouse management system, commonly called a WMS. Workflows routes exceptions to people who can investigate them. The distinction matters when a buyer hears that a warehouse has been scanned. A scan is a dated observation. Turning it into an inventory adjustment involves rules about timing, authority and sufficient evidence. Platform FAQ
A pallet's licence plate number identifies the handling unit. Its rack location is another piece of information, and the number of cases it contains is a third. A photograph of the label can help reconcile the first two while leaving the quantity unresolved. Gather's separate 3D Case Count feature reconstructs the stack from multiple views and compares the resulting count with the system record. That distinction gives the operator a reason to specify the type of count it needs. 3D Case Count
Capture method follows the physical layout. The company describes drone missions for racked locations and a vehicle-mounted option for activity in open areas. A drone can store observations while it is out of network coverage and transfer them afterwards. A manager looking at a dashboard needs to know when the shelf was observed as well as when the result arrived. A pallet moved between those events can make two correct records look inconsistent. Drone missions System integration FAQ
02 / AudienceWho gains from a photographic exception record
A third-party logistics provider has an extra audience for a recount. Its customer may want to understand why a stock adjustment appeared on a statement. A dated capture can give the customer and warehouse manager something specific to discuss. The useful outcome is a resolved discrepancy with a traceable explanation. Collecting additional images of a location that nobody has authority to adjust does little to resolve a claim.
A retailer operating both reserve storage and a busy pick area has a different question. The same pallet might sit untouched for days in reserve and then lose cases every hour once brought to the pick face. Counting both areas at identical intervals may spend capacity on the least informative observations. Gather describes scheduling missions by aisle or zone. An inventory lead can use that flexibility to put the most frequent checks where errors have the greatest chance of reaching a shipment. Coverage and scheduling
03 / WorkflowA proposed reconciliation of a partly depleted pallet
Select one zone where the team already conducts manual cycle counts. Record the expected pallet identifier, location, case quantity and the effective time of the WMS extract. Choose both full and partial pallets of a single stock-keeping unit. Gather documents that product restriction for its multi-angle case count and specifies depth sensing on the Mark II airframe for this feature. Confirm that the fleet offered for this site includes the corresponding hardware. Keep mixed pallets in a separate sample so they can still test location verification without being counted as eligible examples for that specific feature. Case counting and its scope
Run a mission during an agreed operating window and retain the observation times. When a case count differs, first check for intervening picks or a movement recorded late in the system. Ask a counter to inspect a sample without seeing the model's answer. Include matches as well as exceptions. Inspecting only flagged pallets can measure how often a flag is justified, but will miss errors in the apparently correct group.
Divide the results into actions that the team understands. A confirmed pallet at the wrong location needs a location correction. A short count may require investigating picking or receiving. A label that could not be read may need replacing. Each creates a different task, and each should have a named owner. Close the loop by checking that the next observation agrees with the resolution. A repeatedly unreadable label should not appear as a fresh, unexplained stock loss every morning.
04 / PricingWhat the site study needs to price
Gather's commercial entry point is a discussion of the customer's facilities and a platform demonstration. The FAQ describes a coverage study that considers rack geometry, number of pallet positions and desired scan frequency. The public pages reviewed do not give a unit tariff. Bring a map that distinguishes racked aisles from bulk floor storage, and separate the reporting requirements for each area. A single number for warehouse square footage leaves most of the costly design choices unanswered. Demo and commercial enquiry Deployment sizing
| Scope | What the supplier describes | Question for the proposal |
|---|---|---|
| Racked locations | Mapped drone missions and rack markers | Coverage frequency and number of devices |
| Battery handling | Optional automated swapper or manual battery changes | Staff availability for the desired window |
| WMS connection | Integration can return validated transactions | Setup, validation, support and change ownership |
Gather AI's site and deployment FAQ and demo route, consulted 2026-11-21.
Include the cost of acting on the results. Suppose a building has many easily corrected location mismatches during an initial rollout. Resolving the backlog can consume staff time just as the scanning process becomes more efficient. That is real work with potential value. Record it separately from ongoing review time. A sensible cost comparison uses the steady flow of new discrepancies while recognising the one-time effort required to establish a reliable baseline.
05 / DistinctionsA deployment that connects scanning with a real customer requirement
In Gather's GEODIS case study, a customer wanted more frequent counts. The account says four manual counters previously completed 800 location scans in a day and a single drone operator later completed 1,200. These are the supplier's figures for a particular facility. The especially useful detail is how the team looked for misplaced pallets. A historical capture gave it a last known location from which to investigate, even when the WMS could not reconcile the pallet identifier. GEODIS case study
The deployment illustrates the difference between observing and moving goods. Exotec is relevant when the warehouse needs to bring stock to a workstation. Gather can address a different need in the same building, such as identifying discrepancies before stock is requested. Zebra Technologies provides context for the devices and scanning workflows used at individual handling events. A transaction captured perfectly at dispatch still leaves an interval since the last observation in reserve.
06 / QuestionsWhich results can support an adjustment
Ask how the implementation marks a position that could not be observed. A missing image, an apparently empty bay and a clearly identified pallet are three distinct states. Collapsing them into a simple pass or fail can turn a coverage issue into a spurious stock correction. Obtain examples from the proposed site, especially at the top of racks, behind obstructions and where a forklift exposes a different face of the pallet from the aisle view.
Check when information flows back to the WMS. Gather says validated transactions can post through an API, while drone mission results synchronise after the mission. Those statements support different operational expectations. A supervisor releasing the last cases on a disputed order needs to know whether an intervening pick has been taken into account. Document who can authorise a change, what happens if the WMS rejects it and where another shift can see that the exception remains open. Transaction timing
07 / DecisionChoose the visibility gap first
Start with a genuine unresolved question about stock and the person responsible for answering it. The decision becomes much clearer once an operator can name the evidence that would settle a disagreement, the latest time at which that evidence would help, and the action that follows. Then the choice of camera, flight schedule or interface has an operational purpose.
Reserve storage
Test scheduled drone missions against the existing cycle-count process, including unreadable and empty positions.
Case-pick operation
Use eligible single-SKU partial pallets and establish an independent count before evaluating the 3D result.
Third-party logistics
Follow a discrepancy from first capture to an agreed customer-facing explanation and WMS correction.
A business worth understanding.
Suggest your business or one you find interesting. Tell us what you want to understand about its product, positioning, design or workflows.
Suggestions are free. Selection and publication stay with the desk.
- Platform and deployment FAQConsulted
- Drone VisionConsulted
- 3D Case CountConsulted
- GEODIS deploymentConsulted
- Commercial demo and scopingConsulted
- Case-counting method and eligibilityConsulted

