Offer · Every pallet and parcel checked on camera: fewer picking errors, damage proven with a photo
KaizoFlow Supply Chain
KaizoFlow makes the flow of goods in your warehouse visible. Cameras and edge AI at your docks and packing stations count pallets and parcels, read labels, spot damage and compare what they see with your WMS. Every exception arrives with a photo as evidence, a person confirms it, and the damage report for the carrier is drafted for you. Images and stock data stay in your warehouse.
For whom
Warehouses and distribution centres, 3PL operators, couriers, ports and wholesalers with their own warehouse: first in the Szczecin–Świnoujście port area, where receiving, packing and dispatch errors are counted in złoty every month.
Runs on
AI Edge Box Camera at each dock or packing station, plus one AI Office Box Pro that collects the results, runs the language model and calculates the KPIs. Compare AI Box packages →
System diagram
Inputs, outputs and interfaces
What goes into the AI Edge Box at your docks and packing stations, what comes out, and through which interface. Images and stock data stay in your warehouse.
Your network: your data stays here
Inputs
Cameras at docks and packing stations
Images of every pallet, parcel and load, from fixed cameras above the dock and the packing table.
IP cameras (RTSP / ONVIF)Orders, deliveries and scans
What each order, delivery or load should contain, and the scanner reads at the station.
WMS API or EDI, barcode scannersDelivery notes and shipping documents
Delivery notes, CMR and packing lists, to compare with what actually arrived.
E-mail, scans or ERP export
AI Edge Box at each dock + AI Office Box Pro
- 1Image capture at the dock and the packing station
- 2Vision model: counts pallets and parcels, reads labels and SSCC codes, detects damage
- 3Comparison with the WMS: what the camera sees vs what the order says
- 4Exception with the photo as evidence, for a person to confirm
- 5Language model (Bielik) drafts the damage report and the carrier claim
Outputs
Exception at the station
Damaged, mislabelled, missing or extra items shown on the station screen with the photo; the operator confirms.
Station screen or handheldCorrect stock in the WMS
After confirmation the WMS is updated: received quantities, damage status, blocked items.
WMS APIPhoto evidence and carrier claim
Every load photographed at receiving and dispatch; the damage report and the claim drafted for a person to send.
PDF report, e-mailKPI dashboard
Picking errors, damage by carrier and supplier, time per exception and throughput per shift.
Web dashboard in your LAN
Learning loop: every exception the operator confirms becomes a new training sample, so the model keeps learning your goods, labels and packaging.
Interfaces
| Interface | Direction | Protocol / format | What passes |
|---|---|---|---|
| Cameras | Dock and station → Edge Box | IP cameras over RTSP / ONVIF, PoE | Images of pallets, parcels and loads |
| WMS | WMS ↔ AI Box | REST API, database view or EDI, depending on your WMS | Expected contents in; confirmed quantities and damage status out |
| Scanners | Station → Edge Box | Barcode and SSCC scanners (USB, Bluetooth or through the WMS) | Order, delivery and pallet IDs |
| Documents | E-mail / ERP → Office Box | E-mail inbox, PDF scans or ERP export | Delivery notes, CMR, packing lists |
| Operator screen | Operator ↔ Edge Box | Touch screen at the station or a handheld | Exceptions with photos, confirmations |
| Reports and claims | Office Box → your team | PDF and e-mail; HTTPS dashboard in your LAN | Damage reports, carrier claims, KPIs |
| Remote service | Our team ↔ AI Boxes | VPN, under the SLA | Model updates, retraining, support |
Where it works
Three checkpoints and one report
Receiving dock
Counts pallets and parcels, reads SSCC labels and photographs damage before anyone signs the delivery note.
Packing station
Checks the contents of each order against the WMS before the box is closed: wrong, missing or extra items are caught on the spot.
Dispatch dock
Photographs every load before it leaves, so you can prove what condition the goods were in when the carrier took them.
Damage report and claim
From the photos and the delivery note, a local language model drafts the damage report and the claim for the carrier or supplier. A person checks and sends it.
What we measure
KPIs we measure before the pilot and prove after it
We measure these with you before the pilot (the baseline), so the return can be proven in złoty afterwards.
| KPI | Definition | How it turns into money |
|---|---|---|
| Picking error rate | % of orders shipped with a wrong or missing item | Return, re-shipping and handling cost × number of errors; lost customers |
| Damage caught at receiving | % of damage found before the delivery note is signed | Damage you can charge to the carrier or supplier instead of paying for it yourself |
| Claims recovered | Value of damage claims accepted by carriers and suppliers | Direct recovery; a photo with a time stamp strengthens every claim |
| Time per exception | Minutes to find, document and report one exception | Time of clerks and team leaders |
| Throughput | Pallets or parcels handled per hour at a dock | Overtime and extra shifts in peak season |
| Inventory accuracy | Match between the WMS and the physical stock | Fewer stock counts, fewer lost sales |
Rollout
Five stages, each with a result you can check
- 11–2 weeks
Discovery
We choose one dock or packing station, agree the exceptions that cost the most and measure the baseline KPIs. Result: agreed KPIs and success criteria.
- 22–3 weeks
Data
Cameras installed, images of your goods, labels and damage collected and labelled with your team. Result: a training dataset.
- 32–3 weeks
Offline proof of concept
The model is trained and tested on recorded shifts. Result: an accuracy report against manual checks.
- 43–4 weeks
Pilot in shadow mode
The system runs alongside your team without changing the process. Result: hard data on errors caught and time saved.
- 54–6 weeks
Rollout
WMS integration, more stations, damage reports, training and handover. Result: a production system and a subscription.
Return on investment
Example calculation: a distribution warehouse with 50,000 orders a month
Illustrative assumptions; in the discovery call we replace them with your baseline numbers.
| Item | Assumption | Saving per year |
|---|---|---|
| Fewer picking errors | 50,000 orders a month, 0.5% with an error, PLN 150 per error; errors halved | PLN 225,000 |
| More damage recovered | Transport damage of PLN 300,000 a year; 40% more recovered thanks to photo evidence | PLN 120,000 |
| Less manual checking and paperwork | 2 FTE × PLN 90,000 on checks and damage reports; half of that time freed | PLN 90,000 |
| Total benefit | PLN 435,000 | |
| Solution cost (year 1) | Rollout PLN 150,000 + subscription PLN 96,000 | about PLN 250,000 |
Payback in about 7 months in this example. For the wider picture: McKinsey reports that companies which adopted AI early in the supply chain have 15% lower logistics costs, 35% lower inventory and 65% better service levels than their competitors.
“KaizoFlow: what you can see, you can fix. Every pallet, parcel and damage made visible, with a photo as evidence.”
All our offers run on the same on-premise platform: an AI Box in your local network with your own models. Your data does not leave the company, which matters for GDPR, trade secrets and medical data. We take care of the hardware, updates and secure remote service (VPN, SLA).
Start with KaizoFlow Supply Chain
Book a short discovery call. We measure your baseline KPIs first, so the result can be proven in numbers.
Request MVP