Industries · Manufacturing
Manufacturing with clear data,quality checked consistently.
WEB-AZ builds AI solutions for factories and production workshops: in-line visual inspection, shift reports from existing data, verified supplier documents and procedures that are easy to find. Operators, engineers and the quality department decide faster, using information gathered in one place.
Where a factory can gain time and quality
Consistent visual inspection, not spot checks
Scratches, burrs, wrong labels or off-spec colours can slip through between two checks, and the criteria vary from one operator to another. Images kept for every part show the state of the batch at the end of the line.
Shift reporting without re-entry
Quantities produced, scrap, stoppages and their causes noted on sheets or in Excel, then re-entered into the ERP. Entered just once, digitally, the figures reach the manager the same day and are easy to compare between shifts.
Maintenance requests with a history
“Machine 4 is making a strange noise” is often said in the corridor or over the phone. A written request, with a photo and location, does not get lost, is not duplicated and leaves a history for preventive planning.
Supplier documents checked faster
Certificates of conformity, test reports, technical data sheets and declarations of performance arrive as PDFs with different structures. Values extracted and compared automatically with the specification shorten goods receipt.
Technicians’ know-how, within the team’s reach
The settings for a difficult part or the fix for a rare fault are often known only by the experienced technician. Documented and easy to search, they also help the shift when that person is away.
AI solutions for the production floor
We do not replace machine control systems, nor the quality decision. We add a layer that gathers information, makes it easy to find and flags what deserves a person’s attention.
- Visual inspection on the production lineCameras that photograph every part, a model trained on your own examples of good and defective parts, and an operator who confirms every rejection. The images are kept as evidence for quality analysis.
- Visual analysis right on the shop floorImage processing on devices installed next to the line, for a fast response and so that images of confidential products or processes never leave the factory.
- Equipment inventory from nameplatesA technician photographs the nameplate of a motor or pump, and its data lands in the maintenance records or the ERP, sorted into fields. Uncertain values are flagged for checking.
- Supplier document checksIncoming certificates and test reports are read, the relevant values are extracted and compared with the specification in the order. Deviations go to the person responsible for goods receipt or quality.
- Assistant for procedures and work instructionsOperators and technicians ask in everyday language and get the answer from work instructions, machine manuals and maintenance reports, with a reference to the source document.
- Production reports from existing dataAn agent that gathers data from the ERP, the digitised shift sheets and the stoppage log and prepares reports by shift, line and part, with explanations for large variations.
- Shop-floor apps and ERP integrationsSimple tablet forms for shift reporting and maintenance requests, connected to the ERP or the maintenance system, so the information is entered only once.
Use cases · Manufacturing
The examples below are conceptual and show how the solutions could work. They are not delivered projects.
Surface defects on injection-moulded parts
On a plastic injection moulding line, a camera photographs every part as it leaves the mould. Parts with suspicious marks are flagged on a screen next to the operator, who decides whether to reject them. The shift report shows defect types by hour.
Conceptual exampleA maintenance request sent from a tablet
An operator notices an oil leak. They scan the code on the machine with the tablet, take a photo and describe the problem by voice. The request reaches the maintenance system with the machine, the location and a suggested priority, and the team leader confirms it.
Conceptual exampleMaterial certificates checked at goods receipt
With every steel delivery, the inspection certificate is read automatically, and the composition and mechanical properties are compared with the requirements in the order. A batch with values outside tolerance is blocked until the quality department decides.
Conceptual example
Integration with ERP, MES and factory equipment
Factories often combine an ERP, older systems on the line and a lot of paper documentation. These are systems that integrations can be built with, and feasibility is checked for each project.
- ERP (for example SAP, Microsoft Dynamics, Odoo, Charisma)
- MES (manufacturing execution systems)
- CMMS for maintenance
- PLC and SCADA via OPC UA or Modbus
- Industrial cameras and sensors
- Quality management systems and controlled documentation
- SharePoint or file servers
- Excel and Power BI
Integration challenges
- Old machines with no data interface, where information has to be captured through additional sensors or entered manually
- A production network kept separate from the office network for security reasons, which limits what can communicate with the outside
- Lighting, vibration and product variation that require a period of testing on the line before visual inspection can be relied on
- Part codes and names that differ between the ERP, the technical drawings and what is actually used on the shop floor
Safety, quality and confidentiality in the factory
Employee monitoring
Cameras installed for quality control can also capture people. We point them at the product and the process, and any monitoring that may involve employees must be justified, proportionate and communicated, in line with GDPR and employment law.
The quality decision stays human
The system flags; it does not release batches. If you work under a quality management system, your procedures define who approves and how non-conformities are documented.
Separation from safety systems
AI solutions do not control machines and do not replace guarding devices or emergency stop systems. They run alongside them without affecting how they work.
Confidentiality of know-how
Technical drawings, formulations and process parameters are sensitive information. We can process images and documents locally or in controlled environments, with restricted role-based access.
Useful guides
Before you decide, read
- Computer visionComputer vision without facial recognition: applications in manufacturing, logistics and retailAnonymous counting, visual inspection, event detection: what computer vision can do without identifying people and which conditions need to be met.Read the guide
- Automatione-Factura automation: what you can automate in RO e-Factura and what you check by handRO e-Factura can be integrated with your invoicing software, accounting and internal workflows. What can be safely automated, what stays checked by a person and what to ask before you start.Read the guide
Frequently asked questions
How can AI help a factory or production workshop?
Through in-line visual inspection, with the rejection decision confirmed by the operator, production reports compiled from the ERP and digitised shift sheets, checks of supplier documents against the specification and an assistant that searches the work instructions. It does not control machines or release batches.
How many images does visual inspection need to work?
There is no universal number. It depends on how visible the defect is, how much the product varies and the lighting conditions. We start with a collection of images from your line, test and tell you honestly whether the result is good enough for real use.
Do images and production data leave the factory?
Not necessarily. For sensitive information we can run the processing on devices installed in the factory, with only aggregated results going to the cloud, if you want that. Access is granted by role and can be traced in the log.
Can we work with the ERP we already have?
In most cases, yes. We check at the outset what interfaces the ERP has, which data is reliable and what needs cleaning up. If there is no reasonable technical route, we propose an intermediate import format.
What is the recommended first step?
We choose a single measurable problem together, for example a recurring defect type on one line or shift reporting in one department. We run a limited test on real data and decide on that basis whether it is worth expanding.
Can’t find the answer? Write to us and we’ll reply directly.
Show us the line, not just the presentation
Describe your process, the defects you want to catch earlier or the report that is hard to put together. We will come with concrete questions and a proposal for a test on your shop floor.