About Intsoft

Intsoft (Shenzhen) Technology Co., Ltd. is a technology company focused on industrial machine vision and the practical deployment of AI algorithms, and has worked for years on work-standard management for manufacturing lines. Intsoft is the English brand name.

What we do can be stated in one sentence: make the line know whether each operation was completed to standard. That looks simple, yet for decades there has been no good technical answer to it. Result quality is checked by inspection equipment, machine condition is captured by SCADA, but whether a person followed the SOP has relied on manual patrols and after-the-fact investigation. We wrote about why that gap exists from what we see on site.

What we focus on

• Applying industrial machine vision to work-standard management — no general security work, no consumer vision products.
• Packaging complex vision algorithms into configuration that process engineers can maintain themselves, so the barrier to use stays low.
• Solving problems in real production environments rather than chasing benchmark scores in a lab.
• Respecting the constraints of the shop floor: no line stoppage, no PLC rewrite, no extra work for operators. How that is achieved is described in the Visual AI SOP System.

Which industries we serve

How we deliver

How the capability was built

Our capability was not designed in one pass; it was added to one shop-floor failure at a time. That order explains why process validation and MES traceability are bound together in our solution — approaches that treat them as two separate purchases rarely survive contact with a real line.

1. Making it visible first. Getting the action at the workstation captured reliably: choosing the camera, fixing the mounting position, designing the lighting. This step sets the ceiling for everything that follows.
2. Then making it recognisable. Breaking the SOP into work elements that can actually be judged, and building recognition logic for each one. A standard that has not been broken down cannot be judged by an algorithm at all.
3. Then making the decision right. Comparing what was recognised against the standard sequence, handling missed steps, skipped steps and wrong order, and suppressing false alarms down to a density the floor will tolerate.
4. Then making it retrievable. Writing the decision back to MES with a timestamp. Without this step the system is just another alarm box as far as the quality department is concerned.

The team and how it is organised

Vision projects rarely fail on the algorithm; they fail because something on site was not caught. Our team is therefore organised around what a site needs rather than around the usual software-company split — most of the time is spent at customer sites, not in the office.

• Vision and algorithms — imaging design, decision logic and threshold calibration. These require seeing the real workstation before parameters can be fixed.
• Field delivery — installation, cable routing and coordination with the line equipment team, so that the line is not stopped, the PLC is not rewritten and operators gain no extra actions.
• Process consulting — translating a spoken work requirement into a judgeable process target. This is the step most often skipped, and the one that most often causes rework.
• After-sales support — calibration review, rule adjustment and version updates. Whether a system is still running a year later depends on this, not on the go-live day.

How we keep delivery quality honest

Delivery quality does not show in a demonstration; it shows in whether the system is still running three months after go-live. To make that a matter of record rather than a promise, we build several things into the process.

Questions we get about the company