Industrial AI · Supplier selection
Industrial AI in China: how to assess a supplier shortlist
“Industrial AI” covers very different offerings: models, machine vision, software platforms and mobile robots. A useful shortlist starts with the process you want to improve.
What a list of 100 companies can tell you
The article “Top 100 Industrial AI Companies in China” introduces a dossier covering industrial models, machine vision and autonomous mobile robots. Such overviews help identify unfamiliar suppliers and product categories.
Inclusion in a list does not establish suitability for a particular plant. Check recency and methodology, as well as whether two suppliers are actually addressing the same task.
Sources for this section: [1]
Separate suppliers by their role in the process
A model provider, camera-system manufacturer and integrator may all contribute to one solution. Describe their responsibilities separately. The following classification is a working tool, not a ranking.
Scroll horizontally if needed.
| Role | Selection question |
|---|---|
| Model / AI software | Which task does the model solve with our data, and how is output quality assessed? |
| Machine vision | Which defects does the system detect on our parts under our lighting conditions? |
| Mobile robotics / AMR | How does material flow work at crossings and bottlenecks and when orders change? |
| Platform / integration | Who connects the systems, maintains interfaces and resolves faults? |
A reference needs to match the intended use
Ask for a deployment with comparable products, data and operating conditions. In visual inspection, for example, false rejects and missed defects need separate consideration. One “accuracy” percentage does not explain which errors are costly to the plant.
Include work beyond the model: data preparation, machine interfaces, training, monitoring and changes after a product changeover. Request a clear division between the supplier’s responsibilities and those retained by your team.
Evaluate the shortlist against one test brief
A shared test brief makes proposals more comparable. Describe the task, current effort and conditions for a trial. Share sensitive production data only after internal approval and agreement on usage rights.
- Record the success measure and baseline.
- Specify available data, interfaces and the test environment.
- Agree effort, responsibilities and failure handling.
- Decide on a bounded pilot and document the evidence still needed.
Sources
Publication dates and original sources. Company disclosures are not independent performance evidence.
- Sino-Cooperation · Top 100 Industrial AI Companies in China ↗Original language: German