We recently changed the language we use at Zetamotion.
What we previously called a “partnership licence” is now called a Hypercare Subscription.
This is more than a marketing change. It reflects an important distinction between conventional software and AI-powered visual inspection on a working production line.
In a typical software-as-a-service model, the provider gives the customer access to a product. The features are available, documentation and technical support are provided, and responsibility for turning the software into an operational solution largely passes to the customer.
That model works when the problem is self-contained and the user can take ownership of implementation. Machine vision on the factory floor is rarely that simple.
Manufacturers do not need another platform that performs well in a controlled demonstration but leaves their team to solve the difficult final steps. They need an inspection system that works with real products, real defects, real operators and the everyday variability of production.
That is why we believe industrial AI needs Hypercare.
The hardest part begins when AI meets production
A promising proof of concept can show that a model is capable of identifying a defect. It cannot reproduce every condition the system will encounter after deployment.
On the factory floor, lighting changes. Materials reflect differently. Product positioning varies. Acceptable process variation can resemble a defect. New product variants are introduced, inspection criteria evolve and operators encounter edge cases that were not present during the pilot.
The technical model is only one part of the resulting inspection system. Cameras, lighting, calibration, thresholds, pass/fail rules, reporting and operator review workflows all influence whether that system delivers useful decisions.
The real measure of success is therefore not whether an AI model works in isolation. It is whether the complete inspection workflow continues to perform reliably under production conditions.
As Dr Wilhelm Klein, CEO of Zetamotion, put it:
“Production AI is not finished at go-live. In many ways, go-live is where the real learning starts.”
Where conventional provider models create headaches
Many AI and machine vision projects are sold using a familiar sequence: demonstrate the technology, install the system, hand over the licence and move ongoing problems into a support queue.
The software may have been delivered successfully, but the manufacturer is left carrying much of the operational risk.
Several common problems follow.
The implementation burden quietly moves to the manufacturer
A licence may provide access to the software without providing the specialist capacity required to make it productive.
Internal teams can find themselves responsible for collecting and organising inspection data, defining defect categories, labelling images, tuning thresholds, configuring review processes and determining why performance has changed.
These tasks require time and specialist knowledge. They also arrive on top of the team’s existing production and quality responsibilities.
A successful pilot does not guarantee production performance
Controlled demonstrations can hide the variability found on a live line.
A model may perform well on a limited test set but struggle when presented with different batches, finishes, orientations, lighting conditions or harmless surface variation. This is especially challenging for manufacturers working with reflective, transparent, textured or naturally non-uniform materials.
When support is limited to general software troubleshooting, the manufacturer can become trapped between a platform that is technically functioning and an inspection process that is not producing sufficiently reliable decisions.
Every production change becomes a new project
Factories do not stand still. New variants, materials, defect criteria and operating conditions are part of normal manufacturing.
Under a rigid delivery model, each change may trigger a separate statement of work, a long support process or a requirement to rebuild datasets manually. The inspection system becomes difficult to maintain precisely because the production environment continues to evolve.
Accountability becomes fragmented
Industrial inspection often involves software, AI models, cameras, lighting, edge compute, line controls and operator workflows. If different providers are responsible for each component, diagnosing a performance issue can become an exercise in passing responsibility between vendors.
The software provider points to the hardware. The hardware provider points to the data. The integrator points to the model. Meanwhile, the manufacturer still has an inspection problem.
This is one of the biggest weaknesses of the traditional handover model: the customer owns the outcome, while each supplier owns only a narrow component.
Hypercare changes the ownership model
A Hypercare Subscription is based on a different principle: the relationship should be organised around making the deployed inspection system useful, not simply keeping a software licence active.
Zetamotion remains involved in the practical work required to maintain and improve the system. Depending on the deployment, this can include:
- Preparing and curating inspection data.
- Onboarding products and variants.
- Teaching teams how to use the inspection and review workflows.
- Refining defect catalogues and decision criteria.
- Tuning models, thresholds and pass/fail rules.
- Reviewing detections, false rejects and missed edge cases.
- Monitoring performance and production feedback.
- Updating models and workflows as requirements evolve.
- Using human-in-the-loop feedback to support continuous improvement.
The purpose is not to remove manufacturers from the process. Their quality engineers and operators hold essential knowledge about products, defects, tolerances and production reality.
Hypercare creates a working relationship in which that manufacturing expertise can be combined with Zetamotion’s AI inspection capabilities. The customer should not have to become a computer vision specialist to keep the system performing.
In Wil’s words:
“We take ownership of helping the system work.”
What Hypercare means in practice
The difference between a conventional software licence and Hypercare can be summarised simply:
| Conventional software model | Zetamotion Hypercare model |
|---|---|
| Access to a software product | An ongoing AI inspection partnership |
| Customer owns implementation | Shared responsibility for operational performance |
| Support focuses on software issues | Support considers data, models, workflows and production conditions |
| Training often ends at handover | Knowledge transfer continues as the system evolves |
| Model updates are treated as exceptions | Continuous improvement is part of the relationship |
| Production variation is the customer’s problem | Production feedback informs ongoing optimisation |
This does not mean that the system will never need input from the manufacturer. It means the customer is not left alone to interpret every unexpected result, retrain every model or resolve every interaction between the AI and the production process.
Hypercare is not unlimited scope
A hands-on relationship still needs clear commercial and operational boundaries.
A Hypercare Subscription does not mean that every future expansion is automatically included. A new inspection station, major product family, additional hardware, substantial system integration or extensive on-site programme may require separate assessment and scoping.
The distinction is between supporting the deployed system and introducing a materially new deployment.
Clear boundaries protect both sides. Manufacturers know what ongoing care they can expect, while larger changes receive the engineering attention, planning and validation they require.
What changes under Hypercare is the default attitude. When an issue emerges within the deployed inspection system, our first response is not to externalise responsibility to the user. It is to understand what has changed and help the system adapt.
Why continuous improvement matters in machine vision
AI inspection systems learn from examples, but no initial dataset can perfectly represent the future.
A rare defect may appear for the first time months after deployment. A supplier may change a material finish. A new product variant may introduce a slightly different geometry. Operators may identify acceptable variation that the original model treated too cautiously.
Human-in-the-loop review turns these moments into useful feedback. Operators can review results, correct decisions and add production context. That feedback can then inform model refinement, threshold changes and workflow improvements.
Instead of treating every unexpected example as a failure, the system uses production experience to become better aligned with the manufacturer’s real quality requirements.
That is particularly important for difficult inspection cases involving:
- Rare defects and limited examples.
- High numbers of product variants.
- Reflective, transparent or non-uniform surfaces.
- Changing production conditions.
- Fine distinctions between acceptable variation and genuine defects.
- Line-speed inspection and rapid operational decisions.
These are not problems that disappear when a licence is activated. They require an inspection capability that can adapt.
Choosing an AI inspection provider means choosing a working relationship
Manufacturers evaluating an AI inspection solution should look beyond the demonstration and the software feature list.
They should ask:
- Who prepares and maintains the inspection data?
- Who helps define the defect catalogue and acceptance criteria?
- What happens when production conditions change?
- How are false rejects, missed defects and edge cases reviewed?
- Who is responsible for model and workflow optimisation?
- How are new variants introduced?
- Does support cover only the software, or the performance of the inspection workflow?
- Where does ongoing support end and separately scoped work begin?
The answers reveal whether the provider is selling access to technology or taking responsibility for helping that technology succeed in production.
More than software: a partnership around production reality
Zetamotion’s Hypercare Subscription gives manufacturers more than a software licence. It provides access to an embedded AI inspection team that stays involved after deployment.
We help customers monitor performance, adapt models and workflows, onboard changes and respond to the realities of the factory floor. Our aim is to keep each deployed inspection system reliable, useful and ready to evolve with production.
Because manufacturers should not have to choose between buying AI software and building their own internal AI team.
They should be able to work with a partner that takes ownership of helping the system perform.
Want to know whether AI inspection is a practical fit for your production line? Book a feasibility check with Zetamotion and tell us about your products, target defects and current inspection process.
