Measure quality
Retrieval quality in the knowledge system, classification accuracy, answer quality – continuously, per task and model.
Measure and monitor AI systems and run them according to agreed rules.
Operation covers ongoing monitoring, cost control, quality measurement, updates, fallback rules and support. The agreed scope is set for each project.
In brief: AI Operations means operational support for the platform and solutions: ongoing monitoring, approvals, updates, fallback rules, cost control and support.
Tasks
Six things that must run in operation.
Control costs
What does each task cost on which model? Visible in Studio, adjustable via routing.
Latency and utilisation
How fast does each node respond, and where do delays occur?
Updates
Keep models and platform up to date – with evaluation before switching.
Fallbacks
What happens if a model or node fails? Rules that match the data class.
Support and approvals
Contacts, escalation, approval processes – defined before operation starts.
Operating models
Operation: responsibilities per operating model.
We agree the right operating framework in writing for each project.
| Operating model | Nodes | Platform | Application |
|---|---|---|---|
| Self-operated | Your IT | Swiss AI Labs (updates) | Swiss AI Labs / you |
| Co-operated | Your IT or private cloud | Swiss AI Labs | Swiss AI Labs |
| Managed | Private cloud | Swiss AI Labs | Swiss AI Labs |
Further reading
Related topics.
Frequently asked questions
Frequently asked questions about AI Operations
Do you guarantee availability?
We do not state absolute figures on the website. Fallbacks, load balancing and monitoring are designed to absorb outages; we agree specific service levels in the contract.
What happens when a model is switched?
The new model is tested, switched over in Studio and observed in operation. Testing and adjustments to interfaces or the application may be needed.
Do you show internal details?
Publicly, we show the principle, layers and data flow at a high level. Internal topology, routing rules, security rules and version states stay internal.
Is it already in production?
We state the maturity of each building block openly. Phrases such as “is designed to” describe the architecture, not a client deployment. Studio, Agents and parts of Factory run in our own development and operating environment and are continuously extended. Further operating models, multiple nodes and private and local installations will follow step by step.
Next step
Operation after the pilot: settled before it starts.
We clarify the operating model, monitoring and support before anything goes live.