Hosting to suit your needs
Operated in Switzerland, the EU, internationally or on your premises, including hybrid set-ups – matched to the data class.
Operations · Managed AI Operations
So your agents keep working properly – tomorrow too.
An agent is never finished. Models change, data shifts, processes evolve. We operate your AI solutions so that they stay reliable, secure and affordable.
Services
What our operations service includes.
We take responsibility for the running agent – technically, functionally and commercially. You stay in control of rules and approvals.
Round-the-clock monitoring
Availability, run times, error rates and queues are measured continuously, with automatic alerting.
Quality measurement
Spot checks and test cases show whether the agent is performing as well as on day one. We detect deviations before they become noticeable.
Model and version management
New model versions are first tested against your test cases, then rolled out in a controlled way.
Cost control
Transparent reporting on compute and model usage, with budget limits and optimisation suggestions.
Access and logging
Roles and permissions, encrypted connections and a complete log of every agent action.
An overview for your team
A control centre shows pending approvals, completed cases and KPIs – for business units and management.
Support and expansion
A dedicated contact person, changes within the agreed allocation and regular operations reviews.
Service levels
The right support for every level of criticality.
A reporting agent needs different support from an agent that books patient appointments. That is why there are three service levels.
| Service | Basicfor supporting agents | Businessfor business-critical processes | Criticalfor healthcare and 24/7 processes |
|---|---|---|---|
| Support hours | Mon–Fri, 08:00–17:00 | Mon–Fri, 07:00–19:00 | 24 × 7 |
| Incident response | next working day | 4 hours | 1 hour |
| Quality review | quarterly | monthly | continuous, with alerting |
| Model updates | bundled | scheduled, with a test run | scheduled, with a test run and fallback option |
| Operations review | annually | quarterly | monthly |
Indicative values. Binding values are agreed in the service level agreement.
When something goes wrong
An agent that is unsure stops and asks.
Every agent has a fallback procedure. If it detects an unusual case or a malfunction, it hands over to a person instead of guessing.
- Detect
Anomaly
Monitoring or the agent itself reports a deviation from expected values.
- Safeguard
Handover
Affected cases automatically move to manual processing. Operations continue.
- Resolve
Analysis
Our team finds the cause in the log, fixes it and verifies the fix against the test cases.
- Learn
Report
You receive a short report, and the case becomes part of future quality reviews.
Already running AI solutions?
We also take over agents and workflows built by others – following an assessment of the current set-up, with a clear recommendation.