Switching AI provider.
Switching AI provider means you receive model weights, data, configuration and documentation in full, in a documented format, and can keep using them with another vendor or in your own operation. At NexPatch this process is fixed in the contract, within a set deadline, at no extra cost for the export itself.
Why we publish the exit process
Based on our review, hardly any comparable vendor in the DACH market publishes its exit process openly. Most contracts describe in detail how a customer gets started, but barely how they get back out. We reverse that, because we believe a commitment is only worth as much as what happens when a customer actually uses it. So the process sits here, in the open, not buried in the small print of a contract.
For companies bringing private AI infrastructure or AI agents into productive operation, this question is not a side issue. A system woven into internal workflows over months or years creates a dependency on the vendor that can only be justified responsibly if the way back is known from the start, not only once it is actually needed.
What you take with you when you leave
| Component | Format | Deadline |
|---|---|---|
| Model weights | Standard format for open models, directly reusable | Within the period named in the contract after notice |
| Processed data | Structured export, machine-readable | Within the period named in the contract after notice |
| Configuration | Documented settings, approval logic, access rights | Within the period named in the contract after notice |
| Documentation | Full technical description of the architecture | Within the period named in the contract after notice |
The exact deadline is fixed in the contract, based on the scope and complexity of the system in question. We charge nothing extra for the export itself, because we treat the exit as part of our operating commitment, not as a separate service.
The process, in four steps
First, notice. You give notice at the term agreed in the contract, informally, through the named contact from your operations agreement.
Second, the export assessment. Together with you, we establish which models, data and configuration matter to you and in what target format you need them - for handover to a new vendor, say, or for your own operation.
Third, the actual export. We provide model weights, data, configuration and documentation within the agreed deadline, in the formats established beforehand.
Fourth, confirmation. After a successful export, both sides confirm in writing that the handover is complete and the relationship has ended, including deletion of any remaining data on our side, as far as agreed in the contract.
How this connects to our choice of models
That this exit is practical at all is closely tied to our choice of models. Because we prefer language models with open weights, as described at private AI infrastructure, a model is not tied to a single vendor - it can keep running on different infrastructure. What ongoing operations look like up to that point, we describe at Operations, and the related response and service times at Service Level.