Bring your own AI provider: what that actually means, and which products it applies to

Most software that adds "AI features" picks a single model provider, bundles the usage cost into a subscription tier, and gives an organisation no say in where its data actually goes when a prompt is sent. That's a reasonable choice for a consumer product; it's a much harder one to accept for an organisation with its own data residency, vendor-risk and AI-governance obligations.

Across Incident Management and Response, and Development, DevOps and SRE, Noviqent takes the opposite approach: your organisation configures which AI provider handles reasoning - Claude, GPT, Gemini, Grok, or a privately-hosted endpoint you control - and every call uses your own key. Nothing routes through a shared account, nothing is billed to Noviqent, and if you point the configuration at an endpoint you host yourself, no third-party AI vendor is involved in that step at all.

Cloud Compliance is worth calling out separately here, because it doesn't use an AI provider at all - by design. Every compliance, security and cost finding it produces comes from a deterministic rule engine running plain, reviewable logic over your resource configuration: the same input always produces the same finding, every rule's exact logic is documented, and there is no prompt, no model, and nothing to hallucinate. For an audit-scope conversation, that's a meaningfully different (and often easier) story to tell than "we send your data to a language model" - Compliance simply doesn't.

For the two products that do use an AI provider, the design choice draws a clean line around what actually leaves your infrastructure. Ticket data, source code, metrics and logs are gathered and processed on infrastructure Noviqent operates directly; the one step that reaches a third party is the AI reasoning call itself, and it's a configuration choice your organisation makes explicitly, disclosed in full in the sub-processor register rather than buried in a privacy policy.