New or uncertain
Guardian recommends. Evidence, confidence, alternatives and impact are shown. A qualified and authorised human decides.
Technology & Trust
Governed by Design.
Customer-controlled boundaries. Policy and permissions. Qualified and Authorised Human In The Loop. Bounded automation. Auditability. Verification.
Governed. Bounded. Auditable. Verified. Customer-controlled.
01Customer-controlled boundaries
Your policy. Your permissions. Your boundary.
Guardian’s controls are set by the customer. What Guardian can do, at what confidence, in which environments, under whose authority, is defined by policy the organisation owns.
Controls can vary by service, application and environment. An action can be prohibited entirely. Nothing changes silently.
02Operating posture
Read-only first. Graded autonomy.
Guardian starts in a sensing and recommendation posture. Value is demonstrated before any operational action is permitted. When the organisation is ready, controlled action is enabled deliberately, verified in operation, and can be withdrawn.
- Read-only first
- Sensing and recommendation. No operational actions. Evidence and confidence exposed for review.
- Graded autonomy
- Known, pre-approved patterns can progress to bounded operational action, under policy, with verified outcomes.
03QHITL
Qualified and Authorised Human In The Loop.
Guardian never crosses the line from recommendation to action without authority the organisation can defend.
Known and pre-approved
Guardian validates policy, acts within its authority, verifies the outcome, and strengthens the Intelligence Foundation from what it learned.
04Auditability
Every recommendation defensible. Every action reconstructable.
Every recommendation carries the evidence behind it. Every governed action carries a record of who authorised it, on what basis, at what confidence, with what verified outcome. Historical decisions can be reconstructed and challenged.
05Deployment
Inside your secure environment.
Guardian is installed within the customer’s own secure environment for enterprise deployments. Operational data does not need to leave the environment for Guardian to provide its core intelligence.
- Not SaaS by default. Guardian runs inside your controls, on infrastructure you approve.
- Customer data isolation. One customer’s intelligence is not used to train another customer’s Guardian.
- No dependency on external generative AI. Guardian’s core intelligence does not require operational information to be sent to an external LLM service.
- Restricted connectivity supported. Guardian is designed to operate in environments where external connectivity is limited or absent.
06Security
Least privilege. Enterprise identity. Customer-controlled secrets.
Guardian is designed to operate under enterprise security principles. Access is minimised to what is required for the sensing, recommendation or action Guardian is authorised to perform. Credentials and secrets are protected by customer-controlled mechanisms. Enterprise identity providers govern who can do what.
07Integration
Alongside the enterprise stack, not in place of it.
Guardian complements existing enterprise tooling rather than replacing it. Approved monitoring, observability, IT service management, configuration and identity systems remain the systems of record. Guardian turns their evidence into governed, explainable diagnosis, and where approved, into governed action through the automation platforms the organisation already trusts.
- Observability & monitoring. Approved sources of operational evidence.
- ITSM. Recommendations and outcomes flow into the ticketing systems your organisation runs on.
- Configuration & service data. CMDB and service metadata contribute context for service-aware diagnosis.
- Identity. Enterprise IAM governs Guardian access, exactly like any other enterprise service.
- Approved automation. Where governed action is enabled, it flows through the automation platforms the organisation has already approved.
08Control
Immediate return to read-only.
Autonomy is granted and it can be withdrawn. Any authorised human can return Guardian to a sensing and recommendation posture, immediately, without disruption to the underlying platforms Guardian supports.